Jennifer Lackey - The Epistemology of Groups-Oxford University Press (2021)

OXFORD
the epistemology of group

JENNIFER LACKEY
The Epistemology of Group
The Epistemology
of Groups
JENNIFER LACKEY
UNIVERSITY PRESS
UNIVERSITY PRESS
Great Clarendon Street, Oxford, 0X2 6DP,
United Kingdom
Oxford University Press is a department of the University of Oxford.
It furthers the University’s objective of excellence in research, scholarship,
and education by publishing worldwide. Oxford is a registered trade mark of
Oxford University Press in the UK and in certain other countries
© Jennifer Lackey 2021
The moral rights of the author have been asserted
First Edition published in 2021
Impression: 1
All rights reserved. No part of this publication may be reproduced, stored in
a retrieval system, or transmitted, in any form or by any means, without the
prior permission in writing of Oxford University Press, or as expressly permitted
by law, by licence or under terms agreed with the appropriate reprographics
rights organization. Enquiries concerning reproduction outside the scope of the
above should be sent to the Rights Department, Oxford University Press, at the
address above
You must not circulate this work in any other form
and you must impose this same condition on any acquirer
Published in the United States of America by Oxford University Press
198 Madison Avenue, New York, NY 10016, United States of America
British Library Cataloguing in Publication Data
Data available
Library of Congress Control Number: 2020941998
ISBN 978-0-19-965660-8
DOI: 10.1093/oso/9780199656608.001.0001
Printed and bound in Great Britain by
Clays Ltd, Elcograf S.p.A.
Links to third party websites are provided by Oxford in good faith and
for information only Oxford disclaims any responsibility for the materials
contained in any third party website referenced in this work.
To Baron, Isabella, and Catherine
Contents
Acknowledgments ix
Introduction 1
0.1 On the Very Existence of Group Beliefs 4
0.2 The Nature of Groups 5
0.3 Chapter Overviews 12
0.4 The Bigger Picture 15
1. Group Belief: Lessons from Lies and Bullshit 19
1.1 Summative and Non-Summative Views of Group Belief 20
1.2 Group Lies and Group Bullshit 30
1.3 Judgment Fragility 41
1.4 Base Fragility 45
1.5 The Group Agent Account 48
1.6 Conclusion 54
2. What Is Justified Group Belief? 55
2.1 Divergence Arguments 56
2.2 The Paradigmatic Inflationary Non-Summativist View:
The Joint Acceptance Account 59
2.3 Problems for the Joint Acceptance Account 62
2.4 Revisiting Divergence Arguments 67
2.5 Deflationary Summativism, the Group Justification
Paradox, and the Defeater Problem 71
2.6 The Collective Evidence Problem 84
2.7 The Group Normative Obligations Problem 88
2.8 A Condorcet-Inspired Account of Justified Group Belief 91
2.9 The Group Epistemic Agent Account 95
2.10 Central Obj ection to the Group Epistemic Agent Account 107
2.11 Conclusion 109
3. Group Knowledge 111
3.1 Social Knowledge 111
3.2 Social Knowledge and Action 115
3.3 Social Knowledgeand Defeaters 123
3.4 Knowing, Being in a Position to Know, and Should
Have Known 127
Viii CONTENTS
3.5 Collective Knowledge 128

3.6 Conclusion 137
4. Group Assertion 138
4.1 Two Kinds of Group Assertion 139
4.2 Having the Authority to Be a Spokesperson 140
4.3 The Autonomy of Spokespersons 148
4.4 Coordinated and Authority-Based Group Assertion 149
4.5 Two Other Accounts 154
4.6 Group Assertion Is Not Reducible to Individual Assertion 158
4.7 Conclusion 163
5. Group Lies 165
5.1 Individual Lies 166
5.2 Counterexamples to the Traditional View of Lying 167
5.3 Non-Deception Accounts of Lying 169
5.4 Back to Deception 171
5.5 Summativism and Sufficiency 178
5.6 Summativism and Necessity 181
5.7 The Joint Acceptance Account of Group Lies 184
5.8 Group Lies 186
5.9 Conclusion 188
References 189
Index 197
Acknowledgments
I have been thinking about issues related to the epistemology of groups for a
number of years and so there are many people to thank for playing a role in
seeing this project through to the end. For giving me enormously helpful feed
back on one or more of the chapters in this book, I am grateful to Anne Baril,
Jared Bates, Michael Bratman, Jessica Brown, Tom Carson, Fabrizio Cariani,
J. Adam Carter, David Christensen, Michael DePaul, Josh Dever, Don Fallis,
Sandy Goldberg, Alvin Goldman, John Greco, Allan Hazlett, Marija Jankovic,
Nick Leonard, Kirk Ludwig, Eliot Michaelson, Federico Penelas, Jim Pryor,
Florencia Rimoldi, John Searle, Andreas Stokke, and Deb Tollefsen.
Thanks also go to audience members at the University of Warsaw, the
Indiana Philosophical Association meeting at Hanover College, Western
Michigan University, the Workshop on the Epistemology of Groups at
Northwestern University, the University of Buenos Aires, the University of
Warwick, a Social Epistemology Workshop in Helsinki, Finland, the GAP.9
Conference in Osnabriick, Germany, an Invited Symposium at the Eastern
Division of the APA in Washington, D.C., the Epistemic Dependence on
People and Instruments conference in Madrid, Spain, the 3rd Colombian
Conference in Logic, Epistemology, and Philosophy of Science in Bogota,
Colombia, the University of Toronto, Mississauga, the University of
St. Andrews, the Southwest Epistemology Workshop at the University of New
Mexico, the International Workshop on Lying and Deception at Johannes
Gutenberg University in Mainz, Germany, the Collective Intentionality IX
Conference at Indiana University, New York University, the University of
Connecticut, the University of Massachusetts, Amherst, the University of
Georgia, Radboud University, the XVII Congress of the Inter-American
Philosophical Society in Salvador, Brazil, the Social Epistemology Workshop
in St. Andrews, Scotland, the Midwest Epistemology Workshop at Notre
Dame, the University of Texas at Austin, the University of Nebraska—
Lincoln, and students in my graduate seminars at Northwestern University.
I am also grateful to my daughter, Isabella Reed, for her tireless and
meticulous work on the index for this book. Just when I thought I might not
cross the finish line, Isabella stepped in with her ever generous and thoughtful
spirit to provide much-needed assistance with this project.
X ACKNOWLEDGMENTS
My greatest debt of all is to my husband, Baron Reed—my most brilliant

reader, my fiercest champion, and my most constructive critic. Every
sentence in this book has been carefully read by him, often more than once,
and every argument has benefited from his incisiveness, powerful mind,
and ability to invariably see the best version of what I am saying. I have
learned more from Baron than from any other human in my life (except my
mother, who literally taught me how to walk and talk!). In keeping with the
theme of this book, I often feel that Baron and I make up a distinctive philo
sophical group with a shared “mind of our own.”
I am dedicating this book to Baron and to our daughters, Isabella and
Catherine, for enriching my life beyond words: to Baron, for his unwavering
support, witty humor, unparalleled love, and the most powerful intellectual
connection; to Isabella, for an irrepressible generosity of spirit, for wisdom
and courage far beyond her years, and for a deep companionship that con
tinues to surprise me; and to Catherine, for her utterly unique and captivat
ing way of looking at the world, for her seemingly endless creativity, and for
her adventurous yet completely steady heart.
Introduction
In 2005, Volkswagen of America learned that its diesel vehicles could not
meet emissions standards in the United States. Rather than lower the actual
emissions levels, the auto manufacturer inserted software that reported sub
stantially lower emissions levels during testing than were possible when the
vehicles were on the road. Before a team from West Virginia University
uncovered this deception, these “defeat devices” were installed in 11 million
diesel cars sold worldwide between 2008 and 2015. The result was that the
fraudulent test results met emission requirements, but the vehicles “spewed
as much as 40 times more pollution from tailpipes than allowed by the
U.S. Environmental Protection Agency.”1 Since then, scientists at MIT
have found that the excess emissions will cause 60 premature deaths across
the United States and 1200 in Europe, with Germany, Poland, France, and
the Czech Republic being hit the hardest.2
Michael Horn, who was the CEO of Volkswagen Group of America at the
time, testified before the House Energy and Commerce Committee’s over
sight and investigations panel in October of 2015 about this emissions-test
cheating scandal. In response to challenges from lawmakers, Horn said,
“This was a couple of software engineers who put this in for whatever rea
son. To my understanding, this was not a corporate decision. This was
something individuals did.”3 Horn went on to explain that three Volkswagen
employees had been suspended as a result of the software that led to the
fraudulent test results.
In response to Horns testimony, Rep. Chris Collins (R-N.Y.), who is him
self an engineer, said, “I cannot accept VW’s portrayal of this as something
by a couple of rogue software engineers. Suspending three folks—it goes
1 https://www.chicagotribune.com/news/sns-bc-u8--volkswagen-emissions-scheme-
20150921-story.html, accessed August 8, 2019.
2 http://news.mit.edu/2017/volkswagen-emissions-premature-deaths-europe-0303 ,
accessed August 8, 2019.
3 https://www.latimes.com/business/autos/la-fi-hy-vw-hearing-20151009-story.html ,
accessed August 7, 2019.
The Epistemology of Groups. Jennifer Lackey, Oxford University Press (2021). © Jennifer Lackey.
DOI: 10.1093/OSO/9780199656608.003.0001
2 THE EPISTEMOLOGY OF GROUPS
way, way higher than that.”4 Collins continued: “Either your entire
organization is incompetent when it comes to trying to come up with intel
lectual property, and I don’t believe that for a second, or they are complicit
at the highest levels in a massive cover-up that continues today.”5
Volkswagens response to this scandal lies at one end of the spectrum
regarding collective responsibility: the blame is entirely the result of a few
individual employees, with none attaching to the corporation itself.
At the other end of the spectrum lies a case like this: on March 6, 1984,
the United States Department of Defense charged that between 1978 and
1981, the company National Semiconductor had sold them 26 million com
puter chips that had not been properly tested and then had falsified their
records to conceal the fraud.6 These potentially defective chips had been
used in airplane guidance systems, nuclear weapons systems, guided mis
siles, rocket launchers, and other sensitive military equipment. Highlighting
the gravity of the situation, a government official noted that if one of these
computer chips malfunctioned, “You could have a missile that would end
up in Cleveland instead of the intended target.”
Officials at National Semiconductor admitted to both the omission of
required tests and the falsification of relevant documentation and agreed to
pay $1.75 million in penalties for defrauding the government. However, the
company refused to provide the names of any of the individuals who had
participated in the decision to omit the tests and falsify the documents, or
any who had been involved in carrying out these tasks. The legal counsel for
the Department of Defense objected, arguing that “a corporation acts only
through its employees and officers” and thus the government would have no
assurance that National Semiconductor would not engage in the fraud
again. In response, the CEO of National Semiconductor said, “We totally
disagree with the Defense Department’s proposal. We have repeatedly stated
that we accept responsibility as a company and we steadfastly continue to
stand by that statement.” A spokesperson for National Semiconductor later
reiterated this position: “We will see [that our individual people] are not
harmed. We feel it’s a company responsibility, [and this is] a matter of
4 https://www.latimes.com/business/autos/la-fi-hy-vw-hearing-20151009-story.html,
accessed August 7,2019.
5 https://www.washingtonpost.com/news/the-switch/wp/2015/10/07/volkswagens-pulling-
the-plug-on-its-2016-american-diesel-cars/ , accessed August 7,2019.
6 This case, including all of the quotations, are discussed in Velasquez (2003).
INTRODUCTION 3
ethics.” National Semiconductor prevailed: no individual employee was ever

held criminally or civilly liable for the crime. Only the company qua com
pany was penalized.
In contrast to Volkswagens approach, then, National Semiconductor
took complete responsibility for defrauding the government at the level of
the company and denied that any individual employees were deserving
of blame.
These two different ways of characterizing collective responsibility are
closely connected to a central debate in the literature on the epistemology of
groups. On the one hand, deflationary theorists hold that group phenom
ena, such as group beliefs, can be understood entirely in terms of individual
members and their states. According to this approach, the states of collect
ives are not interestingly different than those of individual knowers, and
thus collective epistemology turns out to be largely or completely reducible to
individual epistemology. Inflationary theorists, on the other hand, hold that
group phenomena are importantly over and above, or otherwise distinct
from, individual members and their states. In this way, groups are often said
to crucially have “minds of their own.”7 Settling some of the issues in this
debate lies at the heart of making sense of attributions of collective respon
sibility. If National Semiconductor, for instance, is to be treated as an entity
over and above its members in bearing responsibility for defrauding the
government, then it is essential to determine whether the company believed
or knew that required tests were being omitted and relevant documentation
falsified and whether it lied to the government about its fraudulent behavior.
In the absence of plausible readings of these collective states, it simply
wouldn’t make sense to say that National Semiconductor as a corporation
bears full responsibility for its actions.
A central aim of this book is to make progress in understanding these
crucial notions in collective epistemology—group belief, justified group
belief, group knowledge, group assertion, and group lies—so as to shed light
on whether it is groups, their individual members, or both who ought to be
held responsible for collective actions.
7 See Pettit (2003).

0.1 On the Very Existence of Group Beliefs
One fundamental type of objection I have frequently heard to a project on

the epistemology of groups goes like this: groups are not the proper bearers
of epistemic states. Being a knower, or justifiedly believing a proposition,
requires belief. Belief is a mental state, and groups don’t have mental states
in any robust sense. To the extent that we attribute states of belief or knowledge
to groups, such talk is loose or metaphorical. Anthony Quinton, for
instance, writes:
We do, of course, speak freely of the mental properties and acts of a group
in the way we do of individual people. Groups are said to have beliefs,
emotions, and attitudes and to take decisions and make promises. But
these ways of speaking are plainly metaphorical. To ascribe mental predi
cates to a group is always an indirect way of ascribing such predicates to its
members. With such mental states as beliefs and attitudes, the ascriptions
are of what I have called a summative kind. To say that the industrial
working class is determined to resist anti-trade union laws is to say that all
or most industrial workers are so minded. (Quinton 1975/1976, p. 17)
There are two different views of group belief suggested in this passage. On
the one hand, there is the eliminativist view, according to which it is literally
false that groups believe things and hence group belief attributions are sim
ply metaphorical. On the other hand, there is the deflationary view men
tioned above, according to which it is literally true that groups believe
things, but such claims are made true entirely by individual members of the
groups believing things. On either reading, we might say that a book on the
epistemology of groups is misguided. Groups are not epistemic agents in
their own right because they don’t have proper beliefs and, thus, they don’t
have justified beliefs or knowledge. Talk of group beliefs is either metaphor
ical or fully reducible to the beliefs of individuals.
Since any reader of this book who is sympathetic with this line of thought
will most likely see no point in moving forward, let me offer a very brief
argument right at the outset for rejecting it.
I take as a starting position that groups lie. I will say more about this
throughout the book, especially in Chapters 1 and 5, but this does not seem
particularly contentious. Google “Facebook lies” and a litany of articles
comes up. Corporations have been forced to pay literally billions of dollars
for lying. After a 2009 headline in Business Insider that reads “Pfizer to Pay
INTRODUCTION 5
$2.3 Billion in Biggest Fine Ever for Deceitful Drug Marketing” the first line
says, “There is not always truth in advertising, but when you really lie, you
really pay—especially if you happen to be an enormous drug company.”8
And on a more personal level, imagine that your employer promised to give
you research funds while recruiting you, but you then learned after they
never materialized that the university said this to you while knowing full
well that they did not have the resources to follow through. It seems quite
natural for you to say, not at all metaphorically, “My university lied to me.”
With this in mind, here is an argument:
1. Groups lie.
2. Group lies cannot be understood without groups having genuine
beliefs.
3. Therefore, groups have genuine beliefs.
I will discuss group lies in far more detail in Chapters 1 and 5. But very
briefly, a lie—whether offered by an individual or a group—just is an asser
tion that one does not believe oneself that is made with the intention to be
deceptive. Indeed, even within debates about the details over how to under
stand lying, there is consensus that it crucially involves the absence of belief
on the part of the liar. Premise (1) is thus widely supported by our social
practices, including our notions of moral and criminal responsibility, and
(2) follows from every major account of the nature of lying. Finally, all
I mean by “genuine” is that such beliefs are neither loose talk nor entirely
reducible to the beliefs of individuals. Rather, there is a robust sense in
which groups are believers in their own right.
Of course, I don’t expect this argument to be fully satisfying at this point.
But what I hope it does is get the skeptical reader on board with thinking
that a project on the epistemology of groups is worth exploring.
0.2 The Nature of Groups
Another initial question that might be raised about a project on the epis
temology of groups is what kind of collective entities will be at issue. Groups
8 https://www.businessinsider.com/pfizer-to-pay-23-billion-m-biggest-fine-every-for-
deceitful-advertising-2009-9, accessed August 7, 2019.
obviously come in a variety of forms. At one end of the spectrum, there are
highly structured groups with policies, procedures, and robust forms of
interaction among the members, such as corporations, universities, juries,
and boards; at the other end, there are collections of individuals with no
formal structure or interaction among group members, such as left-handed
Northwestern students and red-haired New Yorkers. And in between these
two ends are groups with varying degrees of structure and interaction, such
as governments, scientific communities, Americans, and women.
Accounts of group phenomena are often directly shaped by which groups
are taken to be paradigmatic. For instance, those who are drawn to restrict
ive views typically focus on highly structured groups with regular inter
action among the members. A particularly clear example of this can be seen
in the work of Frederick F. Schmitt who, following Margaret Gilbert, argues
that “a set of individuals forms a group just in case the members of the set
each openly expresses his or her willingness to act jointly with the other
members of the set” (Schmitt 1994, p. 260). This conception of a group
requires a high level of conscious interaction among the members, and thus
rules out as groups all but very formalized collections. It is, therefore, not
surprising that Schmitt frequently uses a jury as a classic example of a
group. Moreover, this starting point directly impacts his view of other col
lective phenomena, such as justified group belief, where he relies crucially
on the notion of joint acceptance. Clearly, there is a sense in which many
groups, such as the Democratic Party or even Northwestern University, are
not even properly positioned for joint acceptance, as they are large, dis
persed, and made up of members with varying levels of authority.9
9 Others distinguish between collections of individuals and group agents, with very strong
requirements to qualify as the latter. For instance, according to Jesper Kallestrup, “A collective
is a group agent only if (i) its individual members intend that the collective act and form atti
tudes together, i.e. each of these individuals must intend that they together enact the joint per
formance and come to a group attitude. Moreover, (ii) each must intend to do their part, and
(iii) intend to do so because of their belief that others intend to do their bit.... A different but
related set of constraints concerns the office of the collective as fixed by its charter. A collective
is a group agent only if its (founding) members jointly set up common goals and agree on how
to proceed in order to meet them. Both the ends and the means, which are carried out for the
purpose of achieving them, are captured by the groups charter, which is sometimes formally
enshrined in a system of laws, other times its existence is evidenced by the practice of the
group and its members. When these two sets of constraints are met, a collection of individuals
unites in forming a rational agent in its own right” (Kallestrup 2016). It should be clear that
many collectives that act together will fail these conditions, such as a group of strangers who
work together to save a drowning swimmer. In at least some sense, this collection surely might
properly be regarded as a group agent, despite failing Kallestrup’s demanding conditions.
INTRODUCTION 7
In contrast, those who are interested in more permissive accounts of

collective phenomena typically focus on large, unstructured groups, such
as those that have information distributed throughout their members.
Alexander Bird, for instance, takes the scientific community to be a para
digmatic group in his argument on behalf of a phenomenon that he calls
“social knowing,” where group states do not even supervene on the mental
states of the individual members.10 Similarly, Soren Harnow Klausen
recently argued that:
We should allow that the factors which, together with truth (or, in the case
of knowing how, some sort of adequacy to the task in question), are neces
sary and jointly sufficient for group knowledge, can be distributed among
the members of the group. A well-known example of genuinely distrib
uted cognition has been provided by Hutchins (1995), who describes how
a navy vessel crew is able to navigate successfully through the concerted
efforts of many individuals, each of whom carries out a very specialized
task and does not necessarily have any knowledge of the contributions of
others, nor of the more general tasks or the ways in which the different
contributions are merged. I suggest that this kind of example, rather than
that of a jury or a board of directors facing a specific decision, should
serve as a paradigm of collective knowledge. (Klausen 2015, p. 823)
Here, Klausen begins with a paradigm of a group that not only fails to
engage in any sort of collective deliberation, as a jury does, but is also such
that the members are wholly unaware both of the actions of the others and
of the collective goals. If this is the starting point, then any account that
requires joint awareness or activity of any sort will immediately be ruled out.
Attempts have been made in the literature to distinguish these different
kinds of groups in ways that have theoretical or practical significance. One
common distinction that is drawn here is between established groups and
non-established groups, where it is standard to rely on examples of each kind
rather than definitions or criteria. Margaret Gilbert, for instance, writes:
There are ascriptions of cognitive states to two or more people who are
understood to constitute an established group of a specific kind such as a
10 See Bird (2010).

union, court, discussion group, family, and so on. There are also ascriptions
of cognitive states to two or more people without any presumption that
they constitute an already established group. (Gilbert 2004, p. 96)11
While there are certainly important differences between, say, a court and a
collection of left-handed Northwestern students, it is doubtful that “being
established” is the most theoretically or practically relevant feature to focus
on. Suppose, for instance, that unbeknownst to the left-handed students at
Northwestern, I fill out all of the necessary paper work for them to have
official status as a club on campus. Left-handed Northwestern students are
now recognized by the university and thus constitute an established group.
Surely, however, this official status by itself does not change the group in any
deeply significant way. So, the distinction between being established and
non-established doesn’t seem to mark a particularly meaningful difference
between these two kinds of groups.
Christian List and Philip Pettit (2011) have argued that while a corpor
ation is a group, left-handed Northwestern students are a mere collection,
and this difference is grounded in whether the collection in question can
survive changes of membership. They write:
Collections of individuals come in many forms. Some change identity

with any change of membership. An example is the collection of people in
a given room or subway carriage. Other collections have an identity that
can survive changes of membership. Examples are the collections of
people constituting a nation, a university or a purposive organization. We
call the former “mere collections”, the latter “groups”. Our focus here is on
groups. (List and Pettit 2011, p. 44)
Once again, however, it is not clear that an important distinction has been
highlighted. On the one hand, a company may not survive the firing of its
CEO, a cult may not survive the death of its leader, and a political group
may not survive its dictator being overthrown, but there is nonetheless a
clear difference between these groups and left-handed Northwestern stu
dents. On the other hand, a collection of individuals trying to save a beached
whale over the course of 36 hours12 may survive a number of changes
11 See also Lahroodi (2007) and Bird (2010).

12 This is a modified example that List and Pettit (2011) use to illustrate a mere collection.
INTRODUCTION 9
in membership,13 yet there may be significant asymmetries between this

type of group and a medical association or academic department. So, it
seems as though there can be paradigmatic groups unable to survive
changes in membership, and classic instances of mere collections that are
able to do so. Thus, the ability to survive changes in membership also fails
to track a theoretically substantive difference among collective entities.
One distinction that has clear significance along a number of dimen
sions—such as epistemic, moral, legal, and practical—is the groups ability,
or lack thereof, to engage in collective deliberation or reasoning. Such
reasoning involves, at a minimum, a sensitivity to evidence, the capacity to
engage in belief revision, and being the proper subject of normative evalu
ation. For instance, school boards often meet to discuss issues relevant to
their goals, bouncing ideas off of one another, considering and responding
to objections as a group, weighing various pieces of evidence, and revising
their collective beliefs as necessary. In contrast, in the absence of unusual
circumstances, left-handed Northwestern students simply do not engage in
these sorts of activities for several reasons. First, collections of individuals
such as left-handed Northwestern students often simply don’t conceive of
themselves as a group and so wouldn’t consider engaging in any sort of
reasoning with other members. Second, these sorts of collections are some
times widely dispersed across even larger groups of individuals, rendering it
practically impossible to collectively reason, especially when there aren’t
any formal meetings or events at which they can do so. Finally, these groups
of individuals are frequently united by features, such as left-handedness,
that have very little general interest or value.
For ease of expression, let us call those groups capable of engaging
in collective reasoning deliberative groups and those that are not non-
deliberative groups.
There are two points about non-deliberative groups that should be
emphasized here. First, that this sort of group is incapable of engaging in
collective reasoning is simply a contingent feature that arises in the circum
stances in which it is found, one that can certainly change from one minute
to the next. For instance, Northwestern administrators may request that all
left-handed students gather in an auditorium to discuss whether the cam
pus is suitably sensitive to their needs. Prior to this gathering, the group was
13 For instance, a reward that is given in recognition of the efforts of those saving the
beached whale would include all of the people who participated in the rescue at any point. This
group would, then, be the subject of praise and other kinds of normative assessment in ways
that mere collections of individuals are not.
incapable of engaging in collective reasoning and was therefore non-

deliberative. But as soon as they find themselves in an auditorium identified
as sharing a certain salient property, they might then reason together as a
group and thus become deliberative. For instance, each student may now
share with the others evidence about how many classrooms do not have
desks for left-handed students, how many times each student has been
inconvenienced because of his or her left-handedness, and so on, and they
may decide as a group what their position is regarding whether the campus
is appropriately set-up for their needs. So, being non-deliberative is not
necessarily an enduring property of a group.14
Second, despite the fact that non-deliberative groups cannot engage in
collective reasoning, there is a clear sense in which they can nonetheless
have group beliefs, albeit ones that are different from those had by their
deliberative cousins. For instance, suppose that I am in charge of safety at
Northwestern and I wish to determine whether left-handed students regard
the campus as properly designed for their needs. To this end, I send out a
survey for all left-handed students to fill out and, after receiving the results,
I aggregate their judgments via a supermajority procedure. I then report on
this basis that left-handed Northwestern students believe that the campus is
not suitably sensitive to their particular needs. It is not uncommon to think
that there is nothing strained or mistaken about this belief attribution—as a
group, left-handed Northwestern students do hold this belief, just not in the
same way that a group capable of collective reasoning might do so.
While deliberative groups depend for their existence on features such as
an appropriate structure, a set of constitutive rules, accepted social integra
tion, and so on, non-deliberative groups can simply be brought into exist
ence through internal or external interest. For instance, someone who is
either herself left-handed or is simply interested in those who are may be
inspired to survey Northwestern students with this feature, aggregate their
responses, and reveal their belief as a non-deliberative group. The same can
be said for other collections of individuals, such as red-haired New Yorkers.
This interest brings the group into existence, and the surveying and aggre
gating reveals their group belief.
Unlike the features discussed earlier—such as whether a group is estab
lished or can survive changes in membership—the ability to engage in col
lective reasoning has clear significance. For instance, a collection that is
capable of weighing evidence and revising its beliefs as a group can be
14 Of course, being deliberative can also go out of existence.

INTRODUCTION 11
evaluated as rational and irrational in a way that one that is not so capable
cannot. Sure, after conducting the survey of left-handed Northwestern stu
dents, the administration may have evidence that these students have beliefs
that collectively conflict with one another, and they may assess them on this
basis. But this is an assessment of the beliefs of a collection of individuals
rather than of a collective entity. Corresponding to the evaluation of a group
being appropriately judged irrational is the responsibility that such an entity
might collectively bear for this shortcoming. If a school board is irrational
in its belief that the honors program at a local high school ought to be dis
continued, then it can be held responsible as a group for this epistemic
shortcoming. Relatedly, the school board can then act as a collective agent
by deciding to discontinue the honors program and, accordingly, can be
held morally and legally responsible for some of the effects that this move
has on the community. For instance, the school board can be appropriately
deemed self-serving or callous if it is discovered that the decision was made
simply for political benefits. Or the school board can be sued if parents
regard the cessation of the honors program as failing to provide their chil
dren with the educational opportunities that are required by the state. Thus,
there is obvious epistemological, metaphysical, moral, legal, and practical
value that comes with the ability to engage in collective reasoning.
In addition to deliberative and non-deliberative groups, there are what
we may call mere collections, which are sets of individuals that share a com
mon feature though one for which there is no interest, either internal or
external. This is currently the status of left-handed Northwestern students
since neither the members of this collection nor those outside of it have any
interest in left-handed students qua being left handed. As we have seen, this
can certainly change. Left-handed students may organize themselves into a
deliberative group by meeting weekly to discuss their plight, formalizing
rules for making decisions among themselves, and electing a board to offi
cially represent their interests on campus. Alternatively, a right-handed
member of the community worried about issues of fairness may convert
left-handed students into a non-deliberative group through her interest
in them.
There is, however, a feature that is even more general—one that may even
cut across the deliberative/non-deliberative distinction15—that captures the
15 I should note that while I am interested in normative evaluation of both deliberative and
non-deliberative groups, as a general rule the normative evaluation of deliberative groups
tends to be richer and more significant to our broader understanding of the epistemic, moral,
and legal landscape.
kinds of groups that will be the focus of this book; namely, being subject to
normative evaluation. In particular, I will be interested in those groups that
are properly subject to normative assessment, such as praise and blame,
along both epistemic and moral dimensions, and the corresponding attri
butions of responsibility, accountability, and so on, Put succinctly, if we can
properly hold a group, G, responsible for </>-ing, then this is sufficient for
regarding G as a group in the sense relevant for this project.
0.3 Chapter Overviews
This book is divided into five chapters. In Chapter 1,1 take up the question:
how should we understand the sense in which groups have beliefs? In stark
contrast to the quote from Anthony Quinton earlier, the received view in
collective epistemology is that group belief must be understood in non-
summative or inflationary terms. Such views are motivated by cases that
purport to show that a group can be said to properly believe that p, despite
the fact that none of its individual member believes that p. If this is true,
then group belief cannot be understood, even in part, in terms of the beliefs
of individual group members. This is the negative claim of the non-summative
view. The positive claim is that group belief should be characterized in
terms of something that the members do, and this is typically identified as
joint acceptance. Very roughly, group belief is the result of the members
jointly agreeing to accept a given proposition as the groups, even if no
member believes it herself.
In this chapter, I challenge this orthodoxy by raising an entirely new
objection to this general approach to understanding group belief. I show
that joint acceptance accounts crucially lack the resources to be able to
explain how groups can lie and bullshit, and, more generally, I argue that
group belief cannot be determined by states or processes that are under the
direct voluntary control of the members. I also show that such non-summative
views countenance as group beliefs states that are riddled with incoherence
among the bases of the beliefs of the group members. This leaves group
belief without an appropriate mind-to-world direction of fit and renders it
unsuitable for proper epistemic evaluation and collective deliberation.
In addition, I show that the original cases used to motivate non-summative
views can be fully explained without the need to posit group belief.
I then go on to develop and defend a new view, which I call the Group
Agent Account: group belief is determined in part by relations among the
INTRODUCTION 13
bases of the beliefs of members, where these relations arise only at the
collective level, and are crucial especially insofar as the group is able to
function as an agent. At the same time, group belief is partly constituted by
the individual beliefs of members. In this way, the resulting view is neither
strictly summative nor non-summative.
In Chapter 2, I turn to the question of justified group belief, which has
received surprisingly little attention in the literature. Mirroring the debate
regarding group belief, there are those who favor an inflationary approach,
where the justificatory status of group belief involves only actions or fea
tures that take place at the group level, such as the joint acceptance of
reasons. On the other hand, there are those who endorse a deflationary
approach, where justified group belief is understood as nothing more than
the aggregation of the justified beliefs of the groups members.
In this chapter, I raise new objections to both of these approaches.
Against inflationary views, I show that they face what I call the Illegitimate
Manipulation of Evidence Problem, according to which accounts that allow
the justification of group beliefs to be achieved through wholly voluntary
means also permit the evidence available to the group to be illegitimately
manipulated, thereby severing the connection between group epistemic jus
tification and truth-conduciveness. Against deflationary views, I argue, that
they lead to the Group Justification Paradox in which a group ends up
counting as justifiedly believing both thatp and that not-p.
Finally, I develop and defend a positive view of justified group belief that
parallels my account of group belief in Chapter 1 in critical respects, which
I call the Group Epistemic Agent Account: groups are understood as epi
stemic agents in their own right, ones that have evidential and normative
constraints that arise only at the group level, such as a sensitivity to the rela
tions among the evidence possessed by group members and the epistemic
obligations that arise via membership in the group. These constraints bear
significantly on whether groups have justified belief. At the same time, how
ever, group justifiedness on the Group Epistemic Agent Account is still
largely a matter of member justifiedness, where the latter is understood as
involving both beliefs and their bases. The result is a view that neither
inflates nor deflates group epistemology, but instead recognizes that a
group’s justified beliefs are constrained by, but are not ultimately reducible
to, members’ justified beliefs.
In Chapter 3, I take up two quite influential kinds of purported group
knowledge that are inflationary and non-summative in nature, and that
pose direct challenges to the account of justified group belief developed in
Chapter 2. The first, developed and defended in most detail by Alexander

Bird, is often referred to as ‘social knowledge.” A paradigmatic instance of
social knowing is taken to be the so-called knowledge possessed by the
scientific community, where no single individual knows a given proposition,
but the information plays a particular functional role in the community.
The second is “collective knowledge,” which occupies an important place in
United States law. According to the “collective knowledge doctrine,”
knowledge may be imputed to a group by aggregating bits of information
had by its individual members. If these are correct, then my view of justified
group belief is false, particularly the requirement that some of the members
of a group have the relevant justified beliefs themselves. However, I show in
this chapter that both social knowledge and collective knowledge sever the
crucial connection between knowledge and action, and open the door to
serious abuses, not only epistemically, but morally and legally as well. Bits of
information that are merely accessible to group members, or individual
instances of knowledge that are aggregated with no communication, do not
amount to group knowledge in any robust sense. I conclude, then, that
neither social knowledge nor collective knowledge is genuinely group
knowledge, and thus neither poses a challenge to my Group Epistemic
Agent Account.
I turn, in Chapter 4, to understanding what it means for a group to assert
a proposition. It is especially crucial to get a grip on this notion for being in
a position to hold collectives, such as corporations, morally and legally
responsible for what they say. I begin by distinguishing between two kinds
of group assertion—coordinated and authority-based—and I argue that
authority-based group assertion is the core notion. I then show that a defla
tionary view of group assertion, according to which a groups asserting is
understood in terms of individual assertions, is misguided. This is the case
because a group can clearly assert a proposition even when no individual
does. I then develop a positive inflationary view of group assertion accord
ing to which it is the group itself that is the asserter, even though this stand
ardly occurs through a spokesperson(s) or other proxy agent(s) having the
authority to speak on behalf of the group. This is supported by the fact that
paradigmatic features of assertion apply only at the level of the group.
A central virtue of my account is that it provides the framework for distin
guishing when responsibility for an assertion lies at the collective level and
when it should be shouldered by an individual simply speaking for herself.
In the final chapter of this book, I take up group lies. Despite the preva
lence of group lies and their often far-reaching effects, such as those seen in
INTRODUCTION 15
both the Volkswagen and National Semiconductor cases, there has never
before been a philosophical treatment of group lies.16 This chapter begins
the process of filling this surprising gap in the literature by focusing on the
question of what a group lie is. After providing an account of how to under
stand individual lies, I consider, first, whether group lies can be understood
in terms of the lies of the groups members and, second, whether group lies
can be characterized in terms of joint agreement by the groups members to
lie. After showing both views to be misguided, I offer my own account of
group lying, according to which it crucially involves the group offering a
statement. In particular, because what a group says can come apart from
what its individual members say, I argue that a group might lie when
no individual member lies, and a group might fail to lie even though every
individual member does. A central virtue of my account is that it captures
the often subtle and complex relationship that can exist between most
groups and their spokespersons. In this way, my view provides the basis for
understanding how groups are responsible for their lies, as well as for deter
mining when it is appropriate to trace this responsibility to the individual
members of the group and the spokespersons who represent them.
0.4 The Bigger Picture
Let’s return to the two cases we discussed at the start of this Introduction.
Straightforward deflationary views of collective phenomena have the
resources for holding only individual members of groups responsible for
their actions. After all, according to such accounts, only individuals believe,
know, assert, lie and so on, and so there quite literally are no states or actions
of collective entities to bear attributions of praise and blame. Such a frame
work accords very well with the way the Volkswagen of America CEO tried
handling the emissions-test cheating scandal. There were a few rogue soft
ware engineers on this reading, with no responsibility attaching to the cor
poration itself, and so the suspension of these employees fully handled the
matter. But as Rep. Chris Collins (R-N.Y.) and other lawmakers made clear,
this response is not only deeply unsatisfying, it is also wildly implausible.
Deception of this magnitude undoubtedly involves some degree of culpabil
ity at the level of the corporation itself. In particular, installing a device
16 Lackey (2018b) is the only exception.

explicitly designed to produce fraudulent test results in 11 million diesel

cars sold worldwide cannot occur without complicity or, at the very least,
negligence at the highest levels of leadership and oversight. Thus, purely
deflationary views of collective epistemic phenomena lack the resources
both for providing the correct diagnosis in a case such as this and for hold
ing the relevant parties accountable.
Purely inflationary views of collective phenomena, on the other hand,
have the resources for holding only groups responsible for their actions,
allowing individual members to go entirely or largely blameless. For, on
such views, activity that takes place at the collective level determines
whether groups believe, know, assert, lie and so on, and so there quite liter
ally are no individual states or actions of this kind that are part of the ana
lysis to shoulder praise and blame. This framework very nicely captures the
way that National Semiconductor approached its omission of required tests
and the falsification of relevant documentation. The company took full
responsibility for defrauding the government and agreed to pay $1.75 mil
lion in penalties, but vehemently denied that any blame belonged to indi
vidual employees. As the legal counsel for the Department of Defense made
clear, however, corporations act only through their employees and officers,
and so there is no way that this level of illegal activity occurred without
knowledge and involvement on the part of individuals. Hence, strictly infla
tionary views of collective epistemic phenomena lack the resources both for
providing a diagnosis that includes culpable activity on the part of individ
uals and for holding all of the relevant parties accountable.
To be sure, group members might be held accountable for jointly accept
ing the proposition in question on an inflationary view. But there are at least
two ways in which this is inadequate. First, this would distribute responsi
bility equally, as there is no interesting sense in which some might have
jointly accepted more than others. In actual cases, however, it is clear that
members often have significantly different roles in group actions and,
accordingly, bear different degrees of praise and blame. Second, it is import
ant to have a framework that can account for the responsibility members
shoulder in collective action that go beyond joint acceptance. Groups often
have complicated structures, where those at the highest levels of leadership
may avoid joint acceptance altogether but might nonetheless be involved in
collective action through more nuanced and harmful ways.
The views that I develop and defend in this book avoid all of these pit
falls. Because I argue that group belief and group justified belief involve
both the beliefs and epistemic statuses of individual members, and also
INTRODUCTION 17
relations and normative requirements that hold only at the level of the
collective, my views are neither purely inflationary nor deflationary. Rather,
I provide a framework for distributing responsibility across groups and
their individual members. I regard this as a central virtue of this book.
Neither Volkswagen nor National Semiconductor got things right in the
attributions of responsibility, and this led to significant pushback and
dissatisfaction among those impacted by their responses. We need a
framework for fully understanding accountability in such cases, and my
views provide this.
Moreover, while my views of group assertion and group lies are robustly
inflationary, they also have the resources for the proper distribution of
responsibility between collectives and their individual members. For
example, spokespersons have the authority to speak on behalf of groups,
and when they assert or lie in their official capacity as a representative of the
group, it is the group itself that asserts or lies. Similar considerations apply
to collective action more broadly that is performed through a proxy agent.
It is then, quite straightforward, how responsibility attaches to groups on
my view. But this relationship between groups and spokespersons
provides the resources for also holding individual members accountable.
Spokespersons are often chosen by members of the group who bear respon
sibility for not only ensuring that what is conveyed on its behalf is accurate
and well-supported, but also for keeping the spokespersons properly
informed. When things go awry along any of these dimensions, there are
often particular individuals clearly deserving of blame. In addition, spokes
persons are frequently (though crucially not always) members of the groups
they represent. The Chair of my Department, for instance, is both a member
of our group and our spokesperson when it comes to conversations with the
Dean about hiring decisions. We can certainly imagine situations in which
my Department is on the hook for a group decision we made but where the
Chair, as our spokesperson, deserves greater, or a different sort of, blame.
Perhaps she withheld crucial information in our decision-making or
inaccurately conveyed our position in conversation with the Dean. On my
view, the Department would shoulder the responsibility of its assertion con
veyed through the spokesperson, but the Chair could still be uniquely
blamed for her role in the process.
One final issue that is worth addressing at the outset is this: I argue in the
first three chapters of this book on behalf of views of collective phenomena,
such as group belief and justified group belief, that include as an epistemic
anchor, so to speak, the states of individual members. So, for instance, my
accounts of group belief and of justified group belief both require that at
least some of the individual members of the group instantiate the states in
question. Yet in the last two chapters, I defend views of both group assertion
and group lies that are robustly inflationary. In other words, I show that
groups can assert and lie when no member of the groups is even aware of
the proposition in question. What explains this asymmetry in a way that is
not ad hoc?
Robustly inflationary views should be adopted only where a group is
capable of granting authority to another agent or agent-like entity to do
something on its behalf. So, for instance, a group can grant authority to a
lawyer to speak on its behalf, to lie on its behalf, to bullshit on its behalf, and
to act on its behalf. In all of these cases, then, it will be possible for the
group’s actions to be constituted by the actions of another, even when the
group itself is entirely ignorant of the matter. Thus, accounts of all of these
phenomena will be robustly inflationary in nature. In contrast, a group can
not grant authority to another to believe on its behalf, to desire on its behalf,
to justifiedly believe on its behalf, or to know on its behalf. Accordingly,
accounts of all of these phenomena will require that at least some of the
individual members of the group instantiate the states in question.
Group Belief
Lessons from Lies and Bullshit
Groups and other sorts of collective entities are frequently said to believe
things.1 Sarah Huckabee Sanders, for instance, was asked by reporters at White
House press conferences whether the Trump Administration “believes in
climate change”2 or “believes that slavery is wrong.”3 Similarly, it is said on
the website of the ACLU of Illinois that the organization “firmly believes
that rights should not be limited based on a person’s sexual orientation or
gender identity.”4 And, according to the Presidential Commission on the BP
oil spill in the Gulf of Mexico, both BP and Halliburton believed that there
were flaws with the cement used for the well safety device before the
Deepwater Horizon explosion.5
These are just a few examples, but there are countless others. Moreover,
the importance of understanding these claims is clear, both theoretically
1 A discussion of what distinguishes a group from a mere collection of individuals lies

beyond the scope of this chapter, though I discuss this issue briefly in the Introduction. See
also Gilbert (1989 and 2004), Bird (2010), List and Pettit (2011), and Ritchie (2013). I won’t be
directly engaging with those who are more skeptical about the existence of group beliefs. For
arguments of this sort, see Rupert (2005 and 2011) and Huebner (2014). If what I argue in this
chapter is correct, however, then I will have shown that there are compelling reasons to coun
tenance the existence of group beliefs.
2 https://www.washingtonpost.com/news/energy-environment/wp/2017/ll/03/trump-
administration-releases-report-finds-no-convincing-alternative-explanation-for-climate-
change/?utm_term=.a8be9a994df0 , accessed January 15, 2017.
3 https://townhall.com/tipsheet/laurettabrown/2017/10/31/april-ryan-aslcs-white-house-if-
trump-administration-believes-slavery-is-wrong-n2403031, accessed January 15, 2017.
4 https://www.aclu-il.org/en/issues, accessed January 15, 2017.
5 https://www.theguardian.com/environment/2010/oct/29/bp-oil-spill-bp, accessed January
15, 2017. In response to this, "Halliburton said it did not believe that the foam cement design
used on the well caused the incident. ‘Halliburton believes that significant differences between
its internal cement tests and the commissions test results may be due to differences in the
cement materials tested,’ the company said in a statement issued in response to the letter. ‘The
commission tested off-the-shelf cement and additives, whereas Halliburton tested the unique
blend of cement and additives that existed on the rig at the time Halliburton’s tests were con
ducted.’ The company added: ‘Halliburton believes that had BP conducted a cement bond log
test, or had BP and others properly interpreted a negative pressure test, these tests would have
revealed any problems with Halliburton’s cement.’ ”
DOI: 10.1093/oso/9780199656608.003.0002
and practically. If we do not grasp what it is for a group to hold a belief, then
we cannot make sense of our widespread attributions to collective entities6
of actions that they perform, or should have performed, and of the corres
ponding responsibility that they bear. If BP, for instance, believed that there
were problems with a well safety device before the oil spill in the Gulf of
Mexico, then the company should have taken actions to repair it and is there
fore clearly culpable for the massive environmental damage that ensued.
Broadly speaking, there are two approaches to understanding the nature
of group belief. On the one hand, there is the summative view, according to
which group belief is understood as nothing more than the “summation” of
the beliefs of the group’s members. On the other hand, there is the non-
summative view, where groups are regarded as entities with “minds of their
own” and group belief is conceived of as involving actions that take place at
the collective level,7 such as the joint acceptance of a proposition. Despite
the initial plausibility of the summative approach, it is now received wisdom
in collective epistemology that group belief must be understood in non-
summative terms. In this chapter, however, I challenge this orthodoxy by
raising new, and what I regard as decisive, objections to this approach to
group belief. I then go on to develop and defend a new view, which I call the
Group Agent Account: group belief is determined in part by relations
among the bases of the beliefs of members, where these relations arise only
at the level of the collective, and are crucial especially insofar as the group is
able to function as an agent. At the same time, group belief is also largely a
matter of the individual beliefs of members. In this way, the resulting view is
neither strictly summative nor non-summative.
1.1 Summative and Non-Summative

Views of Group Belief
Let’s begin with the traditional summative account, according to which a

group’s believing that p can be understood in terms of the individual mem
bers of the group believing that p. A conservative version of the summative
account (CSA) can be formulated in the following way:
6 I will use “group” and “collective,” and “groups” and “collective entities,” interchangeably.
7 This phrase is from Pettit (2003).
GROUP BELIEF 21
CSA: A group G believes that p if and only if all or most of the members
of G believe thatp.
The CSA correctly subsumes some classic instances of group belief. For
instance, it is plausible to characterize the Northwestern community’s belief
that its institution is in Illinois in terms of all—or at least most—of its mem
bers holding this belief. However, other common examples of group belief
do not seem to fare as well on such a view. Suppose, for instance, that the
President of the university issues a statement saying that Northwestern
believes that the quarter system is not beneficial to students’ academic suc
cess. Suppose further, however, that this belief is held by only a small, yet
powerful constituency of the Northwestern community, such as the admin
istration, or an appointed committee that oversees this matter. It may still be
appropriate to attribute the belief that the quarter system is not beneficial to
students’ academic success to the Northwestern community, despite the fact
that neither all nor most of its members holds this belief. Because of cases
such as this, summative accounts are typically formulated more liberally
(LSA) as follows:
LSA: A group G believes that p if and only if some of the members of G

believe that p.
On this version of the summative view, it may be appropriate to attribute

belief that p to G even if only one of its members believes that p. For
instance, perhaps it is sufficient that the President of the university alone
believes that the quarter system is not beneficial to students’ academic suc
cess to properly attribute this belief to Northwestern. This may also be true
when a CEO holds a belief within a company, a leader holds a belief within
her cult, and a dictator holds a belief within her nation.
Even with this modification, however, the summative account of group
belief is said to suffer from a debilitating objection. In particular, it is argued
that a group can be properly said to believe that p, even when not a single of
its members believes that p. A classic example of this sort of case is where a
group decides to let a view “stand” as that of the group’s, despite the fact that
none of its members actually holds the view in question. For instance, con
sider the following:
philosophy department: The philosophy department at a leading

university is deliberating about the final candidate to whom it will extend
admission to its graduate program. After hours of discussion, all of the

members jointly agree that Jane Smith is the most qualified candidate from
the pool of applicants. However, not a single member of the department
actually believes this; instead, they all think that Jane Smith is the candidate
who is most likely to be approved by the administration.
Here, it is argued that the philosophy department believes that Jane Smith is
the most qualified candidate for admission, even though none of the mem
bers holds this belief. This attribution is supported by the groups actions:
the group asserts that Jane Smith is the most qualified candidate, it defends
this position, it heavily recruits her to join the department, and so on. This
is taken to show that individual belief that p on the part of even one of the
group members is not necessary for the groups believing thatp.8
There are different ways in which the sort of scenario found in phil
osophy department can come about. A standard route is through com
promise. If, say, half of the members of a group believe that candidate x is
the best, and the other half believe candidate y is, they might compromise
and put forward candidate z as their top candidate. Or suppose that one
member of a company believes that the appropriate minimum age for
employment is 18 and another believes it is 16. The group might adopt the
position that it is 17. Another common way for a case such as philosophy
department to arise is through the following of externally imposed rules.
For instance, a jury might come to the conclusion that a defendant is
nnocent because it was instructed to exclude all hearsay evidence, but each
individual juror might nonetheless believe that he is guilty. Similarly, an
evaluating panel might deliver the verdict that a submitted study is unpub
lishable because it does not rise to the exceedingly high standards of the
journal in question, but each member of the group might personally believe
that it is. A further way for the scenario in philosophy department to
arise is through pragmatic considerations. This is one way to understand the
case above, where the members of the philosophy department put forward
Jane Smith as the top candidate because they believe she is the most likely to
be approved by the administration. Or suppose that a group of political
leaders puts forward views, not because any of the individuals believe them,
8 See Gilbert (1989), Schmitt (1994), Tollefsen (2007 and 2009), and Bird (2010) for this
sort of argument. Pricker (2010) provides cases of this sort involving group virtues, but also
suggests that similar considerations apply in the case of group beliefs (see p. 241).
GROUP BELIEF 23
but because collectively they regard these positions as increasing their

chances of being voted back into office.
Opponents of the summative account of group belief also hold that a
group can be properly said to not believe that p, even when every single one
of its members believes that p. Hence, it is argued that individual belief that
p on the part of all of the members of a group is not sufficient for the group’s
believing that p. Consider the following:
philosophy departmenT2: The same philosophy department that is

deliberating about the final candidate to whom it will extend admission to
its graduate program is also such that every single one of its members
believes that the best red pepper hummus in Chicago can be found at
Whole Foods.
Despite the unanimity of individual belief in such a case, it is argued that it

is not correct to say that the philosophy department believes that the best
red pepper hummus in Chicago can be found at Whole Foods. This is
because assessment of red pepper hummus is entirely irrelevant to the goals
and purposes of the group.9
9 This sort of argument can be found in Gilbert (1989) and Schmitt (1994). Schmitt,
borrowing from Gilbert (1989), puts this point as follows: “Two groups may have the same
membership, yet differ in their beliefs. The membership of the Library Committee may
be identical with that of the Food Committee. Yet the two committees might have very
different purposes and accordingly make judgments about quite different issues based on
very different kinds of evidence. Every member of the Library Committee might believe that
there are a million volumes in the library, and so might the Library Committee itself. Yet the
Food Committee holds no such belief. Thus, the summative condition is too weak” (1994, p. 261).
This objection to the sufficiency dimension of summativism focuses on the relevance of
goals and purposes of the group, but there is another sort of counterexample. Gilbert (1987),
for instance, argues that each member of G might believe that p, but G itself would not be said
to believe that p if each member is unwilling or unable to communicate his or her belief thatp.
She writes:
Suppose an anthropologist were to write “The Zuni Tribe believes that the north is
the region of force and destruction.” Now suppose that the writer went on to give
his grounds for this statement as follows:
Each member of the Zuni tribe believes that the north is the region of force and
destruction, but each one is afraid to tell anyone else that he believes this; he is
afraid that the others will mock him, believing that they certainly will not believe it.
What conclusions can be drawn from this? It surely suggests at least that when we
ascribe a belief to a group we are not simply saying that most members of the group
have the belief in question. That is, it is surely not logically sufficient for a group
belief that p that most members of the group believe that p. (Gilbert 1987, p. 187)
Gilbert may wish to distinguish this sort of case from an implicit belief of a group. For
instance, J. Angelo Corlett writes: “...a decision-making group might possess a belief without
formally recognizing or accepting it. And it may do so because a certain belief may be implied
These problems have motivated the now widely accepted non-summative

account of group belief, according to which a group’s believing that p is irre
ducible to some or all of its members believing that p. Such a view holds
that in some very important sense, the group itself believes that p, where
this is understood as over and above, or otherwise distinct from, any indi
vidual member believing thatp.
There are two central versions of non-summativism. The first and per
haps most widely accepted is what we may call the joint acceptance account
(hereafter, JAA), a prominent expression of which is offered by Margaret
Gilbert in the following passage:
JAA: A group G believes that p if and only if the members of G jointly

accept that p.
The members of G jointly accept thatp if and only if it is common knowledge
in G that the members of G individually have intentionally and openly...
expressed their willingness jointly to accept that p with the other members
of G. (Gilbert 1989, p. 306)10
A key aspect of such an account is that joint acceptance does not require
belief on the part of a single member of the group in question. She writes:
It should be understood that: (1) Joint acceptance of a proposition p by a

group whose members are X, Y, and Z, does not entail that there is some
subset of the set comprising X, Y, and Z such that all the members of that
subset individually believe thatp. (2) One who participates in joint accept
ance of p thereby accepts an obligation to do what he can to bring it about
that any joint endeavors... among the members of G be conducted on the
assumption that p is true. He is entitled to expect others’ support in bring
ing this about. (3) One does not have to accept an obligation to believe or to
try to believe that p. However, (4) if one does believe something that is
inconsistent with p, one is required at least not to express that belief baldly.
(Gilbert 1989, pp. 306-7)
by one of the beliefs it accepts. For example, it would seem that each group of the requisite sort
believes implicitly that it is a group. Otherwise, the group might not qualify as a conglomerate”
(Corlett 2007, p. 236).
10 Gilbert (1987, 1993, 1994, 2002, and 2004), Schmitt (1994), Tuomela (1992), and
Tollefsen (2007 and 2009) also hold a joint acceptance view of group belief. I will discuss
Tuomela’s particular account in some detail later in this chapter.
GROUP BELIEF 25
Thus, according to Gilberts non-summative view, so long as a group jointly

accepts that p in the way described above, such a group is said to believe
that p.11
On a joint acceptance account of a group’s believing thatp, then, it is nei
ther necessary nor sufficient that some of its individual members believe
that p. It is not necessary because joint acceptance by the group members
does not require individual belief on their part, and it is not sufficient
because individual belief by the group members does not involve their joint
acceptance of the proposition in question.12 According to this account,
then, the philosophy department in the first case above believes that Jane
Smith is the most qualified candidate for admission, even though none of
the members hold this belief, precisely because they jointly agree to let this
position stand as the groups. Moreover, the philosophy department in the
second case does not believe that the best red pepper hummus in Chicago
can be found at Whole Foods, even though every single member of the
group possesses this belief, because the members never jointly accepted
such a claim. Thus, such an account delivers the correct intuitive result in
both instances.13
11 Elsewhere, Gilbert writes, “what is both logically necessary and logically sufficient for the
truth of the ascription of group belief... is... that all or most members of the group have
expressed willingness to let a certain view stand’ as the view of the group” (Gilbert 1989, p. 289).
12 For our purposes, it is sufficient to note that there is a difference between belief and
acceptance—that accepting thatp is not the same as believing thatp, and vice versa. For detailed
discussions about this difference, see, for instance, van Fraassen (1980), Stalnaker (1984),
Cohen (1989 and 1992), Wray (2001), and Hakli (2007 and 2011). While there is certainly not
consensus among these authors about what precisely this distinction amounts to, below are
four differences that have been cited between acceptances and beliefs:
1. One can accept propositions that one does not believe, whereas one cannot believe what
one does not accept.
2. Acceptance often results from a consideration of ones goals, and this results from adopting
a policy to pursue a particular goal.
3. Belief results in a feeling, in particular, a feeling that something is true.
4. Acceptance can be voluntary, whereas belief cannot. (Wray 2001, p. 325)
13 It should be noted that, at best, the joint acceptance account captures the beliefs held only
by what I called in the Introduction deliberative groups. These groups are distinguished from
non-deliberative groups by their ability to engage in collective reasoning, where this includes
deliberation, revision, and a sensitivity to evidence, all understood collectively. What is central
for our purposes here is that members of non-deliberative groups are not capable of engaging
in joint acceptance in the way required by the view. For instance, if I survey left-handed
Northwestern students and aggregate their beliefs via a majority aggregation rule, it may be
perfectly appropriate to say via this method that Northwestern students believe that there are
not enough desks to suit their needs. But given that the students themselves do not even
identify as a group, they will be unable to collectively deliberate about their needs or to jointly
accept this proposition. Thus, to the extent that a unified account of the beliefs properly
attributed to both deliberative and non-deliberative groups is desirable, the joint acceptance
account fails.
There is, however, an immediate problem facing the version of the joint
acceptance account proposed by Gilbert: groups are often large, with com
mittees or boards that are appointed to make decisions on behalf of the
group as a whole. For instance, consider the following:
medical association: The Board of Directors of the American

Academy of Pediatrics convenes and decides that its official position is that
there are significant health benefits to circumcision, which it proceeds to
publish in all of its relevant materials.14 Despite this, all of the doctors who
are members of the American Academy of Pediatrics recognize that the
evidence is inconclusive, and so have some lingering doubts that prevent
them from individually holding this belief.
As it stands, the JAA cannot countenance group belief in medical asso

ciation since the members of the American Academy of Pediatrics fail to
jointly accept that there are significant health benefits to circumcision. In
particular, only a very small percentage of the groups members—namely,
the Board of Directors—satisfies the requisite joint acceptance condition.
This structure of a collective entity is quite commonplace: groups are often
vast, rendering it practically difficult if not impossible to have each member
engage in any sort of joint activity. Thus, a smaller, more manageable body
is either elected or appointed to represent and make decisions for the larger
group. Given that medical association is certainly in the spirit of pre
cisely those sorts of cases that the joint acceptance view of group belief was
designed to accommodate, the JAA requires modification.
Raimo Tuomela proposes a different version of the joint acceptance
account that is formulated to avoid exactly the worries found with the
JAA. Specifically, he offers the following:
JAA2: G believes that p in the social and normative circumstances C if and

only if in C there are operative members Ap..., Am of G in respective pos
itions Pr.., Pm such that: (T) the agents Ap..., Am, when they are perform
ing their social tasks in their positions Pf.., Pm and due to exercising the
relevant authority system of G, (intensionally) jointly accept that p, and
because of this exercise of authority system, they ought to continue to
accept and positionally believe it;
14 This is how “hierarchical groups” in the sense characterized in Goldman (2004) function.
GROUP BELIEF 27
(2’) there is a mutual belief among the operative members Ap..., An to the
effect that (1’);
(3’) because of (T), the (full-fledged and adequately informed) non
operative members of G tend tacitly to accept—or at least ought to accept—
p, as members of G; and
(4’) there is a mutual belief in G to the effect that (3’). (Tuomela 1992,
pp. 295-6)
Tuomela’s account of group belief is quite similar to Gilbert’s, though

crucially he requires only that “operative members” engage in the joint
acceptance of the proposition in question. Operative members, according
to Tuomela, are those who are responsible for the group belief having the
content that it does which, in turn, is determined by the rules and regula
tions of the group in question. For instance, in the case of a corporation or a
large company, the board of directors may be the operative members while
the employees who work on the assembly line or in the housekeeping
department may be non-operative members. Given this amendment, the
JAA2, unlike the JAA, delivers the verdict that the American Academy of
Pediatrics believes that there are significant health benefits to circumcision
in medical association since the Board of Directors is obviously
comprised of operative members in the relevant sense.15 In what follows,
15 While the JAA2 does much better than the JAA when board- or committee-governed
groups are at issue, they both have problems countenancing certain beliefs of groups that are
dictatorially based. Consider, for example, the following:
catholic church: The Pope in his official capacity solemnly declares that the use of
assisted reproductive technologies is immoral according to the Catholic Church. While the
Pope himself believes this, this declaration was made without the Pope discussing the issue
with any other member of the Church, including the cardinals and bishops with whom he
works most closely.
In this sort of case, it may be quite natural to say that the Catholic Church believes that it is
immoral to use assisted reproductive technologies. However, while the proponent of the JAA2
may argue that dictatorially governed groups have only one operative member—the dictator—
it does not seem correct to say that there is any joint acceptance occurring. There may be
acceptance, to be sure, since the Pope may both believe and accept that assisted reproductive
technologies are immoral. But the very notion of something being joint presupposes that there
is more than one person involved. Given this, the JAA2 seems incapable of accounting for at
least many of the beliefs held by dictatorially-governed groups.
While I regard this as a genuine problem for any joint acceptance account, my central aim at
this particular point is to isolate a paradigmatic non-summative account of group belief that is
best able to capture the relevant data. Given that the JAA2 is better able to explain the beliefs
found in cases such as medical association than the JAA is, it will be the central target in
the discussion that follows.
then, I will take the JAA2 as the paradigmatic joint acceptance account of
group belief.16
The second version of non-summativism commonly accepted in the lit
erature is what we might call the premise-based aggregation account (here
after, PBAA), a central proponent of which is Philip Pettit. Like other
non-summativists, Pettit grounds his view in the argument that a group can
be properly said to believe that p, even when not a single of its members
believes thatp. Unlike other views, however, he locates his project within a
judgment aggregation framework. “Aggregation procedures are mechanisms
a multimember group can use to combine (aggregate’) the individual beliefs
or judgments held by the group members into collective beliefs or judg
ments endorsed by the group as a whole” (List 2005, p. 25).17 For instance, a
dictatorial procedure, “whereby the collective judgments are always those of
some antecedently fixed group member (the ‘dictator’)” (List 2005, p. 28)
understands the belief of a group in terms of the beliefs of a single member—
the dictator. A majority procedure, “whereby a group judges a given propos
ition to be true whenever a majority of group members judges it to be true,”
understands the belief of a group in terms of the beliefs of a majority of
its individual members (List 2005, p. 27). These are simply two examples
of judgment aggregation procedures; as we will soon see, there are cer
tainly others.
With this in mind, Pettit asks us to consider the following case:
factory: The employees of a factory are deciding whether to forgo a

pay-raise in order to spend the saved money on implementing a set of
workplace safety measures. The employees are supposed to make their deci
sion on the basis of considering three separable issues: “first, how serious
the danger is; second, how effective the safety measures that a pay-sacrifice
would buy is likely to be; and third, whether the pay-sacrifice is bearable for
members individually. If an employee thinks that the danger is sufficiently
serious, the safety measure sufficiently effective, and the pay-sacrifice suffi
ciently bearable, he or she will vote for the sacrifice; otherwise he or she
will vote against” (Pettit 2003, p. 171). Imagine now that the factory’s three
employees vote in the following way:
16 See also Tuomela (1993 and 1995).

17 For more on the theory of judgment aggregation, see List and Pettit (2002 and 2004),
Dietrich (2005), List (2005), and Pauly and van Hees (2006).
GROUP BELIEF 29
Serious danger? Effective measure? Bearable loss? Pay sacrifice?

A. Yes No Yes No
B. No Yes Yes No
C. Yes Yes No No
In factory, all three members of the group believe that the pay sacrifice
should not be made since each individual votes “No” in the conclusion col
umn. However, the group itself might decide to arrive at their collective
belief via a premise-based aggregation procedure, whereby the groups belief
is determined by the majority of votes found in the premise columns.
According to Pettit, if the group belief is determined by how the members
vote on the premises, then the group conclusion is to accept the pay sacri
fice since there are more “Yes’s than “No”s in each of the premise columns.
In such a case, “the group will form a judgment on the question of the pay
sacrifice that is directly in conflict with the unanimous vote of its members.
It will form a judgment that is in the starkest possible discontinuity with the
corresponding judgments of its members” (Pettit 2003, p. 183). This diver
gence between the belief of a group and the beliefs of its members has come
to be known as the doctrinal paradox or the discursive dilemma and it has
motivated Pettit to conclude that groups are intentional subjects that are
distinct from, and exist “over and beyond,” their individual members. He
writes: “These discontinuities between collective judgments and intentions,
on the one hand, and the judgments and intentions of members, on the
other, make vivid the sense in which a social integrate is an intentional sub
ject that is distinct from its members” (Pettit 2003, p. 184).
According to Pettit, then, a groups believing that p can be understood in
terms of combining the majority of individual beliefs that p held by a groups
members via a premise-based aggregation procedure, so long as the collec
tion of individuals is in fact a group. Moreover, given the considerations
raised in the context of discussing the joint acceptance account, we can add
that the members in question should be operative ones. Thus, a more pre
cise formulation of the view is:
PBAA: A group G believes that p if and only if the majority of the opera
tive members’ votes in the premise columns are that p.18
18 I should emphasize that while there are different aggregation procedures, and thus differ
ent ways to understand group belief in judgment aggregation terms, the premise-based proced
ure is the only one of these that results in a non-summativist conception of group belief. This is
why I formulated the PBAA as capturing both necessary and sufficient conditions for group
I now want to turn to two phenomena that have never before been discussed
in the collective epistemology literature; namely, group lies and group
bullshit.
1.2 Group Lies and Group Bullshit
To begin, I take the following to be a paradigmatic group lie:
tobacco company: Philip Morris, one of the largest tobacco companies

in the world, is aware of the massive amounts of scientific evidence reveal
ing not only the addictiveness of smoking, but also the links it has with lung
cancer and heart disease. While the members of the board of directors of
the company believe this conclusion, they all jointly agree that, because of
what is at stake financially, the official position of Philip Morris is that
smoking is neither highly addictive nor detrimental to one’s health, which is
then published in all of their advertising materials.
Since it is not my purpose in this chapter to provide an account of lying, I

will simply use the one that I will argue for in Chapter 5:19
LIE: A lies to B if and only if (1) A states thatp to B, (2) A believes thatp is
false, and (3) A intends to be deceptive20 to B with respect to whether p in
stating thatp.
Hence, in order for a group, G, to lie, I will assume that all three of these
conditions need to be satisfied: G must state thatp where G believes thatp is
false, and G must have the deliberate intention to be deceptive.21
belief—that is, when group belief is understood non-summatively on a judgment aggregation

model, a group believes thatp when and only the majority of the operative members’ votes in the
premise columns are that p. This is compatible with there being other, summative conceptions
of group belief on a judgment aggregation model, such as those that result from a dictatorial or
majority procedure. I will discuss this point in a bit more detail later in this chapter.
19 See also Lackey (2013). I should note that the only condition that will be crucial to the
arguments in this chapter is (2), and even among competing views of what it is to lie, a condi
tion of this sort is accepted.
20 For arguments against the inclusion of a condition involving deception in an account of
lying, see Sorensen (2007 and 2010), Fallis (2009), and Carson (2010). For a response, see
Lackey (2013) and Chapter 5.
21 I will develop this account in far more detail in Chapter 5, but this will suffice for present
purposes.
GROUP BELIEF 31
While there are many questions that can be asked about the nature of
group belief, group intention, and so on, that both (1) and (2) are satisfied
by Philip Morris in tobacco company is at least plausible. This is prima
facie supported by the fact that a cursory internet search of “tobacco com
pany and lies” brings up a litany of articles and websites defending precisely
the verdict that tobacco companies have lied to the public. For instance, in
a 2005 article in the Los Angeles Times, it was reported that “Jurors in Los
Angeles County Superior Court found that Philip Morris had concealed
information about the risks and addictiveness of smoking, with deliberate
intent to defraud smokers such as Fredric Reller of Marina del Rey, who
died in September 2003 at the age of 64.... In Fredric Reliefs videotaped
deposition, ‘he admitted that he was ashamed and embarrassed that he had
believed Philip Morris’s lies and deceit that there was no valid scientific
proof that their cigarettes caused lung cancer.’”22 I take it, then, that the
scenario described in tobacco company, which resembles the actual
case of Philip Morris in some crucial respects, is precisely the sort of scen
ario in which we feel comfortable attributing a lie to a group.
Given that it is clear that there are group lies, I now want to propose the
following desideratum for any plausible account of group belief:
Group Lie Desideratum: An adequate account of group belief should have

the resources for distinguishing between, on the one hand, a group’s assert
ing its belief that p and, on the other hand, paradigmatic instances of a
group’s lying regarding thatp.
I do not think that the Group Lie Desideratum needs much argument: if an
account of group belief cannot discriminate between paradigmatic instances
of group belief and clear instances where group belief is absent, the account
is fundamentally misguided. But the problem here is not just that a view’s
inability to satisfy the Group Lie Desideratum reveals its deep failure to cap
ture the nature of group belief. There are also important moral and legal
reasons for wanting to hold groups, such as corporations, businesses, and
governments, responsible both for their lies and for the consequences that
follow from them. In a case such as tobacco company, for instance, it is
not just an intellectual curiosity whether an account of group belief gets the
verdict right—it also matters so that we can properly hold Philip Morris
22 https://www.latimes.com/archives/la-xpm-2005-mar-05-fi-smoke5-story.html , accessed
July 29, 2020.
morally and legally responsible for its lies about the health risks involved in
smoking and the deaths that resulted from them.
With this in mind, I will now show that the two dominant non-summative
accounts of group belief in the literature—the JAA2 and the PBAA—are
incapable of satisfying the Group Lie Desideratum.
Let’s begin with the joint acceptance account. The first point to notice is
that, according to the JAA2, Philip Morris believes that smoking is neither
highly addictive nor detrimental to one’s health in tobacco company.
The operative members of the company—namely, the board of directors—
not only jointly accept this proposition, but also do so through the exercise
of their authority and with mutual belief. Moreover, given that the power
possessed by the board of directors is part of the very structure of the
company, the non-operative members of the group tacitly accept this
proposition and do so with awareness. Thus, conditions (1’) through (4’) are
satisfied by Philip Morris, thereby resulting in the group believing that
smoking is neither highly addictive nor detrimental to one’s health. Given
that the scenario described in tobacco company is a paradigm of
a group lying, yet the JAA2 regards it as a perfectly ordinary instance of a
group reporting its belief, the joint acceptance account of group belief is not
only incorrect, but fundamentally so.
There is a sense in which this conclusion should not come as a surprise:
the situation described in tobacco company is nearly identical in struc
ture to that found in medical association, with the exception that the
motivation for the joint acceptance in the former case is financial gain while
that in the latter case is the overall health of the nation. Since non-summative
accounts of group belief do not place any conditions on the motivation
for the joint acceptance in question, this difference is simply silent on
whether the state in question is a group lie or the reporting of a group belief.
Otherwise put, tobacco company describes a paradigmatic group lie
while medical association describes a paradigmatic group belief for
the proponent of the joint acceptance account, yet such states are indistin
guishable on such a view.23
Once we see this, it becomes clear that there are other phenomena in the
neighborhood of group lies that the joint acceptance account also fails to
distinguish from group belief. For instance, while Harry Frankfurt’s notion
of bullshit has, as far as I know, never been discussed in connection with
23 For ease of expression, I shall often compare group lies and group beliefs. However,
strictly speaking, this should be read as comparing acts of lying and acts of reporting a belief.
GROUP BELIEF 33
collective entities, it nonetheless seems clear that groups can bullshit just as
individuals can. For instance, consider the following:
oil company: After the oil spill in the Gulf of Mexico, BP began spraying
dispersants in the clean-up process that have been widely criticized by
environmental groups for their level of toxicity. In response to this outcry,
the executive management team of BP convened and its members jointly
accepted that the dispersants being used are safe and pose no threat to the
environment, a view that was then made public through all of the major
media outlets. It turns out that BP s executive management team arrived
at this view with an utter disregard for the truth—it simply served their
purpose of financial and reputational preservation.
The scenario in oil company is a classic instance of what we may call

group bullshit, which Frankfurt describes in the individual case as follows:
It is impossible for someone to lie unless he thinks he knows the truth.

Producing bullshit requires no such conviction. A person who lies is
thereby responding to the truth, and he is to that extent respectful of it.
When an honest man speaks, he says only what he believes to be true; and
for the liar, it is correspondingly indispensable that he considers his state
ments to be false. For the bullshitter, however, all these bets are off: he is nei
ther on the side of the true nor on the side of the false. His eye is not on the
facts at all, as the eyes of the honest man and the liar are, except insofar as
they may be pertinent to his interest in getting away with what he says. He
does not care whether the things he says describe reality correctly. He just
picks them out, or makes them up, to suit his purpose.
(Frankfurt 2005, pp. 55-6)
Whereas the group in tobacco company believes that smoking is highly

addictive and detrimental to one’s health, but then asserts that this is not the
case with the deliberate intention to deceive, the group in oil company
simply fails to believe that the dispersants they are using are safe and pose
no threat to the environment, but then asserts that this is the case solely to
serve their purposes. In both the former case of a group lie and the latter
instance of group bullshit, however, the relevant group belief seems to be
absent. Yet, as was the case in tobacco company, the joint acceptance
account regards the group state in oil company as a straightforward
instance of group belief. Once again, this is clear not only because the
operative members of BP—namely, the executive management team—jointly

accept that the dispersants they are using in the gulf oil spill are safe and
pose no threat to the environment, but also because the other conditions of
the JAA2 are plausibly satisfied as well. So, the joint acceptance account
delivers the verdict that there is the group belief that p, both in paradigmatic
cases where the group believes that not-p and in paradigmatic cases where
the group simply fails to believe thatp.
This motivates a second desideratum of an adequate account of group
belief, which we can characterize as follows:
Group Bullshit Desideratum: An adequate account of group belief should

have the resources for distinguishing between, on the one hand, a groups
asserting its belief that p and, on the other hand, paradigmatic instances of a
groups bullshitting thatp.
Since both a lie and bullshit undeniably involve the absence of belief, the
satisfaction of both the Group Lie Desideratum and the Group Bullshit
Desideratum are non-negotiable for a tenable account of group belief. To
my mind, the fact that the joint acceptance account is incapable of meeting
these desiderata is a decisive objection to this conception of group belief.
It is worth pointing out that these particular objections—involving group
lies and group bullshit—have not been previously raised to a joint accept
ance account of group belief. The standard problem raised against such a
view is that paradigmatic instances of group belief function differently in
important ways than beliefs in individuals do. For instance, it has been
argued by K. Brad Wray, A.W.M. Meijers, and Raul Hakli that group belief
is far more directly voluntary than it is in the individual case.24 Consider,
again, medical association, which involves the board of directors of
the American Academy of Pediatrics jointly agreeing that there are signifi
cant health benefits to circumcision. When group belief is determined by
the official position arrived at by a decision-making body in this way, the
members can simply decide that the group believes that p, whereas individ
uals do not seem capable of just deciding to believe that p in this way.25
Similarly, it has been argued that group belief in this sense is far less gov
erned by evidence than it is when an individual’s doxastic states are at issue.
24 See Wray (2001 and 2003), Meijers (2002), and Hakli (2007 and 2011).
25 For a classic defense of the view that individuals lack direct voluntary control over their
beliefs, see Alston (1988).
GROUP BELIEF 35
For instance, K. Brad Wray claims that groups, unlike individual agents, can
choose to believe based on their goals and Christopher McMahon contends
that groups often defend as true positions that they adopt for purely instru
mental reasons.26 In medical association, the goal of the AAP may be
to produce the best health for the greatest number of children, and so the
board of directors may choose to downplay their personal doubts in an
effort to further this broader aim. This way of belief formation, it is argued,
is unavailable in the individual case, where doxastic attitudes are far more
directly sensitive to evidence.
Of course, responses to these objections can and have been offered on
behalf of the joint acceptance account. For instance, while group belief on
this view may be more voluntary than individual belief, this may simply be
a matter of degree rather than of kind. For individuals surely have voluntary
control over methods of belief acquisition, sensitivity to evidence, and so
on, all of which directly affect which beliefs are formed. Moreover, it has
been questioned whether we should expect features of individual phenom
ena, such as belief, to always be possessed by their collective counterparts.27
Perhaps because entities at the individual and collective levels are so differ
ent, it shouldn’t be surprising for them to have radically different properties.
Finally, individuals often do have beliefs for purely instrumental reasons. A
politician, for instance, may believe that he is doing what is best for the
country because it is expedient, better for his image, and so on.
My central point here, however, is not to evaluate these objections in any
sort of detail but, rather, to point out that the inability of the joint accept
ance account to satisfy the Group Lie and the Group Bullshit Desiderata has
not been noticed in the literature on group belief. Moreover, while responses
have been offered to the classic objections to this view, I will now show that
there are not plausible responses to be offered to this inability.
First, a proponent of the joint acceptance account of group belief may
attempt to resist my conclusion by arguing as follows: if the central differ
ence between tobacco company and oil company, on the one hand,
and medical association, on the other hand, lies with the motivation
for the joint acceptance in question, why can’t a condition simply be added
to the JAA2 requiring a certain kind of motivation needed for group belief?
What might such a condition look like? It cannot simply require that the
joint acceptance not be motivated by the intention to deceive, for such an
26 See Wray (2003) and McMahon (2003), respectively.

27 See Gilbert and Pilchman (2014).
intention is lacking in instances of group bullshit and yet there is still the
absence of group belief. It also cannot require that the joint acceptance not
be motivated by an utter disregard for the truth, for, as Frankfurt says above,
the liar is respectful of the truth—it is just that this respect is used to conceal
the truth from the liar’s audience.
It may be better, then, to add a positive condition: perhaps the joint
acceptance needs to be motivated by a sensitivity to the truth, or to the
available evidence, or to some other epistemically proper feature. This
proposal, however, is doomed to failure for at least two reasons. First, the
motivation for the joint acceptance in medical association is the
overall health of the nation, not a sensitivity to an epistemically significant
property Given that the proponent of the joint acceptance view regards
medical association as a classic case of group belief, it would hardly
help the view to add a condition to group belief that the American Academy
of Pediatrics would fail to satisfy. Second, wishful thinking can certainly
produce belief, both at the individual and at the group level, but clearly a
positive epistemic requirement would not be satisfied here. In particular,
belief that results from wishful thinking is not sensitive to an epistemically
proper feature, despite the fact that it is undeniably a belief.
A second strategy that the proponent of the joint acceptance account
might take for resisting my objection is to flesh out a way of allowing for
group lies and group bullshit within the framework of the view. Here is how
it might go for the former: suppose that when a given group deliberates
about the question whether p, the members jointly accept that p, but then
also jointly agree to spread it about that not-p with the deliberate intention
to deceive the public. Thus, their joint acceptance of that p amounts to
group belief on the JAA2. This, combined with their agreement to convey
that not-p with the intention to deceive, results in both conditions of the
traditional conception of lying being satisfied. It is, then, possible to distin
guish between group belief and group lies on the joint acceptance account.
Although this scenario as described is certainly possible, so, too, is it pos
sible for the members of a group to move directly to jointly agreeing that
not-p and then spreading this about to the public with the deliberate inten
tion to deceive, as is done in tobacco company. It may, of course, be
obvious that all of the individuals in the group believe that p, but it clearly
doesn’t follow from this that the group also believes that p, given the non-
summative nature of the JAA2. Thus, tobacco company still represents
a paradigmatic instance of group lying that is not explainable by the joint
acceptance account.
GROUP BELIEF 37
The situation is even worse in the case of group bullshit, where there do
not seem to be any resources within the joint acceptance account for distin
guishing it from group belief. For if a group jointly accepts one thing but
then agrees to report another, this simply collapses into a group lie. If the
group instead jointly accepts a proposition with an utter disregard for the
truth, this simply turns out to be a classic case of group belief for the pro
ponent of the JAA2. There is simply no room in between to account for
group bullshit.28
Let us now turn to the premise-based aggregation account of group belief
and evaluate how this non-summativist view fares with respect to the Group
Lie Desideratum and the Group Bullshit Desideratum. To begin, consider
the following case:
JUDGMENT AGGREGATION TOBACCO COMPANY (or JA-TOBACCO

company) : The board members of Philip Morris are discussing whether
cigarette smoking is safe to the health of smokers. The board members are
supposed to make their decision on the basis of considering three separable
issues: first, whether the available evidence supports the conclusion that
smoking is not connected to lung cancer; second, whether there is reason to
think that smoking does not cause emphysema; and third, whether there is
data supporting that there is not a link between smoking and heart disease.
If a board member thinks that the evidence supports that smoking is not
connected to cancer, does not cause emphysema, and is not linked to heart
disease, he or she will vote that smoking is safe to the health of smokers;
otherwise he or she will vote that it is not. The board members vote in the
following way:
No lung cancer? No emphysema? No heart disease? Safe to health?

A. Yes No Yes No
B. No Yes Yes No
C. Yes Yes No No
After the voting, the board members decide that, because of what is at stake
financially, Philip Morris will publish in all of their advertising materials
that smoking is safe to the health of smokers.
28 For another argument against the joint acceptance account of belief, one that draws on
some of my arguments involving defeaters in Chapter 3, see Carter (2015).
Following Pettit, one way of determining the group’s belief injA-TOBACCO

company is via a premise-based aggregation procedure. On this account,
there are more “Yes’s than “No’s in each of the premise columns, so the
group believes that cigarette smoking is safe to the health of the smokers.
Indeed, Pettit’s very solution to the conflict between the individual beliefs
and the group belief in the original case is to conclude that while the group
believes that the company should forgo a pay-raise in order to spend the
saved money on implementing a set of workplace safety measures, no single
individual employee believes this. Similarly, then, the conclusion in ja-
tobacco company should be that while the group believes that cigarette
smoking is safe to the health of smokers, no single individual board member
of Philip Morris believes this. When the company then reports in their
advertising materials that smoking is safe, they are simply reporting the
belief of the group.
But doesn’t this leave us with the same problem afflicting the joint accept
ance account: namely, that the situation in ja-tobacco company intui
tively appears to be a paradigmatic group lie, and yet the view at issue
countenances it as a standard instance of reporting a group belief? To make
this case even stronger, we can imagine that the individual votes of “Yes” in
the premise columns are motivated at least in part by economic consider
ations, though not ones incompatible with belief. For instance, perhaps the
board members were inclined to look for conclusive or definitive evidence
linking smoking with lung cancer, emphysema, or heart disease before vot
ing “No” in one of the premise columns. Were the economic advantages of
selling cigarettes not present, we can imagine that their standards for believ
ing negatively would have been lower. Given this, the very fact that there are
more “Yes’s than “No’s in each of the premise columns in ja-tobacco
company is in large part the result of the financial gain promised by min
imizing the health risks of smoking. So, while each board member individually
believes that smoking is detrimental to the health of smokers, the collective
view of Philip Morris is that it is safe and this group belief is explainable by
the company’s desire for economic benefits. When the company then pub
lishes this view in all of its advertising materials—again for financial gain—
this appears to be a classic example of a group lie, and yet the PBAA regards
it as a straightforward instance of reporting a group belief.
If this is still doubted, imagine the board members sitting in the confer
ence room at Philip Morris looking at the table showing the results of their
votes. Each knows that he or she individually believes that smoking is not safe
to the health of smokers and yet each also agrees that Philip Morris should
GROUP BELIEF 39
publish in their advertising materials that smoking is safe. Moreover, assume

that it is also clear to each board member that the decision to publish the
view that smoking is safe to the health of smokers is made so as to avoid the
risk of massive financial loss. That this latter choice is sufficient for satisfying
the “intention to deceive” component of lying should be clear if it is noted
that the board could have decided to instead report that the data regarding
smoking and various diseases is mixed or inconclusive. Thus, I take it as
undeniable that in ja-tobacco company, we find a paradigmatic group
lie, despite the PBAA’s verdict that a group belief is present. Note that we
would hardly regard it as sufficient to combat accusations of Philip Morris’s
moral and legal responsibility for the smoking-related deaths of people for
the board members to respond that they were using a premise-based aggre
gation procedure in arriving at their collective view and hence actually
believed that smoking is safe. And while attributions of moral and legal
responsibility may not always track beliefs, they at least do so frequently,
and thus they provide even further reason for concluding that the group is
lying in ja-tobacco company.
Of course, it may be immediately asked why we wouldn’t simply aggre
gate the judgments of the board members in ja-tobacco company via
a different procedure. For the problem is generated in the first place by rely
ing on a premise-based aggregation procedure. But in addition to the dicta
torial and majority procedures mentioned above, there are, among others, a
supermajority procedure, whereby a group believes a given proposition to
be true whenever a supermajority of group members believes it to be true; a
unanimity procedure, “whereby the group makes a judgment on a propos
ition if and only if the group members unanimously endorse that judg
ment,” (List 2005, p. 30); and a conclusion-based procedure, whereby the
groups belief is determined by the majority of votes found in the conclusion
columns. Clearly, if a conclusion-based aggregation procedure is used in
ja-tobacco company, then the result is that each board member and
the group as a whole believe that smoking is not safe to one’s health.
There are, however, at least two problems with this move. First, there is
nothing in the judgment aggregation view that rules out using the premise
based aggregation procedure or dictates the use of a conclusion-based rule
in a case such as ja-tobacco company. Thus, it can simply be stipulated
in the case that the group’s view will be determined by its votes on the prem
ises, perhaps because the board members agreed upon this strategy from
the outset, or because they decided to aggregate this way after seeing the
results of the voting, or because it is written into Philip Morris’s bylaws that
group beliefs will be grounded in the members’ beliefs about the premises,
or because this is the most promising way of achieving the rationality of the
group. Second, recall the dialectic of the chapter: I am arguing that non-
summative accounts of group belief lack the resources for accounting for
group lies and group bullshit. Out of the relevant aggregation procedures,
the only one that supports non-summativism is the premise-based rule. For
each of the others understands the belief of the group in terms of the beliefs
of some individual or set of members, for example, the dictator, the major
ity of the members, the supermajority, and so on. If Pettit were to respond
to my challenge that non-summativism cannot accommodate group lies by
proposing the use of a procedure that supports summativism, this would
hardly save the account.
Let us now turn to the PBAAs ability, or lack thereof, to adequately
explain group bullshit. There are two scenarios to consider here, each with a
different outcome. On the one hand, if the proponent of the PBAA counten
ances group belief in cases where individual members of a collective entity
vote positively on an issue despite not personally holding the belief, then
problematic instances of group bullshit immediately arise. For we can sim
ply imagine ja-tobacco company exactly as it is described, except that
each of the board members votes with an utter disregard for the truth of the
claims. When the judgments are then aggregated via a premise-based pro
cedure and it is reported to the public that smoking is safe solely for financial
gain, the PBAA regards this as a straightforward case of a group asserting
its belief when the more plausible verdict is that Philip Morris is simply
bullshitting the public. On the other hand, if a proponent of the PBAA
countenances group belief only in cases where individual members of a col
lective entity vote on an issue because they personally hold the belief in
question, then we can again understand ja-tobacco company exactly
as it is described, except that the decisions to use a premise-based aggrega
tion procedure and to report the result that smoking is safe are made with
out any regard for the truth. This appears to be a classic example of group
bullshit, and yet the PBAA regards it as a straightforward instance of the
report of a group belief.
I should emphasize that my claim, as was the case with respect to the
joint acceptance account, is not that there aren’t any conceivable scenarios
in which the PBAA could plausibly explain an instance of a group lie or of
group bullshit. Here is one: suppose that in ja-tobacco company, every
member of the board at Philip Morris votes “No” in each of the above premise
and conclusion columns. On every way of aggregating the group’s judgments,
GROUP BELIEF 41
then, Philip Morris believes that smoking is not safe. Now suppose further
that despite knowing that they individually and collectively believe that
smoking is not safe, the board nonetheless decides to report to the public
that it is safe, either with the intention to deceive them or with an utter dis
regard for the truth. This is a case where the PBAA can plausibly explain a
group lie and group bullshit, respectively. However, there are two reasons
this does not affect the arguments in this chapter. First, my arguments show
that there are paradigmatic group lies and bullshit that the PBAA coun
tenances as straightforward instances of reporting group beliefs. This is
certainly compatible with there being some other cases of group lies and
bullshit that such an account can adequately accommodate. Second, my
arguments are targeting non-summative accounts of group belief. The only
judgment aggregation procedure that supports a non-summative account of
group belief is a premise-based one in a scenario such as ja-tobacco
company. And, as I said above, it does not respond to my objection that
non-summativism cannot capture group lies to propose cases in which a
summative aggregation procedure can.
1.3 Judgment Fragility
While I regard the failure to satisfy the Group Lie and the Group Bullshit
Desiderata as decisive objections to the joint acceptance and premise-based
aggregation accounts of group belief, there are two further considerations
against these views that should be discussed. In this section, I’ll focus on the
phenomenon of judgment fragility.
Let’s begin with the JAA2. Notice, first, that group members may jointly
accept thatp, not with an utter disregard for the truth—as is the case with
bullshit—but with little regard for the truth. This can happen especially clearly
when a view is adopted by a group for pragmatic reasons. For instance,
consider the following:
history department: The History Department at a leading university

is deliberating about the final candidate to whom it will extend admission to
its graduate program. After hours of discussion, there is still widespread
disagreement over whether Mary Jones or Thomas Brown is the most quali
fied applicant remaining in the pool. With three minutes left to the meeting
and the Chair announcing that they will need to convene again tomorrow if
a decision cannot be reached, one member proposes a different applicant
from their shortlist for admission, Robert Lee. Despite the fact that not a
single member of the department actually believes that Lee is the most
qualified candidate for the last spot, they all jointly accept this proposition
so as to end the department meeting on time and to avoid having to devote
another day to such matters. The History Department then proceeds to
report to the Graduate School that its position is that Robert Lee is the most
qualified applicant for the last spot of admission.
In history department, the members of the History Department spent

hours discussing the applicant pool for their graduate program and com
piled a short list of candidates through attention to the candidates’ qualifica
tions, so there is clearly not a complete disregard for the truth in their
overall deliberative process. Nevertheless, their jointly accepting and then
reporting that Robert Lee is the most qualified candidate for the last spot is
entirely motivated by their practical desires to end the meeting on time and
to avoid devoting another day to this issue. This results in the group state
being riddled with what we might call judgment fragility. Lets say that a
groups judgment is fragile in this sense if the following holds: were the
members of the group to deliberate about the same body of evidence at T1
and T2 with no relevant difference in the information that emerges via the
deliberation, it is very likely that the group’s judgments would diverge
between T1 and T2. Group belief, I suggest, is, ceteris paribus,29 incompat
ible with judgment fragility of this sort. This is because belief, whether it is
at the individual or the group level, is a relatively settled state. We wouldn’t
say, for instance, that you believe that I’m trustworthy if you change your
mind about this question every few minutes without any corresponding
change in the relevant evidence. Similarly, we shouldn’t say that the History
Department believes that Robert Lee is the best candidate for admission if
every time we send the group into a room to deliberate about this ques
tion, without any difference in evidence, a different answer emerges. Thus,
despite the fact that the group’s state in history department counts as
a straightforward instance of group belief on the JAA2, its judgment fragil
ity renders this the wrong verdict.
A similar case can be used to illustrate a problem with the PBAA. To see
this, consider the following:
29 This clause is intended to capture highly unusual cases where it might be argued that
belief can come and go without a change in the evidence, such as through direct brain inter
vention. I’m grateful to Nathan Lauifer for a comment that led to the addition of this clause.
GROUP BELIEF 43
JUDGMENT AGGREGATION HISTORY DEPARTMENT (or JA-HISTORY

department): A three-member History Department at a leading
university is deliberating about the final candidate to whom it will extend
admission to its graduate program. The members are supposed to make
their decision on the basis of considering three separable issues: first, whether
the applicants writing sample is the most impressive; second, whether the
students letters are the strongest; and third, whether the persons previous
coursework is the best. If a member thinks that the student s writing sample
is the most impressive, the letters are the strongest, and the previous course
work is the best, he or she will vote that the candidate is the best one;
otherwise he or she will vote that the candidate is not. Regarding one of the
candidates on the short list, Robert Lee, the votes are as follows:
Best writing sample? Best letters? Best course work? Best candidate?
A. Yes No Yes No
B. No Yes Yes No
C. Yes Yes No No
After the voting, the Chair announces that they will need to convene again
tomorrow if a decision cannot be reached and so the members decide to use
a premise-based aggregation procedure to arrive at the History
Department’s view entirely because they do not wish to meet again. This
results in the group’s believing that Robert Lee is the best candidate for
admission, which is then reported to the administration.
As was the case with history department, the members of the History
Department in this case deliberate about the applicants for their graduate
program with attention paid to the candidates’ qualifications, so there is
clearly not a complete disregard for the truth. But they also choose to use
a premise-based aggregation procedure entirely to avoid an additional
departmental meeting, which renders the resulting state subject to judg
ment fragility. In particular, were the members of the group to deliberate
about the same body of evidence at a different time with no relevant change
in the information that emerges via the deliberation, it is very likely that the
History Department would have a different view. This is because the mem
bers’ decision to aggregate their beliefs via a premise-based procedure was
guided solely by a contingent practical constraint that there is no reason
to suppose will emerge in another context. For this reason, the History
Department again does not seem to believe that Robert Lee is the best can
didate for admission, despite the PBAA’s verdict that it does.
I want to take a step back to discuss a feature that the phenomenon of

judgment fragility shares with the problems raised by group lies and group
bullshit, and to suggest a deeper diagnosis of what has gone wrong with
non-summative accounts of group beliefs. Notice, first, that both joint
acceptance and the selection of aggregation procedures are acts that are
under the direct voluntary control of the members of a group. In particular,
it is this voluntary control that enables the members of Philip Morris to
simply decide to jointly accept that smoking is safe, that allows the mem
bers of the History Department to choose to accept that Robert Lee is the
best candidate for admission, and that permits these same members to
embrace a premise-based aggregation procedure to achieve a specific out
come. Because of this, members can also be guided by a range of factors that
are utterly disconnected from the way the world is—from the economic
concerns of Philip Morris to the fleeting whims and desires of departmental
colleagues to end a meeting on time. Herein lies the problem: beliefs have a
mind-to-world direction offit. For instance, it has been argued that beliefs
aim at the truth and thus aim to fit the world, or that beliefs are satisfied or
proper when they fit the world. Regardless of the details, however, they are
importantly different from desires, which have a world to-mind direction of
it. Desires aim for the world to be a certain way and are satisfied when the
world fits them. As Mark Platts says, “beliefs should be changed to fit with
the world, not vice versa” while “the world, crudely, should be changed to fit
with our desires, not vice versa” (Platts 1979, p. 257).30 But group belief,
when understood according to non-summative accounts, can crucially lack
this mind-to-world fit. Philip Morris in tobacco company is not aim
ing to conform its state to the world, or even to be responsive to the way the
world is, when its members jointly accept that smoking is safe. Quite the
contrary; Philip Morris’s state is responsive to the way it wants the world to
be and thus has more in common with a desire than a belief. In particular,
the company wants it to be the case that smoking is safe and jointly accepts
that it is so in an effort to bring about the consequences that would follow
were the world in fact this way. In this sense, Philip Morris’s state has more
of a world-to-mind direction of fit.31
30 Platts himself does not endorse this view.

31 I mentioned earlier that a standard objection to non-summativist accounts is that group
belief ends up being far more directly voluntary than it is in the individual case. While my
argument here also partially relies on the voluntariness of both joint acceptance and the selec
tion of aggregation procedures, my central concern is that the structure of belief ends up hav
ing the wrong direction of fit on a non-summativist model.
GROUP BELIEF 45
Given that a mind-to-world fit is one of the identifying features of belief,

and the non-summativist is unable to secure this fit for group belief, it is
clear that we need to look elsewhere to understand this phenomenon.
1.4 Base Fragility
Let us turn to the final problem for non-summativism, what I call base
fragility, and begin with how it afflicts the JAA2. Consider, for instance, the
following case:
English department: The English Department at a leading univer

sity is deliberating about the final candidate to whom they will extend
admission to their graduate program. All of the members jointly accept that
the best candidate for admission is Sarah Peters, but half of them agree to
this because they believe that she is a highly qualified applicant and half of
them agree to this because they believe that she is a highly unqualified
applicant. The latter half of the department is made up of a contingency of
disgruntled employees who wish to sabotage their own department and
regard “the best candidate for admission” as the applicant who will most
likely pull the programs rankings down.
Once again, the joint acceptance account regards this as a straightforward

instance of group belief. But notice: because the members of the English
Department jointly accept the proposition that Sarah Peters is the best can
didate for admission for different and indeed competing reasons, the result
ing state has a base fragility to it that is not present in standard cases of
belief. Let us say that a group’s state is base fragile if the bases of a significant
subset of its members’ beliefs conflict with the bases of another signifi
cant subset of its members’ beliefs. The English Departments view about Sarah
Peters is clearly base fragile in this sense. This can be seen by noticing that
any future evidence that the English Department acquires regarding Sarah
Peters’s qualifications, whether it is for or against them, will count against
the group belief. For instance, evidence on behalf of Peters’s qualifications
will sway the disgruntled employees away from continuing to regard her as
the best candidate for admission and evidence that undermines her qualifi
cations will persuade the other half of the department that she is no longer
the best candidate for admission. This example can, of course, be even further
complicated so that one-quarter of the group’s members believe that p for
reason q, one-quarter believes that p for reason r, and so on. Hie more
heterogeneous the grounding for the joint acceptance is among the group
members, the more base fragile the resulting state is.
Base fragility of this sort is, I claim, incompatible with group belief for at
least two reasons. First, group beliefs have to be the sorts of things that are
properly subject to epistemic evaluation, and states that are base fragile are
not. In particular, group beliefs have to be evaluable as rational or irrational,
justified or unjustified, undefeated or defeated, and so on. When a groups
belief is held in the face of such base fragility, however, no single coherent
evaluation can be given. If, for instance, the English Department gets fur
ther evidence against Sarah Peters’s qualifications, does this render its belief
that she is the best candidate irrational, unjustified, or defeated? No single
answer can be given here. When viewed in light of one set of bases, the evi
dence counts against the belief, but when viewed in light of another set, it
counts in favor of it. This deep lack of unity reveals that the state that is
purported to be a single one belonging to a group is in fact a collection of
individual beliefs.
This is not to say, of course, that all of the members of a group need to
hold a belief for the same reasons. Indeed, one of the epistemic virtues of
group belief is that a group’s members might all hold a belief for different,
mutually supporting reasons. This can render the resulting state better off
epistemically than any of the individual states taken alone, The point here is
that the group’s belief cannot be base fragile, where this means that the
bases of the individually-held beliefs are wildly conflicting.
Second, group beliefs have to be the sorts of things that can coherently
figure into collective deliberation about future actions of the group, and
states that are base fragile cannot. For instance, if the English Department is
deliberating about how to act, they should arrive at conclusions typical of a
group that believes that Sarah Peters is the best candidate for admission-
such as nominating her for a university fellowship or writing her an out
standing letter of support. But this is not in fact how things will turn out, as
half of the members will be deliberating in ways that are typical of a group
that believes that Sarah Peters is the worst candidate for admission. Thus,
there will be widespread disagreement among the members about future
actions related to this proposition, ultimately leading either to inertia, inco
herence, or a change in the bases of some of the members.
It is not difficult to see that a problem involving base fragility arises with
respect to the PBAA, too. We can simply leave ja-history department
GROUP BELIEF 47
as it is, except we can imagine that the department members’ votes on the
premises are riddled with base fragility. For instance, department member
A might vote that Robert Lee’s writing sample is the best for reason q—say,
that it is the most historical—while member C votes that it is the best for
reason that it is not the most historical. This might happen if, for
instance, A and C have different interpretations of the role that the histor
ical elements play in the writing sample—perhaps A regards such elements
as the central focus of the paper, while C regards them as merely incidental
aspects supporting a non-historical claim. Similarly, department member B
might vote that Lee’s letters are the best because of reason r—that they
emphasize how professional he is—while member C votes that they are the
best for reason ~r—that they focus on how he is not professional. C might
be looking for a pure lover of the discipline rather than a highly profession
alized candidate, and perhaps B and C have competing visions of what pro
fessional activity involves. Finally, department member A might vote that
Lee’s course work is the best because of reason 5—that his courses reveal the
most breadth—and member B might vote that his course work is the best
because of reason —that his courses reveal the least breadth. Perhaps B
values depth over breadth, and A and B disagree over whether Lee’s course
work has breadth because only A regards interdisciplinary work as relevant
to his evaluation. Thus, the votes are as follows:
Best writing sample? Best letters? Best course work? Best candidate?
A. Yes/g No Yes/s No
B. No Yzslr Yes/~s No
C. Yes/~g Yes/~r No No
As should be clear, the PBAA regards this as a clear instance in which the
History Department believes that Robert Lee is the best candidate for
admission. But as was the case in English department, the resulting
state is base fragile in a way that renders it unfit for proper epistemic evalu
ation. If, for instance, the History Department gets further evidence about
Lee’s writing sample, it is unclear whether it would render its belief that Lee
is the best candidate irrational, unjustified, or defeated. When viewed in
light of one set of bases, the evidence might count against the belief, but
when viewed in light of another set, it might count in favor of it. As we saw
above, this sort of base fragility is incompatible with group belief.
Thus, the phenomenon of base fragility provides a further reason to
reject non-summativism about group belief.
1.5 The Group Agent Account
We have seen that a central problem facing the two classic non-summative
accounts of group belief—the JAA2 and the PBAA—is their inability to
satisfy the Group Lie and the Group Bullshit desiderata. I take this to be a
decisive reason to reject such accounts. But notice: this argument also goes
some distance toward resurrecting a broadly summative approach to group
belief. Here is why: summativism was traditionally regarded as the intuitive
approach to understanding the phenomenon of group belief. What under
mined this view were precisely cases such as philosophy department
and philosophy departments, which purported to show that indi
vidual belief thatp on the part of a groups members is neither necessary nor
sufficient for the group believing thatp. If the scenario described in phil
osophy department is indistinguishable from paradigmatic group lies
and instances of bullshit, however, then surely we should no longer grant
that the philosophy department, in jointly accepting that Jane Smith is the
most qualified candidate for admission to their graduate program, clearly
believes this proposition. So, one of the central reasons for rejecting sum
mativism in the first place no longer holds.
We have also seen, however, that issues about fragility, particularly at the
level of the bases, rule out understanding group belief entirely in summative
terms. For even if every member of a group believes thatp, they might do so
for wildly conflicting reasons, which renders the resulting state unfit for
epistemic evaluation and for future deliberation in relation to group action.
This, then, prevents the state from being a group belief. Indeed, consider
ations about judgment and base fragility show that, in a deeply important
sense, group belief is crucially connected to our understanding of groups
as agents in their own right. When individuals make up a group, there are
relations that arise among their beliefs that can only be properly assessed at
the level of the collective. Whether these relations are together coherent
or incoherent, for instance, is critical in assessing whether a belief state is
appropriate for figuring in the groups actions. And the nature of these rela
tions at the collective level directly impact whether a groups action is
rational or justified in light of its belief states.
I thus propose the following account of group belief, which avoids all of
the problems afflicting rival views:
Group Agent Account: A group, G, believes that p if and only if: (1) there
is a significant percentage of G’s operative members who believe that p, and
GROUP BELIEF 49
(2) are such that adding together the bases of their beliefs that p yields a
belief set that is not substantively incoherent,32*
There are five features to note about the Group Agent Account. First, the
addition of condition (1), which necessitates belief on the part of a signifi
cant percentage of operative members, enables my view to satisfy the Group
Lie and the Group Bullshit Desiderata. This can be seen by noticing that
such a condition fails to be satisfied in tobacco company since not a
single member of the board of directors of Philip Morris believes that smok
ing is neither highly addictive nor detrimental to one’s health. It also fails to
be fulfilled in oil company, as not a single member of the executive man
agement team of BP believes that the dispersants they are using are safe.
Thus, the Group Agent Account is able to accommodate the verdict that
Philip Morris is lying in the former case and BP is bullshitting in the latter,
thereby having the resources for distinguishing between a groups asserting
its belief, on the one hand, and its lying or bullshitting on the other.
Second, the Group Agent Account avoids the problem posed by judg
ment fragility since group belief is not determined by factors, such as joint
acceptance or choices about which aggregation procedure to use, that are
under the direct voluntary control of the groups members. Indeed, it is pre
cisely because of this level of voluntarism that judgment fragility arises. In
history department, for instance, the members simply choose to
accept Robert Lee as the best candidate for admission, and it is this choice—
grounded entirely in pragmatic factors—that renders the groups judgment
fragile. Were they to deliberate about the same issue with the same evidence,
though without the worry about the meeting ending in five minutes, they
very likely would have arrived at a different conclusion. Similarly, the choice
about which aggregation procedure to use in ja-history department
is taken up directly, simply because the group wishes to arrive at the desired
outcome. This leaves group belief on both views subject to the fleeting
whims and temporary desires of its members. In contrast, condition (1) of
the Group Agent Account ties group belief intimately to individual beliefs,
so that the level of voluntary control at the former level is no greater than it
is at the latter level. This has the consequence that group belief is no more
riddled with judgment fragility than individual belief is.
32 One lesson that is often drawn from the Preface Paradox is that there are some kinds of
incoherence that are not irrational. The addition of “substantively” is intended to permit a
group to have a belief even when there is this sort of incoherence.
Third, and related, the Group Agent Account gets the direction of fit right
for group belief. In particular, since individual beliefs have a mind-to-world
direction of fit, and since individual beliefs provide the building blocks
for group belief on my view, group belief also has a mind-to-world direc
tion of fit.
Fourth, because condition (2) of the Group Agent Account requires that
the bases of the individual beliefs of the operative members not be incoher
ent, it avoids the problem posed by base fragility. In particular, in cases such
as English department, the wildly conflicting reasons that the mem
bers have for accepting that Sarah Peters is the best candidate for admission
leads to the failure of (2), and thus the resulting state fails to qualify as a
group belief. How should we characterize the requirement that the reasons
not be incoherent? There are various ways of understanding this. One
option is to understand coherence in terms of evidential support and inco
herence in terms of a lack thereof.33 Alternatively, coherence can be under
stood in terms of accuracy-dominance avoidance, in the sense that, for a
coherent set of beliefs, there is no rival belief set that is never worse and
sometimes better than it with respect to overall accuracy, and incoherence
could then be fleshed out accordingly.34 Here, I take no stand on how pre
cisely this concept should be understood. It suffices for my purposes that
there is, intuitively, the presence of incoherence among the members’ bases
in cases such as English department, and that there are available
accounts that can explain this.
Fifth, even though the Group Agent Account is not a simple summativist
one, it nonetheless faces the relevance objection posed by cases like phil
osophy department!. Recall that the objection here is that every
member of a group may believe that p, but believing that p may be entirely
irrelevant to the purpose and goals of the group. So, for instance, every
member of the philosophy department may in fact believe that the best red
pepper hummus in Chicago is at Whole Foods, but this may be completely
disconnected from the focus and objectives of the collective entity. The
problem is that, on my view, the philosophy department ends up believing
that the best red pepper hummus in Chicago is at Whole Foods—so long as
the bases of the individual beliefs aren’t incoherent—even though this is
said to be the wrong intuitive verdict.
33 See Kolodny (2007). 34 See Briggs, Cariani, Easwaran, and Fitelson (2014).
GROUP BELIEF 51
By way of response, notice that there is a difference between a group

having a belief, on the one hand, and a group having a relevant or important
belief, on the other. There is nothing peculiar in itself in saying that the phil
osophy department believes that the best red pepper hummus in Chicago
can be found at Whole Foods. It is just that such a belief is typically of so
little interest to us that we wouldn’t overtly make this attribution to the
group. But this is true in the individual case as well, and yet we wouldn’t
withhold belief here. For instance, you most likely believe that oranges don’t
grow on kangaroos and that the Earth is more than 20 years old, but only
under highly unusual circumstances would I explicitly attribute these beliefs
to you. Why? Because such beliefs are of very little interest to me. It doesn’t
follow from this, however, that you don’t hold such beliefs. According to the
Group Agent Account, the same is true in the group case.
In fact, it is worth pointing out that the non-summativist who appeals to
this sort of relevance objection is committed to a very counterintuitive con
ception of belief, group or otherwise. For instance, suppose that every
member of PETA believes that Citizen Kane is the greatest film of all time.
On this view, PETA fails to believe this at T1 since it is irrelevant to its goals,
but then believes this at T2, when its President announces that PETA will
now be extensively evaluating the depiction of animals in films. Nothing
has changed about the psychology of any of the group’s members or the
propositions they accept, yet they now have a belief simply because of an
announcement from the group’s President. This conclusion strikes me as a
further reason to doubt the force of this objection from relevance.
Sixth, it is worth returning briefly to the cases with which this chapter
began, for it may be thought that they still pose a problem for the Group
Agent Account. For instance, if the intuitive description of philosophy
department is that the group believes that Jane Smith is the most quali
fied candidate for admission to its graduate program despite not a single of
its members believing this, then the mere fact that it is structurally identical
to a group lie and group bullshit does not undermine the intuitiveness of
this description. In other words, regardless of its similarity to cases where
group belief is clearly absent, philosophy department describes a
scenario where group belief seems to be present.
By way of response, let me say that it is not at all clear to me that the
intuitive response here is that the group holds the beliefs in question. In
fact, there are many other plausible ways to describe this case that do not
involve belief at all. For instance, we can say that the philosophy depart
ment’s official position is that Jane Smith is the most qualified candidate for
admission to its graduate program, or that the philosophy department has

decided to accept this,35 or that this is its public view, and so on. All of these
characterizations make clear without invoking the notion of group belief
that the group bears a relationship to the proposition in question that none
of the individuals may share, and it does so without being committed to
anything such as a group mind that is over and above the minds of any
individual members. I would say something similar in the case of a jury,
where they come to a conclusion about the guilt or innocence of a defendant
because of the rules they are instructed to follow, despite the fact that not
a single juror in fact believes it. In such a case, I would say that the group’s
verdict is that, say, the defendant is innocent, despite the fact that not a single
member believes this to be true.
So, the intuitiveness of the group’s having a belief that no individual
member does in philosophy department strikes me as merely apparent.
Further support for this conclusion derives from considering an individ
ual analogue of this case. Suppose, for instance, that only a single member
of the philosophy department reports to the administration that Jane Smith
is the most qualified candidate for admission to its graduate program,
despite the fact that she does not believe that this is the case. How would we
describe this situation? The standard view is that the member of the phil
osophy department accepts, but does not believe, the proposition in ques
tion. But then why wouldn’t we say this in cases that are identical in all
respects except that a group is substituted for an individual? Why would the
mere fact that a collective entity is involved transform the psychological
state of acceptance into belief? Since there does not appear to be a compel
ling answer to this question, philosophy department does not motiv
ate the rejection of the Group Agent Account.
Finally, let me offer a few general words about group belief, and the very
important lessons we have learned about it from reflecting on group lies
(and related phenomena, such as group bullshit). My general view will be
met with resistance from opponents on two radically different sides.36 On
the one side, there will be those who hold that there simply are no group
35 This response has been developed in detail by Wray (2001), Meijers (2002), and
Hakli (2007).
36 It might be thought that this distinction simply maps the difference between summativism
and non-summativism, but this would not be quite right. For instance, while a summativist
might understand group belief in terms of individual beliefs, this need not be understood as a
form of group belief eliminativism. I am here interested in contrasting those who think we
shouldn’t even be theorizing about group beliefs, since talk of this phenomenon is simply
metaphorical.
GROUP BELIEF 53
beliefs, and any talk to the contrary is metaphorical. According to this

position, treating group belief as a phenomenon worthy of philosophical
treatment in its own right is deeply mistaken. On the other side, there will
be those who maintain that groups have “minds of their own,”37 and that
their mental states are over and above, or distinct from, any mental states
of their individual members. Group belief, on this view, does not even par
tially depend on individual belief, so the two phenomena are importantly
different. What I hope to have shown in this chapter is that paying close
attention to group lies reveals that both of these sides are wrong. If we take
as our starting point what I regard as an undeniable fact—namely, that
groups lie—then it becomes clear both that groups genuinely have beliefs,
and that they need to be anchored by individual beliefs.
To see this, notice first that when we talk about groups lying, this is not
simply metaphorical. When Fredric Reller said in his videotaped deposition
that he believed Philip Morris’s lies that there is no valid scientific evidence
that cigarettes cause lung cancer,38 he wasn’t speaking loosely—he was
attributing a full-blown lie to Philip Morris, just as he would to you or me.
Groups can lie, and when they do, the consequences can be catastrophic.
But notice: in order to understand what it is for a group to lie, we need
to have a robust conception of what it is for a group to have a belief. For, on
every plausible conception of lying, even those that are in deep disagree
ment, a necessary condition is that the liar either believes that what is said is
false, or fails to believe that it is true.39 Given this, the very notion of group
belief is at the heart of understanding group lies. Since I take it as undeni
able that groups lie, I also take it to be clear that they have beliefs, too.40
As we have seen in this chapter, however, the notion of group belief that
is needed when theorizing about group lies cannot be understood in terms
of the non-summative proposals in the offing, as they deliver the wrong
results in cases of paradigmatic lies. This is why group belief needs to be
anchored by individual beliefs. Otherwise, it turns out that features such as
economic motivations can wholly determine whether a group holds a belief
and, thereby, whether they’ve told a lie. If, for instance, it is in Philip Morris’s
financial interests to believe that smoking is safe, then all that needs to be
37 See, for instance, Pettit (2003).

38 See the earlier discussion of this in section 1.2 of this chapter.
39 See, for instance, the references in note 20.
40 This is compatible with granting that many phenomena that are called “group beliefs” in
fact are not. As I said earlier, I would regard many of these states as the group’s official position,
acceptance, verdict, and so on.
done to deny culpability for deceiving smokers on most non-summative

accounts is to get the operative members of the group in a room and have
them agree that smoking is safe. Voila: Philip Morris now believes that
smoking is safe, and thus there is no lying when this is reported to the public,
even in a court of law. If this conclusion strikes you as deeply wrong, as it
does me, then you should consider taking on board the view of group belief
defended in this chapter.
1.6 Conclusion
We have seen that non-summative accounts fail to satisfy the Group Lie and
the Group Bullshit Desiderata, and thus that group belief cannot be deter
mined by states or processes that are under the direct voluntary control of
the members. We have also seen that non-summative accounts incorrectly
countenance as group beliefs states that are riddled with judgment and
base fragility. This leaves group belief without a mind-to-world direction of
fit and renders it unsuitable for proper epistemic evaluation and collective
deliberation. Thus, the current orthodoxy in epistemology according to
which non-summativism is the only game in town is deeply mistaken.
In place of non-summativism, I defended the Group Agent Account.
On my view, group belief is largely a matter of the beliefs of individual
members, yet it is also importantly constrained by relations that arise only
at the level of the group, especially as it is an agent. The result is a view that
not only renders group belief incompatible with judgment and base
fragility, it also satisfies the Group Lie and Group Bullshit Desiderata,
thereby providing the resources for holding groups responsible for their lies
and bullshit.
What Is Justified Group Belief?
As we saw in Chapter 1, groups are often said to believe things. Some of

these beliefs amount to knowledge while others do not, with epistemic justi
fication being one of the central features distinguishing these two categories.
But how should we understand a groups justifiedly believing that p?1
The importance of this question is clear, both theoretically and practic
ally. If we do not understand the justification of group beliefs, then we
cannot make sense of our widespread epistemic attributions to collective
entities—of evidence that they have, or should have, and of propositions
that they know, or should have known. Moreover, the justificatory status of
such beliefs matters a great deal to whether groups are morally and legally
responsible for certain actions and, accordingly, the extent to which they
ought to be held accountable. For instance, if the Bush Administration justi
fiedly believed that Iraq did not have weapons of mass destruction, then not
only did the Administration lie to the public in saying that it did, but it is
also fully culpable for the hundreds of thousands of lives needlessly lost in
the Iraq war.
Despite this, the topic of group justification has received surprisingly
little attention in the literature, with the few who have addressed it falling
into one of two camps. On the one hand, there are those who favor an infla
tionary approach, where groups are treated as entities that can float freely
from the epistemic status of their members’ beliefs. For these theorists, the
justificatory status of group belief involves only actions or features that
take place at the group level, such as the joint acceptance of reasons. On
the other hand, there are those who endorse a deflationary approach, where
justified group belief is understood as nothing more than the aggregation of
the justified beliefs of the groups members.
In this chapter, I raise new objections to both of these approaches. If I am
right, we need to look in an altogether different place for an adequate
1 I will frequently speak simply of a “group justifiedly believing” a proposition, “group

justification,” or “group justifiedness.” All of these locutions should be understood as involving
group epistemic justification.
DOI: 10.1093/OSO/9780199656608.003.0003
account of justified group belief. From these objections emerges the skeleton
of the positive view that I go on to develop and defend, which parallels
my account of group belief in the previous chapter in critical respects and
which I call the Group Epistemic Agent Account: groups are epistemic
agents in their own right, with justified beliefs that respond to both evidence
and normative requirements that arise only at the group level, but which
are nonetheless importantly constrained by the epistemic status of the
beliefs of their individual members.
2.1 Divergence Arguments
Before turning to particular approaches to group justification, two points

of clarification about the topic of this chapter are in order. First, the
analysandum of all of the views under consideration is doxastic, rather
than propositional, justification. Thus, the question to be answered is
when a group has a justified belief—that is, justifiedly believes that p—
rather than has justification for believing a proposition without necessar
ily believing it—that is, is justified in believing thatp. Second, the views
at issue are exclusively concerned with epistemic justification, which is
the kind of justification that is integral to converting true belief into
knowledge.2 Practical and moral justification will not figure into the
discussion.
With these points in mind, let us begin with the inflationary approach to
the justification of group beliefs. The primary support for this view comes
from divergence arguments, which purport to show that there can be a
divergence between the justificatory status of a group’s beliefs and the status
of the beliefs of the group’s members. In particular, it is claimed that a group
can justifiedly believe that p, even though not a single one of its members
justifiedly believes that p. There are two central kinds of cases that purport
to establish these conclusions. Let’s call them different evidence cases and
different epistemic risk settings cases.
An instance of the first kind can be seen in the following:
2 Of course, this is not to say that knowledge is nothing more than justified true belief (see,
for instance, Gettier (1963)), but that epistemic justification is key to distinguishing between
mere true belief and knowledge.
WHAT IS JUSTIFIED GROUP BELIEF? 57
different evidence: A jury is deliberating about whether the

defendant in a murder trial is innocent or guilty. Each member of the jury is
privy to evidence that the defendant was seen fleeing the scene of the crime
with blood spatter on his clothes, but it is grounded in hearsay that, though
reliable, was ruled as inadmissible by the judge. Given only the admissible
evidence, the jury as a group justifiedly believes that the defendant is innocent,
but not a single juror justifiedly believes this proposition because it is defeated
for each of them as individuals by the relevant reliable hearsay evidence.
Cases of this sort are prevalent in the collective epistemology literature, but
Frederick F. Schmitt provides the most developed and detailed version.
According to Schmitt (1994), different evidence cases successfully function
as divergence arguments only when they involve chartered groups, where
“[a] chartered group is one founded to perform a particular action or
actions of a certain kind,” and “has no life apart from its office” (1994,
pp. 272-3). In other words, chartered groups must function only in their
offices or risk ceasing to exist. The U.S. Congress, the Sierra Club, and juries
are all groups of this sort. Moreover, given the particular charter of a group,
it may be governed by special epistemic standards, such as the exclusion of
hearsay in a court of law. Because of this,
...a nonlegal group may fail to be justified in a belief because a member

possesses countervailing hearsay. A court, on the other hand, would not
lose its justification merely because a member possesses countervailing
hearsay. And this is because in its legal capacity, the court rightly excludes
hearsay, and its legal capacity is the only capacity in which it operates.
(Schmitt 1994, p. 274)
Since the jury in different evidence is a chartered group, Schmitt

argues that its charter prohibits it from considering the hearsay evidence
about the defendant fleeing the scene of the crime with blood spatter on his
clothes. Without this crucial testimony, the jury justifiedly believes that the
defendant is innocent of the murder in question. But since the jurors qua
individuals are not governed by these special standards of available reasons,
they each have a defeater provided by the hearsay evidence for believing
in the defendant’s innocence. Thus, the jury justifiedly believes that the
defendant is innocent despite the fact that not a single individual member
justifiedly holds this belief.
An instance of the second kind of case is as follows:
different risk settings: A philosophy department has been given

permission to hire an assistant professor and appoints a sub-committee of
three persons for this task. After considering the application of Fred Jones,
the individual members and the committee have different epistemic risk
settings with regard to accepting the proposition that Jones is a qualified
candidate. “These risk settings determine how much evidence is necessary
for acceptance. So, while both the individuals and the group as a whole
consider precisely the same evidence and they assign the same weight to the
evidence, the group reaches its threshold for acceptance while no individual
member has reached her threshold for acceptance. And, since there is no
epistemically preferred threshold, both the group and the members are
equally epistemically rational.” (Mathiesen 2011, p. 41)
This case, due to Kay Mathiesen, relies on differences in tolerance for epi
stemic risk that agents may have. For instance, it has been argued that if one
agent is an epistemic risk-taker, while another is epistemically cautious, it is
possible for them to have access to the same evidence and yet the former
rationally believes thatp while the latter rationally suspends belief regarding
the question whether p.3 Moreover, these differences in epistemic risk set
tings can be determined by pragmatic factors. Mathiesen writes:
A practical agent has certain goals or interests. In the case of the hiring
committee, its practical goals have been determined by the charge from
the departmental committee to determine a set of “qualified” candidates
for the position. The practical goals of the members may be quite different
from those of the group. For instance, the individuals may “personally”
prefer to be very skeptical that anyone is truly qualified. But, given that
as a group they need to present a set of names to the department, such
skepticism would be out of place in group reasoning.
(Mathiesen 2011, p. 40)
Because of the practical interests of the hiring committee in different

risk settings, the group is more of an epistemic risk-taker than any of
the individual members. According to Mathiesen, this has the result that
even though the amount of evidence available to both the group and the
3 See Levi (1962), Fallis (2006), Riggs (2008), and Mathiesen (2011).
members is sufficient for justified belief, only the formers risk settings
permit belief. Thus, the hiring committee justifiedly believes that Jones is a
qualified candidate, despite the fact that not a single individual member
justifiedly holds this belief.
Divergence arguments involving different evidence and different risk set
tings are said to support two conclusions—a negative and a positive one.
The former is:
Non-Summativism: A group, G, justifiedly believing that p cannot be

understood only in terms of some or all of G’s members justifiedly believ
ing that p.
The positive conclusion is:
Inflationism: A group, G, justifiedly believing that p is understood in

terms of the group itself justifiedly believing that p, where this is over and
above, or otherwise distinct from, the individual members of G justifiedly
believing that p.
According to inflationary non-summativism, then, a group justifiedly believ

ing that p is irreducible to all or some of its members justifiedly believing
that p; instead, the group itself is the epistemic subject of such justified
belief.4 In what follows, we will take a closer look at the paradigmatic version
of inflationary non-summativism: the joint acceptance account.
2.2 The Paradigmatic Inflationary Non-Summativist View:

The Joint Acceptance Account
The most widely accepted inflationary view of group justification is what

we may call the joint acceptance account (hereafter, the JAA5). One version
4 It might be asked whether it is the groups belief, or its justification, that is over and above,
or otherwise distinct from, the individual members of the group justifiedly believing thatp. To
the extent that justification and belief can be considered entirely separately when doxastic
justification is in question, the issues in this chapter will concern justification. But most
theorists who are inflationary about justification are also inflationary about belief, that is, they
argue that both epistemic justification and belief are fundamentally a matter of joint accept
ance. See, for instance, Schmitt (1994).
5 This is obviously to be distinguished from the joint acceptance account of group belief
discussed in Chapter 1.
of the JAA is developed and defended by Schmitt in his (1994), where

he argues:
JAA-S: A group G justifiedly believes that p if and only if G has good reason
to believe that
p, and believes thatp for this reason6
where
G has a reason r to believe that p if and only if all members of G would
properly express openly a willingness to accept r jointly as the group’s reason
to believe that p. (Schmitt 1994, p. 265)7
On this view, then, whether a reason counts as possessed by a group is

determined entirely via its joint acceptance by the group’s members, and the
epistemic goodness or badness of this reason can then, in turn, be fleshed
out in terms of traditional justification-conferring features, such as being
produced by a reliable process, being grounded in adequate evidence, and
so on. Schmitts preferred explanation of the epistemic goodness of a
groups reason is reliabilist, and though there are interesting questions about
the reliability of group belief, the aspect of his view that is a substantive
contribution to collective epistemology is his joint acceptance account of
group reasons.
According to Schmitt, while joint acceptance determines group reasons,
“[t]he reference to what members would properly do is needed because
the reasons possessed by the group include those that are available within
and to the group, not merely those the members actually jointly accept as
reasons” (Schmitt 1994, p. 266, original emphasis). For instance, suppose
that the members of the Humane Society of the United States do not expli
citly jointly accept, that the moral wrongness of animal cruelty gives them a
reason to believe that dog fighting should be opposed; nevertheless, this
reason might be available to the group via all of their other commitments.
Moreover, though Schmitt does not mention these virtues of his view, the
inclusion of what members would do is also necessary to account for cases
where r seems to be possessed by a group, despite the fact that not all
6 I added “and believes that p for this reason” to Schmitt’s account; otherwise, group beliefs
and group reasons will be entirely disconnected from one another.
7 Because Schmitt talks about groups having reasons, I will adopt this locution in the dis
cussion that follows. But where relevant, this should be understood as groups not only having
these reasons, but basing the beliefs in question on these reasons.
members of the group actually jointly accept r. This can happen when, for
instance, a group member is out of town or ill and is therefore not present
when the relevant deliberation takes place, or when the group is so large
that actual joint acceptance is practically impossible. In such cases, so long
as all of the group members would jointly accept r, it counts as a reason that
the group has.8
Another version of the JAA is defended by Raul Hakli. According
to Hakli:
JAA-H: A group G justifiedly believes9 that p collectively “if and only if

the group can successfully defend p against reasonable challenges by pro
viding reasons or evidence that are collectively acceptable to the group and
that support p according to the epistemic principles collectively accepted
in the epistemic community of the group. The epistemic community deter
mines what counts as a successful defence and as a reasonable challenge.”
(Hakli 2011, p. 150)
Just as Schmitt provides a joint acceptance account of group reasons, and

then defends a reliabilist account of what makes these reasons epistemically
good ones, Hakli endorses a joint acceptance account of group reasons and
then develops a dialectical view of what makes these reasons epistemically
good ones. In particular, a group has only “reasons or evidence that are col
lectively acceptable to the group.” What makes these reasons or evidence
good is if they can be used to successfully defend p against reasonable
challenges, which, in turn, is determined by what the epistemic community
collectively accepts.
For proponents of divergence arguments, the central virtue of the JAA is
its ability to account for how groups can justifiedly hold beliefs that no sin
gle member justifiedly believes, thereby supporting inflationism about
group justification. For instance, in different evidence, that the jury
justifiedly believes that the defendant is innocent without a single individual
8 Beyond this, Schmitt says that he will not offer an account of what proper acceptance is.
The key point for my purposes, however, is that “proper” is not an epistemic notion. As Schmitt
says, “proper joint acceptance of a reason is not the same as the reasons being good. Joint
acceptance of r as a reason may be proper even if the reason is bad” (1994, p. 266). Instead,
“proper joint acceptance” will often be determined by the structure or procedural requirements
of the group in question.
9 It should be noted that Hakli provides an account of the justification of group acceptances
rather than of group beliefs. None of the arguments in this chapter, however, turn on this
distinction.
member justifiedly holding this belief can presumably be explained according

to both versions of the JAA: the members of the jury would jointly express
openly a willingness to accept the admissible evidence as their reason to
believe that the defendant is innocent. Since we can assume that this
admissible evidence is both reliably produced10 and capable of providing a
successful defense against reasonable challenges, the jury justifiedly believes
this proposition. However, the individual members not only don’t believe
that the defendant is innocent, they also don’t have justification for believ
ing it, as the hearsay evidence provides them with a defeater.
Similar considerations apply in different risk settings: the mem
bers of the hiring committee would jointly express openly a willingness to
accept the available evidence as their reason to believe that Jones is a quali
fied candidate. Since we can again assume that this evidence is reliably
produced and capable of providing a successful defense against reasonable
challenges, the hiring committee justifiedly believes this proposition. But
because the members qua individuals are more epistemically cautious, they
do not believe Jones is a qualified candidate, and thus they don’t believe this
justifiedly either. This results in the hiring committee justifiedly believing
that Jones is a qualified candidate, despite the fact that not a single individ
ual member justifiedly holds this belief. Thus, the JAA handles classic diver
gence arguments with ease.
2.3 Problems for the Joint Acceptance Account
Objections may be raised to both the reliabilist and dialectical components

of the JAA-S and JAA-H, respectively Regarding the former, for instance, it
is not at all clear that reliability will appropriately track the inflationist’s
intuitions about epistemic justification. If the jury in the above case forms a
belief that the defendant is innocent purely on the basis of admissible evi
dence, is this produced by a reliable process? Clearly, no. Forming beliefs by
ignoring relevant evidence is a paradigm of an unreliable process, so how
will the joint acceptance theorist who is also a reliabilist achieve the desired
verdict in different evidence? With respect to the latter, there are
well-known objections to dialectical accounts of individual epistemic justi
fication that can be raised here, too. For instance, how is such a view going
10 I actually challenge this assumption below.

to handle really persuasive speakers who nonetheless offer bad epistemic

reasons, or highly gullible epistemic communities who very readily accept
poor defenses as successful ones?
But what I want to do here is challenge the core tenet of the JAA—namely,
the grounding of group reasons in joint acceptance. This will cut across
every existing inflationary, non-summative account of group justification in
the literature.
To begin, it will be helpful to clarify two features of the JAA. First, in
order for a group to be said to possess a reason, it cannot be required that all
members of the group be such that they would express openly11 a willing
ness to jointly accept r as the groups reason to believe that p. This would
make group justification too hard to come by. For instance, suppose that a
philosophy department is deliberating about whether to offer a position
in their graduate program to a highly qualified female applicant. All of the
members of the department would properly express a willingness to jointly
accept that the excellence of this womans writing sample is a reason to
believe that she should be admitted, except for one, whose sexism would
invariably prevent such agreement. The philosophy department here clearly
has a reason to believe that she should be admitted to the graduate program,
regardless of how we understand the nature of reasons. In particular, that a
sexist member of the department would never accept the excellence of this
womans writing sample as a reason to admit her affects neither whether the
group believes that her writing sample is excellent nor whether the writing
sample is in fact excellent. Given this, regardless of whether one thinks that
reasons are psychological states, factive states, or both, a single biased member
of a group steadfastly refusing to accept a given reason under any circum
stances is not sufficient for the group to lack that reason. Thus, the JAA
needs to be understood as requiring that only some of the members are such
that they would engage in the relevant joint acceptance.
Of course, it is not enough that this is true of just any members of the
group in question. Groups have members with vastly different roles, only
some of whom have the authority or power to determine certain outcomes
for the group as a whole. As we saw in the previous chapter, those who have
the relevant decision-making authority are often called operative members.12
For instance, the custodians of a law firm might have the authority to deter
mine whether the hallway traffic gives the firm a reason to install hardwood
11 For the sake of ease of expression, I will often drop the “openly.”
12 See, for instance, Tuomela (2004).
floors rather than carpeting, but not whether the details of a case provide a
reason to file a motion to dismiss on behalf of a client. The JAA should,
then, be understood as requiring not only that some members are such that
they would properly express a willingness to jointly accept r as the groups
reason to believe thatp, but also that these members are operative ones,13
With these points in mind, I now want to turn to what I take to be a
decisive objection to all versions of the joint acceptance account, one that
shows that the JAA makes group justification far too easy to come by.
Consider the following:
ignoring evidence: Philip Morris is one of the largest tobacco com

panies in the world, and each of its operative members is individually aware
of the massive amounts of scientific evidence revealing not only the addic
tiveness of smoking, but also the links it has with lung cancer and heart
disease. Moreover, each individual member believes that the dangers of
smoking give the company a reason to believe that warning labels should be
placed on cigarette boxes. However, because of what is at stake financially
and legally, none of these members would properly express a willingness to
accept that the dangers of smoking give Philip Morris a reason to believe
that it should put warning labels on cigarette boxes.
Does Philip Morris have a reason to believe that it should put warning labels
on cigarette boxes? Clearly, yes. Every member of this group is aware of
the scientific evidence showing the dangers of smoking and, accordingly,
believes that warning labels should be put on cigarette boxes. The mere fact
that the company is illegitimately ignoring relevant evidence through dog
matically and steadfastly refusing to jointly accept facts that are not to its
liking should not result in its not having this reason, too. This conclusion is
supported by the fact that we would surely hold Philip Morris responsible
for the ill effects caused by smoking precisely because we take it to have a
good reason to warn people about the dangers of cigarettes. Yet, according
to the JAA, Philip Morris does not have a reason to put warning labels on
cigarette boxes. Indeed, were the company to do so, it would be acting with
out a reason.
13 Thus, one of the conditions of Raimo Tuomelas inflationary account of group justifica
tion is that ‘'There is a special social justificatory dimension in that at least the operative group
members... must share a justifying joint reason for.. .p” (Tuomela 2004,113).
Consider, now, another case:
fabricating evidence: Philip Morris is one of the largest tobacco

companies in the world, and each of its operative members is individually
aware of the massive amounts of scientific evidence revealing not only the
addictiveness of smoking, but also the links it has with lung cancer and
heart disease. Entirely because of what is at stake financially and legally,
however, each of these members decides to jointly accept that all of the sci
entists working on the relationship between smoking and health problems
are liars. Given this, they also jointly accept that the duplicity of the scien
tists gives Philip Morris a reason to believe that the results of the studies
showing a connection between smoking and lung cancer and heart disease
are unreliable.
It is obvious that Philip Morris does not have a good reason to believe that
the results of studies showing a connection between smoking and health
problems are unreliable, but I think it is also clear that it doesn’t even have
a bad reason. To see this, notice that the members completely fabricate,
for purely financial and legal motives, that the scientists working on these
issues are liars, and thereby jointly accept that this provides the company
with a reason to reject the studies as unreliable. But surely this is not suffi
cient for a group to have a reason. One way to see this is that reasons, even
bad ones, are often taken to provide excuses for actions grounded in them.
Suppose that my student says that the reason she didn’t cite the sources on
which she relied is that she believed it wasn’t necessary to acknowledge
material assigned for our class. This reason might not justify her plagiarism,
but it does provide an excuse for her failure to cite the relevant sources. I
may, for instance, be able to understand her behavior, explain why such an
excellent student ended up engaging in academic dishonesty, and ultimately
hold her less responsible for the act than if she had knowingly failed to pro
vide the necessary citations. In fabricating evidence, however, there
is no sense whatsoever in which the members’ joint acceptance of a made-
up claim provides Philip Morris with an excuse for regarding the scientific
studies as unreliable. Indeed, rather than lessening responsibility, as a bad
reason might do, the company seems more guilty of the actions grounded in
the acceptance of the unreliability of the scientific evidence, since the fabri
cation of evidence was knowingly and willfully done. According to the JAA,
however, Philip Morris has a reason to believe that the results of the studies
showing a connection between smoking and lung cancer and heart disease
are unreliable.
It is just a small step from here to show that the JAA also leads to problem
atic results regarding the epistemic justification of group beliefs. Consider
ignoring evidence: given that all of the evidence showing that smoking
is dangerous is not available to the group because of the members’ refusal
to jointly accept it, none of it is part of the justificatory basis of the group’s
belief. It is, then, not at all difficult to imagine scenarios in which the
remaining evidence leaves the group justifiedly believing that smoking does
not pose any health hazards. For instance, the group might have access to
some studies that, though reliably conducted, had a very limited sample of
subjects, none of whom happened to develop lung cancer or heart disease
despite years of smoking. In this case, Philip Morris’s “belief” that smoking
is not unhealthy would be reliably formed, capable of successful defense,
and, given the total evidence available, well-grounded, thereby being epistem-
ically justified. But this result is absurd.
This is the sense in which the JAA makes the justification of group
beliefs far too easy to come by. Any group can manipulate the available
evidence through what it chooses to accept or reject, and thereby wind up
with beliefs that count as epistemically justified even when they clearly
are not.
The upshot of these considerations is that joint acceptance cannot ground
the justification of group beliefs. On the one hand, ignoring evidence
shows that the relevant kind of joint acceptance by the members of a given
group is not necessary for a group to possess a reason, since Philip Morris
has a reason to believe that warning labels should be placed on cigarette
boxes even in the absence of joint acceptance of that claim by its members.
On the other hand, fabricating evidence reveals that the relevant
kind of joint acceptance by the members of a given group is not sufficient for
a group to possess a reason, since Philip Morris does not have a reason to
believe that the results of the studies showing a connection between smok
ing and health problems are unreliable, despite there being joint acceptance
of that claim by its members. What both of these cases make clear is that
group justification cannot be wholly determined by factors over which the
members of the group have direct voluntary control. For it is this voluntary
control that enables the members of Philip Morris to simply decide to not
jointly accept what they should, and to jointly accept what they should not.
Because of this, it is possible for joint acceptance to be guided by factors that
are utterly disconnected from the truth, such as the economic and legal goals
of a company. Thus, any account of group justification that relies entirely on
joint acceptance succumbs to what I shall call the Illegitimate Manipulation

of Evidence Problem (IMEP):
IMEP: If the justification of group beliefs can be achieved through wholly

voluntary means, then the evidence available to the group can be illegitimately
manipulated, thereby severing the connection between group epistemic
justification and truth-conduciveness.
Given that the JAA clearly faces the IMEP, we need to look elsewhere for an
account of the justification of group beliefs.
Of course, the proponent of the JAA might substantially revise the view
so that there are epistemic constraints on both a groups having a reason and
the reason being a good one. For instance, perhaps a group has a reason, r,
to believe thatp if and only if its operative members would properly express
a willingness to jointly accept r as the group’s reason to believe thatp, where
“properly” is understood in distinctively epistemic terms. On this view, the
joint acceptance in question would have to be determined, not by the will of
the operative members, but by the evidence available to them. Otherwise,
there would be no way to ensure that Philip Morris has the relevant reason
in IGNORING EVIDENCE, but lacks it in FABRICATING EVIDENCE.
The problem with this approach, however, is that it ceases to be an infla
tionary non-summative account of group justification. To see this, consider
how this revised version of the JAA would handle ignoring evidence:
despite the fact that the operative members would not jointly accept that the
dangers of smoking give Philip Morris a reason to believe that it should put
warning labels on cigarette boxes, the group nonetheless has this reason
because of the evidence available to its members. But in what way is this a
joint acceptance account when all of the work is done by the available evidence
and joint acceptance is utterly irrelevant to whether the group has a reason?
This point can be put in the form of a dilemma: either group reasons are
determined by joint acceptance or they are not. If they are, the view suc
cumbs to the Illegitimate Manipulation of Evidence Problem. If they are
not, the view is not a joint acceptance account. Either way, inflationary non-
summativism is left wanting.
2.4 Revisiting Divergence Arguments
Divergence arguments provide the central grounding for an inflationary

approach to group justification. We have seen that the paradigmatic version
of such an approach—the joint acceptance account—has serious problems

that motivate its rejection. But where does that leave us vis-^-vis divergence
arguments? My goal in this section of the chapter is to argue that the two
main divergence arguments independently fail. This should just about close
the door to inflationism about group justification.
Let’s begin with different evidence. Recall that the standard inter
pretation of this sort of case is that, while the jury as a group justifiedly
believes that the defendant is innocent, none of the jurors justifiedly believe
this proposition because their justification is defeated by the relevant reli
able hearsay evidence. But why should we think that the notion of justifica
tion is epistemic in both evaluations? The reason that there might be the
inclination to say that the jury justifiedly believes that the defendant is
innocent is because hearsay evidence is deemed inadmissible by the court.
However, being inadmissible is clearly not the same as being unreliable or
otherwise non-truth-conducive. Consider, for instance, that hearsay evidence
is generally inadmissible because a witness needs to be “brought to testify in
court on the stand, where he may be probed and cross-examined as to the
grounds of his assertion and of his qualifications to make it” (Wigmore
1904, p. 437). The problem with hearsay evidence mentioned here is not that
it is more likely to be unreliable or lacking in evidential value, but rather
that the opposing side is denied the possibility of confronting the source of
the information. This is a practical or procedural concern, but not necessarily
an epistemic one. This is made clear by the fact that we can imagine a piece
of hearsay evidence that has been produced by a far more reliable process and
is better grounded in evidence than a piece of firsthand evidence. Nevertheless,
the former would be inadmissible in a court of law, while the latter would
not be. Given this, the mere fact that something is ruled inadmissible does
not necessarily reveal anything about its epistemic status.
Applying these considerations to different evidence, the reliable
hearsay evidence that the defendant was seen fleeing the scene of the crime
with blood spatter on his clothes is highly epistemically relevant to the jury’s
beliefs, even if the rules of the court prohibit it from being factored into
their verdict. This shows that while both the jury and the individual jurors
justifiedly believe that the defendant is guilty in an epistemic sense, the jury
is legally justified in believing that the defendant is innocent. This is because,
as was mentioned above, the law’s exclusion of hearsay evidence can be
radically disconnected from truth-conduciveness, which is precisely what we
find in different evidence. Thus, this case fails to establish what diver
gence arguments purport to show: namely, that the epistemic justification of
a group’s beliefs can diverge from the epistemic justification of the beliefs of
its individual members.14
Now Schmitt may respond to this argument by reminding us that a jury
is a chartered group and must therefore function according to its charter. As
he says, “the court rightly excludes hearsay, and its legal capacity is the only
capacity in which it operates” (Schmitt 1994, p. 274). Given this, insofar as
the jury considers the hearsay evidence in question and forms a belief in the
defendants guilt on this basis, it has ceased to be a jury. Thus, it is not the
case that the jury justifiedly believes that the defendant is guilty.15
But this response will not do. Surely, juries can make mistakes or break
the rules and still remain a jury. This happens with groups all the time.
A basketball team might break the rules by its players repeatedly double
dribbling the ball, and yet it still remains a basketball team playing basket
ball. It is just a bad basketball team playing a very poor game of basketball.
Similarly, a jury might consider hearsay evidence when forming its belief
about a defendant’s innocence or guilt and nonetheless remain a jury
engaged in deliberation. It is just a jury that has broken the rules. Moreover,
unlike the basketball team, so long as its verdict is grounded only in admis
sible evidence, it is not even clear that the jury is overall a bad one. Of
course, if a group breaks enough of the rules, or the right kind of rules, such
as those that are constitutive, it might cease to be the group in question. If
the players carry the ball across the court and never dribble it or attempt to
make a basket, then perhaps they no longer make up a basketball team. The
central point I wish to emphasize here, however, is that the mere fact that a
chartered group breaks a rule of its charter does not lead to the group no
longer existing. Given this, combined with the fact that the jury in differ
ent evidence is considering hearsay evidence only in the formation of
14 In response to this move, Kallestrup (2016) writes: “the key here is that the relevant
standards that govern different juries are epistemic in the sense that they fix the types and
strengths of evidence which can be brought to bear when juries reach a decision (or form a
belief). So, while a jury decision is strictly a legal act, its justification is an epistemic property of
that group. For such justification is a matter of the jury basing their decision on permissible
and strong enough evidence, which in turn is constrained by those standards. And because the
standards may differ from jury to jury, so will the epistemic properties of arriving at justified
decisions.” This response misses the point that the legal standard excluding hearsay is not
always truth-conducive, and so the justification in question is legal, not epistemic in nature.
The mere fact that the legal standards govern evidence is clearly not sufficient for rendering
beliefs that meet these standards epistemically justified. A corporation could adopt standards
of evidence that rule out considering scientific studies that challenge the safety of their prod
ucts. Surely, the beliefs that follow these standards would not thereby be epistemically justified.
15 lam grateful to Mark Thomson for raising this point.
its belief and not in issuing its verdict, there is no reason to conclude that it
is not the jury that believes that the defendant is guilty.
A further objection to Schmitts strategy for defending his reading of
different Evidence is that linking epistemic justification with the
charter of a group succumbs to a version of the Illegitimate Manipulation of
Evidence Problem. Schmitt considers the example of the charter of a court or
jury to exclude hearsay evidence as admissible, but there are no constraints
on the charters of groups. Given this, what prevents a group from being
formed whose primary charter is, for instance, to exclude any evidence that
conflicts with its belief that aliens have visited Roswell, New Mexico? In
such a case, the group could end up justifiedly believing that aliens have
visited Roswell simply because it is illegitimately restricting the available
evidence. Clearly, this is unacceptable.
Let’s now turn to different risk settings. Recall that Mathiesens
interpretation of this case is that while the hiring committee justifiedly
believes that Jones is a qualified candidate, not a single member justifiedly
holds this belief because none believe this proposition. This is due to the
fact that the members of the group are more epistemically cautious than the
group is as a whole. I want to challenge the claim that the diverging doxastic
states are both justified, but I want to do this by questioning the role of
epistemic risk settings. According to Mathiesen, such risk settings can be
determined by pragmatic factors. This is crucial to different risk set
tings, as the reason the hiring committee is less epistemically cautious is
because it has been given a charge by the department to present a set of
qualified candidates for the job. Given this, skepticism would be “out
of place” in the reasoning of the group. But this opens the door to a version
of the Illegitimate Manipulation of Evidence Problem: if epistemic risk set
tings can be determined by practical interests, and such settings can justify
different doxastic states, what prevents groups from manipulating their risk
settings precisely to suit their unwarranted practical purposes? For instance,
given the financial interests of Philip Morris, it certainly makes sense from a
practical point of view for it to be extraordinarily cautious when it comes to
accepting the testimony of scientists about the health hazards of smoking.
Given this, we can end up with Philip Morris being epistemically justified in
withholding belief about the dangers of smoking because of its extraordin
arily high standards for evidence, even when belief is clearly called for. This
shows that it is highly questionable whether risk settings can do the work
that Mathiesen needs them to.
Thus, we have compelling reasons to reject both the best examples of
the inflationary approach and the divergence arguments meant to support
them. This goes a long way toward closing the door on inflationary group
epistemology. In particular, the joint acceptance account is not only the
dominant version of inflationary non-summativism, but divergence argu
ments grounded in cases such as different evidence and different
risk settings are the primary defense offered for such an approach. If
this account and these arguments fail, then so does the central case for
inflationary non-summativism.
2.5 Deflationary Summativism, the Group Justification

Paradox, and the Defeater Problem
Given the serious problems facing an inflationary approach to understand

ing the justification of group beliefs that were developed in the previous
chapter, a natural response is to move toward a deflationary one. The most
widely accepted deflationary view is summativism, according to which the
justification of a group’s belief is understood simply in terms of the justifica
tion of the individual members’ beliefs. More precisely, there are two aspects
to such a view, corresponding to those found with respect to inflationary
non-summativism. The negative thesis is:
Deflationism: A group, G, justifiedly believing that p does not involve the

group itself justifiedly believing thatp, where this is over and above, or other
wise distinct from, the individual members of G justifiedly believing thatp.
The positive thesis is:
Summativism: A group, G, justifiedly believing that p is understood only

in terms of some or all of G’s members justifiedly believing thatp.
Deflationary summativism draws inspiration from a judgment aggregation

framework.16 As may be recalled from Chapter 1, “Aggregation procedures
are mechanisms a multimember group can use to combine (‘aggregate’) the
individual beliefs or judgments held by the group members into collective
16 While one can appeal to the resources of the judgment aggregation framework to support
a deflationary summativist view of justified group belief (as Alvin Goldman does below), it is
important to note that not all aggregation procedures support this approach. For instance, the
premise-based aggregation procedure discussed in Chapter 1 could be reframed in terms of
justified belief, rather than merely belief, and result in a case where the group holds a justified
belief that no member of the group does.
beliefs or judgments endorsed by the group as a whole” (List 2005, p. 25).17

For instance, a dictatorial procedure, “whereby the collective judgments are
always those of some antecedently fixed group member (the ‘dictator’)”
(List 2005, p. 28), understands the judgment of the group in terms of the
judgment of a single member—the dictator. A majority procedure, “whereby
a group judges a given proposition to be true whenever a majority of group
members judges it to be true,” understands the judgment of the group in
terms of the judgments of a majority of its individual members (List 2005,
p. 27). A supermajority procedure, whereby a group judges a given propos
ition to be true whenever a supermajority of group members judges it to
be true, understands the judgment of the group in terms of the judgments
of a supermajority of its individual members. And a unanimity procedure,
“whereby the group makes a judgment on a proposition if and only if the
group members unanimously endorse that judgment,” (List 2005, p. 30)
understands the judgment of the group in terms of the unanimous agree
ment of all of its members. Though there are obvious differences between
these views, they all characterize the judgment of a group in terms of the
judgments of the individual members.
This framework for aggregating member judgments into collective ones
can easily be extended to justified beliefs. Indeed, Alvin Goldman does just
this18 and, in so doing, provides the most detailed deflationary summativist
view to date.19
One of the first questions to address in developing such an aggregative
view of justified belief is how to understand the relationship between group
belief and group justifiedness. While Goldman doesn’t explicitly endorse a
particular account of group belief, he follows List and Pettit (2011) in work
ing within a framework in which group beliefs are the result of a function
17 For more on the theory of judgment aggregation, see List and Pettit (2002 and 2004),
Dietrich (2005), List (2005), Pauly and van Hees (2006), and Cariani (2011).
18 See Goldman (2014).
19 For views of collective knowledge that are summative in nature in one way or another, see
Corlett (1996 and 2007), Mokyr (2002), and Tuomela (2011). For instance, Mokyr (2002)
argues that, under the right circumstances, “society ‘knows’ something if at least one member
does” (p. 4), thereby espousing a sufficiency claim. In contrast, Tuomela focuses on necessity,
arguing that “a group cannot know unless its members or at least some of them know the item
in question” (Tuomela 2011, p. 85). At the same time, however, Tuomela argues that “when
[a group, g] believes that p, the members of g, collectively considered, will be assumed to
believe (accept) thatp when functioning as group members and thus be collectively committed
to p. Their private beliefs related to P (here coveringp and -p) can be different from those they
adopt as members of g” (Tuomela 2011, p. 86). Thus, for Tuomela, a group cannot know that p
without some of its members knowing that p, but a group can know thatp despite the fact that
none of its members privately believe thatp.
that takes profiles of individual members’ beliefs as inputs and yields collective
beliefs as outputs. Goldman calls such a mapping a belief aggregation func
tion, or BAE Examples of BAFs mirror those for judgments discussed
above; for instance, according to the majoritarian rule, a group believes that
p if and only if a majority of its members believe thatp, and so on.
What determines whether a BAF is an appropriate belief-forming rule
for a group? As we saw in Chapter 1, there are various possible answers to
this question. On one view, groups are able to select their own BAFs as they
see fit. On another, BAFs are determined entirely by the socio-psychological
forces that are operative within the groups structure without any choice or
input from the group itself. Goldman doesn’t commit himself to a particular
approach here.
What he does commit himself to, however, is that whatever account of
belief is endorsed, it does not bear a necessary connection to a theory of
group justifiedness. For instance, consider the following example of what
Goldman calls a justification aggregation function, or JAF:
JAF-1: If at least sixty percent of G’s members justifiedly believe that p,

then G too is justified in believing thatp. (Goldman 2014, p. 17)
JAF-1 is a supermajoritarian rule for group justifiedness, according to which

a group justifiedly believes that p if and only if a supermajority of its
members justifiedly believe that p. But, according to Goldman, BAFs and
JAFs can and often do diverge for a given group. For instance, while a
majoritarian BAF might be used for arriving at a group’s belief, it is none
theless perfectly acceptable for a supermajoritarian JAF to be relied upon
for determining this same group’s justifiedness. He writes:
Notice that JAF-1 mirrors the [supermajoritarian belief aggregation function

previously discussed]. There is no necessary connection, however, in the
sense that a JAF must always sanction,’ or approve of, whatever BAF a given
group selects. On the contrary, a given BAF may be one that a suitable JAF
would classify as unsuitable for generating justified group beliefs.
(Goldman 2014, p. 17)
According to Goldman, then, an account of group belief need not constrain

a theory of group justifiedness, nor need the latter constrain the former.
There are, however, reasons to doubt this claim. To see this, consider the
following:
disconnect: It is part of the bylaws of the Vegetarian Club at

Northwestern University (VCNU) that the elected President of the club
determines the beliefs for the entire group. Given this dictatorial BAF, com
bined with the Presidents true belief that vegan burgers are healthier than
hamburgers, it is the VCNU’s belief that vegan burgers are healthier than
hamburgers. At the same time, a supermajoritarian JAF is used for arriving
at the VCNU’s justifiedness in holding this belief. Since 60 of the 100 members
justifiedly believe that vegan burgers are healthier than hamburgers, the
VCNU justifiedly believes this. However, the President herself holds this belief
purely because of wishful thinking and is thus not among the 60 members
who believe this justifiedly.
In disconnect, a dictatorial BAF is combined with a supermajoritarian

JAF which, according to Goldman, seems to be a legitimate pairing. But two
results follow from this, both of which are problematic. First, the process or
basis responsible for the formation of the group’s belief is entirely discon
nected from the justifying features of this very belief. In particular, while
the President solely determines the VCNU’s belief, 60 different members of
the group affect whether this particular belief is justified. This is quite an
odd result. It would be on a par in the individual case with saying that while
reason is responsible for your believing thatp, its justifiedness is determined
entirely by testimony. Second, while the belief itself is formed through an
epistemically baseless and unreliable process—wishful thinking—it none
theless ends up being justifiedly held. This is tantamount to saying that the
origin of a belief can be wholly without epistemic significance to its justi
fiedness. Such a result should be regarded as epistemically unacceptable,
especially by a process reliabilist such as Goldman.
What is even more objectionable here, however, is that the door is left
wide open for Gettier cases to abound.20 For if group belief and group justi
fication can be determined by entirely different aggregation functions, then
not only can the process responsible for the formation of the group’s belief
be disconnected from its justifying features, the truth of the belief can also
be disconnected from its being justified. For instance, in disconnect, it is
simply a matter of luck that the President of the VCNU ends up with a true
belief that vegan burgers are healthier than hamburgers, since forming
beliefs purely on the basis of wishful thinking is surely not likely to result in
20 For the original Gettier cases, see Gettier (1963).

mostly true beliefs. Given this, it is similarly a matter of luck that the VCNU
ends up holding this true belief. But then there is no connection between
the truth of the group’s belief and its justifiedness, since the 60 members
who believe that vegan burgers are healthier than hamburgers for good
reasons have nothing at all to do with the formation of the group’s true
belief in the first place. This description of the situation perfectly parallels
the classic diagnosis of Gettier cases, and reveals the extent to which this
move of separating group belief and group justifiedness leaves group know
ledge vulnerable to being Gettiered.
Of course, an obvious solution to this problem is to require of any JAF
that it aggregates the justified beliefs of at least the very members of the
group who are responsible for the formation of the groups’ belief. But this is
just to deny the original claim that accounts of group belief and group justi
fiedness can float freely of one another. Thus, BAFs and JAFs must work in
synch with one another, lest cases like disconnect proliferate.21
With this in mind, lets turn to deflationary summativism. According to
Goldman, there are two different conceptions of group justification within
an aggregative framework—what he calls horizontal and vertical justified
ness. The best way to understand these notions is to consider the follow
ing case:
different bases: Gisa group whose members consist of 100 guards

at the British Museum, Each of the first 20 guards, M1-M20, justi
fiedly believes that guard Albert is planning an inside theft of a famous
painting (= A). By deduction from A, each of them infers the (existential)
proposition that there is a guard who is planning such a theft (= T). The
remaining 80 guards do not believe and are not justified in believing
A. Each of the second 20 guards, M21-M40, justifiedly believes that Bernard
is planning an inside theft (= B) and deductively infers T from B. The other
80 members do not believe B and are not justified in believing B. Each of a
third group of 20 members, M41-M60 justifiedly believes that guard Cecil is
planning an inside theft (= C) and deductively infers T from C. The 80
others do not believe and are not justified in believing C. Thus, 60 members
of G (justifiedly) believe T by deduction from some premise he/she justi
fiedly believes. (Goldman 2014, p. 16)
21 Goldman’s process reliabilism might have the resources for responding to this problem,
but only by virtue of a de facto connection between BAFs and JAFs. Thus, my point still holds
that these two aggregation functions cannot work independently of one another.
Most of the leading aggregation procedures—for example, supermajoritarian

and majoritarian—have the result that G believes T. But does G justifiedly
believe this proposition? Goldman writes:
...G’s belief in T maybe considered from two perspectives: the horizontal

perspective and the vertical perspective. The horizontal perspective
addresses the question of the J-status of Gs belief in T solely in terms of
other beliefs of G, i.e„ group-level beliefs.... G’s belief in T is unjustified in
terms of horizontal J-dependence. This is because, although G believes T, G
does not infer T from any justified group-level belief of its own. The situ
ation is different, however, when we consider G s belief in T by reference
to vertical J-dependence. Consider all of the members’ beliefs in T and the
proportion of them that are justified....given...[the] vertical criterion of
J-dependence, G’s belief in T is justified (because 60% of G’s members justi
fiedly believe T). (Goldman 2014, p. 18)
On Goldmans view, then, G justifiedly believes that someone is planning an

inside theft at the museum when vertical justifiedness is considered, that is,
when the justificational status of the groups belief is determined, not by the
group’s beliefs, but by all of the members’ relevant beliefs and the propor
tion of them that are justified. In particular, because the individual mem
bers have different bases for their beliefs that someone is planning a theft,
there is no group-level basis from which the group’s belief to this effect can
be justifiedly inferred. Hence, there is no horizontal justification. But if the
group’s belief that someone is planning a theft is viewed independently of a
group-level basis, then the proportion of the members’ relevant beliefs that
are justified render it vertically justified. Indeed, it is precisely this vertical
perspective that Goldman adopts when offering his positive account of the
justification of group beliefs.
Moreover, Goldman claims that it is preferable to think of justifiedness
as a matter of degree, and thus to regard it as a gradable notion instead of
a categorical one. Rather than sketch a full-blown theory of justificational
gradability for collective entities, however, he offers a few sample principles
so as to give a sense of the results such a theory will deliver. Assuming that
members’ doxastic attitudes have categorical justificational status, a central
principle is the following:
(GJ) If a group belief that p is aggregated based on a profile of member

attitudes toward that p, then (ceteris paribus) the greater the proportion of
members who justifiedly believe that p and the smaller the proportion of
members who justifiedly reject thatp, the greater the group’s level, or grade,
of justifiedness in believing thatp. (Goldman 2014, p. 28)
On Goldmans view, the justificational statuses of members’ doxastic atti

tudes depend on the processes by which they severally arrived at their
respective attitudes and, as (GJ) makes clear, the justificational status of the
group belief depends on the justificational statuses of the members’ atti
tudes. Put succinctly, group justifiedness increases with a greater percentage
of individual member justifiedness.22
(GJ) is not only intuitively plausible, it also is easily supported by apply
ing an aggregative framework to group justifiedness. Despite this, I will
argue in what follows that (GJ) should be rejected, as it leads to what I call
the Group Justification Paradox and the Defeater Problem.
To begin, let us compare different bases with the following version
of the case:
conflicting bases: G is a group whose members consist of 100

guards at the British Museum, M -M , each of whom justifiedly believes
that an inside theft of a famous painting is being planned by only one of a
total of five possible guards—Albert, Bernard, Cecil, David, and Edmund.
Each of the first 20 guards, M -M20, justifiedly believes that only guard
Albert is planning the inside theft (= A). By deduction from A, each of them
infers the (existential) proposition that there is a guard who is planning
such a theft (= T). The remaining 80 guards do not believe and are not
justified in believing A. Each of the second 20 guards, M21-M40, justifiedly
believes that only Bernard is planning the inside theft (= B) and deductively
infers T from B. The other 80 guards do not believe and are not justified in
believing B. Each of a third group of 20 members, M41-M60 justifiedly
believes that only guard Cecil is planning the inside theft (= C) and deduct
ively infers T from C. The 80 others do not believe and are not justified in
believing C. Each of a fourth group of 20 members, Mgl-M80 justifiedly
believes that only guard David is planning the inside theft (=D) and
deductively infers T from D. The remaining 80 guards do not believe and
22 An immediate problem with (GJ) is that the degree (or level or grade) to which any indi
vidual member of the group justifiedly believes that p does not play any role in determining the
degree (or level or grade) of the groups justifiedness in believing thatp; only the proportion of
members who justifiedly believes thatp is relevant. I am grateful to an anonymous reviewer for
raising this point.
are not justified in believing D. The final group of 20 members, M81-M100

justifiedly believes that only guard Edmund is planning the inside theft
(= E) and deductively infers T from E. The 80 others do not believe and are
not justified in believing E. Thus, 100 members of G justifiedly believe T by
deduction from some premise he/she justifiedly believes.23
In the original different bases, 60 out of 100 members of G justifiedly

believe that there is a guard who is planning an inside theft of a famous
painting at the museum and, thus, the group itself justifiedly believes this
proposition. In conflicting bases, 100 out of 100 members of G justi
fiedly believe that there is a guard who is planning an inside theft of a famous
painting at the museum and, thus, the group again justifiedly believes this
proposition. According to (GJ), then, the group’s level of justifiedness is
greater in conflicting bases than it is in different bases since the
proportion of members who justifiedly hold the relevant proposition in the
former is greater than in the latter.
But let us take a closer look at conflicting bases. Each of the first
20 guards, Mx-M , justifiedly believes that only guard Albert is planning
the inside theft. Given this, combined with the fact that all of the guards are
aware that Albert, Bernard, Cecil, David, and Edmund are the only possible
thieves, each of these 20 guards also justifiedly believes that Bernard,
Cecil, David, and Edmund are not planning the theft. Each of the second
20 guards, M21-M40, justifiedly believes that only Bernard is planning the
inside theft and, given their other background beliefs, also justifiedly
believes that Albert, Cecil, David, and Edmund are not planning the theft.
Similar considerations apply with respect to the other three subgroups: each
believes that one, and only one, guard is planning the theft, and believes
that the other four possible guards are not so planning.24
We are now in a position to see the Group Justification Paradox unfold:
for each of the five possible candidates of the theft in question, 80 out of
100 guards justifiedly believe that he is not planning it. According to nearly
every judgment aggregation function, it follows from this that the group, G,
also justifiedly believes that each of the five possible candidates is not plan
ning the theft. Since the group justifiedly recognizes that Albert, Bernard,
Cecil, David, and Edmund are the only possible candidates for planning the
23 This case is similar to those involving base fragility discussed in the Chapter 1.
24 For further discussion of some of the issues surrounding groups with conflicting bases,
see Cariani (2013).
theft, the group justifiedly believing that none of them is planning the theft
amounts to the group justifiedly believing that no one is planning the theft.
But, according to the vertical perspective, G also justifiedly believes that
someone is planning an inside theft of a famous painting at the British
Museum since 100 members justifiedly believe this. (GJ) thus leads to what
we might call the Group Justification Paradox: G ends up justifiedly
believing both that no one is planning the theft and that someone is plan
ning the theft.25
Now, it might be noticed that this paradox relies on accepting that con
junction is closed for justified group belief. In particular, the group justi
fiedly believes that each of the five guards in question is not planning the
theft. Thus, G believes that it is not Albert, not Bernard, not Cecil, and so
on. Let us represent the group’s justified beliefs here as follows:
~A
~B
~C
~D
~E
In addition, the group justifiedly believes that someone is planning the theft
and that Albert, Bernard, Cecil, David, and Edmund are the only five pos
sible candidates. Thus, G justifiedly believes:
(AvBvCvDvE)
The contradiction found in the Group Justification Paradox is then gener

ated by the closure of conjunction, that is, by moving from the first set of
justified group beliefs above to the following:
(~A & ~B & ~C & ~D & ~E)
This can be rewritten as the group justifiedly believing:
-(AvBvCvDvE)
25 It should be noted that the Group Justification Paradox is analogous to the general result
in judgment aggregation theory that no supermajority rule short of unanimity will always
secure a deductively closed and consistent set of collective attitudes. (I am grateful to an
anonymous referee for this point.)
Thus, the result is that G justifiedly believes both (AvBvCvDvE) and

~(A v B v C vD vE), that is, both that someone is planning the theft and
that no one is planning the theft. Hence, a contradiction. Given this, a pro
ponent of (GJ) might respond to the Group Justification Paradox by deny
ing that conjunction is closed for justified group belief, thereby avoiding the
contradiction in question.
But notice that even if the closure of conjunction is rejected, G is still left
with an obviously inconsistent set of beliefs, even if not an outright contra
dictory one. Specifically, since all 100 guards justifiedly believe that some
one is planning the theft, the group justifiedly believes this, too. Moreover, all
100 guards justifiedly believe that the thief must be either Albert, Bernard,
Cecil, David, or Edmund and thus that one of the propositions—A, B, C, D,
or E—must be true. Given the vertical picture of justification, it follows that
the group also believes one of these propositions is true. Thus, G justifiedly
believes:
(AvBvCvDvE)
With respect to each of these five propositions, however, 80 out of 100

guards justifiedly believe it is false. On nearly every judgment aggregation
function, this means that the group itself justifiedly believes that A, B, C, D,
and E are false. Hence, G also justifiedly believes:
If the vertical picture is correct, then, G justifiedly believes an obviously

inconsistent set of propositions, which is enough for the Group Justification
Paradox to undermine (GJ).
It should be further noted that unlike with some other paradoxes, this is
not an inconsistent set of beliefs that it is nevertheless reasonable to have.
For example, the Preface Paradox envisions an author apologizing for the
errors that are contained in her book. In so doing, the author has ensured
that there is at least one error in the book because she is now committed
to an inconsistent set of claims: each of the individual claims made in
the book, plus the claim that at least one of them is false. Nevertheless, the
authors apology in the preface is epistemically reasonable, given the

excellent grounds we all have for our own fallibility.26 But notice: the author
in the Preface Paradox does not have any particular reason to revise one
claim rather than another. Looking at the grounds for holding any particular
belief, the author would weigh the evidence in its favor against the very small
chance that it is incorrect. Given this, were the author to think through
the matter, she wouldn’t change any of her beliefs. This is not the case with the
Group Justification Paradox, though. For each of the claims that one of the
five guards is planning the inside theft, 80 members of the group do have
evidence against it that they would present. Were the group to collectively
deliberate, it would have a great deal of work to do before it could reach a
stable position. Moreover, with respect to the Preface Paradox, we can sup
pose that the beliefs are largely independent, in the sense that they can be
accepted or rejected without this having any implication for the other claims
in the book. But this is not the case with the Group Justification Paradox:
accepting one sub-group’s claims necessarily means rejecting the claims of
the other sub-groups. For this reason, they can’t all be comfortably encom
passed in a single point of view. Finally, the author in the Preface Paradox
can act consistently by accepting each claim individually while also not, for
instance, betting on all of the claims being true. The group in the Group
Justification Paradox, however, cannot act consistently. In discussion with
the police, for instance, the group will be both advising that one of the five
guards is planning the theft and then ruling out each of them as the suspect.
Thus, while the inconsistency in some paradoxes might be rationally toler
able, the inconsistency in the Group Justification Paradox is not.
Thus, that (GJ) succumbs to the Group Justification Paradox is sufficient
for calling this view into question. However, reflecting on conflicting
bases also enables us to see that (GJ) is false. For despite the fact that the
proportion of members in conflicting bases who justifiedly believe
that someone is planning the theft in question is greater than the proportion
of members in different bases who justifiedly believe this, the group’s
level of justifiedness is lower in the former than it is in the latter. This is
because the group in conflicting bases has a defeater for believing that
someone is planning an inside theft of a famous painting at the museum,
but not in different bases. Let’s call this the Defeater Problem for (GJ).
26 For the original Preface Paradox, see Makinson (1965). See also the Lottery Paradox in
Kyburg (1961). See Klein (1985), Foley (1987), Christensen (2004), and Makinson (2012) for
arguments that consistency is not a requirement of individual rational belief.
To see this, we should first take a brief detour through of defeaters. There
are two central kinds of defeaters that are typically taken to be incompatible
with justification. First, there are what we might call psychological defeaters,
which can be either rebutting or undercutting, A psychological defeater is a
doubt or belief that is had by S, and indicates that S’s belief that p is either
false (i.e., rebutting) or unreliably formed or sustained (i.e., undercutting).
Defeaters in this sense function by virtue of being had by S, regardless of
their truth-value or epistemic status.27 Second, there are what we might
call normative defeaters, which can also be either rebutting or undercutting.
A normative defeater is a doubt or belief that S ought to have, and indicates
that S’s belief that p is either false (i.e., rebutting) or unreliably formed or
sustained (i.e., undercutting). Defeaters in this sense function by virtue of
being doubts or beliefs that S should have (whether or not S does have
them), given the presence of certain available evidence.28 The underlying
thought here is that certain lands of doubts and beliefs—either that a sub
ject has or should have—contribute epistemically unacceptable irrationality
to doxastic systems and, accordingly, justification can be defeated or under
mined by them.
Moreover, a defeater may itself be either defeated or undefeated. Suppose,
for instance, that Harold believes that there is a bobcat in his backyard
because he saw it there this morning, but Rosemary tells him, and he
thereby comes to believe, that the animal is instead a lynx. In such a case,
the justification that Harold had for believing that there is a bobcat in his
backyard has been defeated by the rebutting belief that he acquires on the
basis of Rosemary’s testimony. But since psychological defeaters can them
selves be beliefs, they, too, are candidates for defeat. For instance, suppose
that Harold consults a North American wildlife book and discovers that the
white tip of the animal’s tail confirms that it was indeed a bobcat, thereby
providing him with a defeater-defeater for his original belief that there is a
bobcat in his backyard. And, as should be suspected, defeater-defeaters can
27 For various views of what I call psychological defeaters see, for example, Bonjour (1980
and 1985), Nozick (1981), Goldman (1986), Pollock (1986), Plantinga (1993), Bergmann
(1997), Reed (2006), and Lackey (2008).
28 For discussions involving what I call normative defeaters, approached in a number of
different ways, see Bonjour (1980 and 1985), Goldman (1986), Fricker (1987 and 1994),
Chisholm (1989), Burge (1993 and 1997), McDowell (1994), Audi (1997 and 1998), Williams
(1999), Bonjour and Sosa (2003), Hawthorne (2004), Reed (2006), and Lackey (2008). What all
of these discussions have in common is simply the idea that evidence can defeat knowledge
(justification) even when the subject does not form any corresponding doubts or beliefs from
the evidence in question.
also be defeated by further doubts and beliefs, which, in turn, can be

defeated by further doubts and beliefs, and so on. Similar considerations
involving evidence, rather than doubts and beliefs, apply in the case of nor
mative defeaters. When one has a defeater for one’s belief that p that is not
itself defeated, one has what is called an undefeated defeater for one’s belief
that p. It is the presence of undefeated defeaters, not merely of defeaters,
that is incompatible with justification and, thus, knowledge.
With these points in mind, let us return to the scenario in conflicting
bases. It should be fairly clear that G has a rebutting psychological defeater
for believing that someone is planning an inside theft of a famous painting
at the museum that is not itself defeated. In particular, G’s belief that no one
is planning such a theft indicates that G’s belief that someone is planning such
a theft is false, and hence the target belief’s justification has been defeated.29
This means that G’s belief that someone is planning a theft is unjustified,
despite the fact that every member of G justifiedly believes this. In contrast,
there is no reason to regard the group’s belief as defeated in different
bases in the same way. For even if we add to the original case that each of
the 100 guards justifiedly believes that the inside theft is being planned by
only one of a total of five possible guards—Albert, Bernard, Cecil, David,
and Edmund—the most we get is that 40 percent believe that it is not Albert,
40 percent believe that it is not Bernard, 40 percent believe that it is not
Cecil, 60 percent believe that it is not David, and 60 percent believe that it is
not Edmund. This in no way provides the group with the belief that no one
is planning an inside theft at the museum. Given this, G’s level of justifiedness
is greater in different bases than it is in conflicting bases, despite
the fact that the proportion of members in the former who justifiedly believe
that someone is planning the theft in question is lower than the proportion
of members in the latter who justifiedly believe this. The Defeater Problem
thus shows that (GJ) is false.30
29 It might be thought that the Defeater Problem undercuts the Group Justification Paradox.
For if we can take one of the beliefs to be defeated, perhaps there isn’t a paradox after all. By
way of response, notice that every paradox with contradictory beliefs can be seen as involving
rebutting defeaters, but this doesn’t make them any less paradoxical.
30 It is worth noting that, though Goldman includes a “ceteris paribus” clause in (GJ), these
problems cannot be subsumed under it. For (GJ) just is a reflection of the notion of vertical
justification: according to this principle, group justifiedness increases with a greater percentage
of individual member justifiedness, and this just amounts to permitting group justifiedness to
be determined independently of a group-level basis for that justification. But recall that what
motivated the conception of vertical justification in the first place was different bases.
Given that conflicting bases is an extension of this paradigmatic instance of vertical jus
tification, there is no plausible sense in which it can be relegated to the ceteris paribus clause.
2.6 The Collective Evidence Problem
We have seen, then, that (GJ) succumbs to both the Group Justification
Paradox and the Defeater Problem. So where does this leave us? At the very
least, the two problems afflicting (GJ) show that group justifiedness cannot
be determined by aggregating individual member justifiedness independ
ently of the relevant bases. Indeed, it is precisely because (GJ) permits such
independence that the problems stemming from conflicting bases
arise. For if group justifiedness were a matter of aggregating members’ doxastic
states plus their bases, then the group in conflicting bases wouldn’t
justifiedly believe that someone is planning an inside theft of a famous
painting at the British Museum since there is no single justified belief + base
combination that is had by at least a majority of the group’s members. But if
the group doesn’t hold the justified belief that someone is planning a theft,
then there is no paradox and there is no belief to be defeated.
The problem with this strategy, however, is that it also has the result that
the group in different bases doesn’t justifiedly believe that someone
is planning a theft at the British Museum, which was the very case used
to motivate vertical justification. So, if the goal is to avoid the Group
Justification Paradox and the Defeater Problem while also retaining the
notion of vertical justifiedness for groups, we need to look elsewhere.
To this end, let’s consider the weakest conclusion that can be drawn from
these two problems. Consider this: the central feature that distinguishes
different bases from conflicting bases is that, as their names
suggest, the bases of the individual members’ beliefs conflict in the latter
but not necessarily in the former. For as the original case is described,
20 percent of the group believes that Albert is planning a theft of the
museum, 20 percent believes that Bernard is planning a theft, and 20 percent
believes that Cecil is planning a theft. By deduction, each of these groups
infers the (existential) proposition that there is a guard who is planning such
a theft. But since it is not built into the case that they believe that only one
guard is planning such a theft, it is open for them to believe that more than
one guard is. Thus, all 60 out of 100 guards might be correct in their respective
beliefs because it is possible that Albert, Bernard, and Cecil are together
planning a theft, different bases, then, does not necessarily involve
bases that conflict. So, the weakest conclusion that can be drawn from the
Group Justification Paradox and the Defeater Problem is that group justi
fiedness cannot aggregate individual member justifiedness when the latter
involves conflicting bases.
Perhaps, then, the spirit of (GJ) can be saved by revising it as follows:
(GJ1) If a group belief that p is aggregated based on a profile of member

attitudes toward that p and the individual members’ bases for believing that
p are non-conflicting, then (ceteris paribus) the greater the proportion of
members who justifiedly reject that p, the greater the groups level, or grade,
of justifiedness in believing that p,
(GJ1) preserves group justifiedness in different bases—at least when it

is read in the way specified above—but rules it out in conflicting
bases, which is exactly what the proponent of vertical justifiedness needs.
But now consider the following:
non-conflicting bases: G is a group whose members consist of 100

guards at the British Museum, Mx-M , each of whom justifiedly believes
that a man was responsible for an inside theft of a famous painting. Each of
the first 20 guards, MT-M20, justifiedly believes that the thief exited a men’s
bathroom right before the theft (= B). From B, each of them infers the prop
osition that it was a man who committed the theft (= M). The remaining 80
guards do not believe and are not justified in believing B. Each of the second
20 guards, M21-M40, justifiedly believes that the thief has a goatee (= G) and
infers M from G. The other 80 members do not believe and are not justified
in believing G. Each of a third group of 20 members, M41-M60, justifiedly
believes that the thief was greeted as “sir” (= S) while walking into the
museum and infers M from S. The 80 others do not believe and are not justi
fied in believing S. Each of a fourth group of 20 members, M61-M80, justifiedly
believes that the thief was talking in a baritone voice (=V) and infers M
from V. The remaining 80 members do not believe and are not justified in
believing V. The final group of 20 members, , justifiedly believes
that the thief’s name is William (= W) and infers M from W. The other
80 members do not believe and are not justified in believing W. Thus,
100 members of G justifiedly believe M by inference from some premise
he/she justifiedly believes.
At the same time, however, each subgroup of 20 guards also has counterevi
dence for the basis of the justified beliefs of a different subgroup. M1-M20
justifiedly believe that not-G since they have evidence that the thief’s goatee
is fake. M21-M40 justifiedly believe that not-B since they have evidence that
the bathroom that the thief was seen exiting is in fact a family bathroom.
M41~M60 justifiedly believe that not-W since they have evidence that the
thief has been using a pseudonym. M61-M80 justifiedly believes that not-S
since they have evidence that it was actually the thief’s companion who was
greeted as ‘sir” upon entering the museum. Finally, Mgl-M100 justifiedly
believe that not-V since they have evidence that the baritone voice heard at
the scene of the crime wasn’t the thief’s but was instead a recording.
As the name suggests, the bases of the members’ beliefs in this case are non
conflicting and, indeed, they are mutually supporting. In particular, when
B, G, S, V, and W are taken together, they provide powerful support—far
more than any piece of evidence taken in isolation—for concluding that the
person responsible for an inside theft of a famous painting at the British
Museum is a man. Thus, non-conflicting bases clearly satisfies the
relevant part of the antecedent of (GJ1) and results in the group justifiedly
holding this belief about the thief.
But notice the second part of the case: in addition to the bases of the
members’ beliefs being non-conflicting, each subgroup of 20 guards has
counterevidence for the basis of the justified beliefs of a different subgroup.
So, for instance, while the basis for guards M1-M20 justifiedly believing that
it was a man who committed the theft is that the thief exited a men’s bath
room, they also justifiedly believe that the thief’s goatee is fake, which is
counterevidence for the basis of the belief for guards M21-M40. Similar con
siderations apply to all of the subgroups, resulting in there being no basis
that does not have 20 percent of the group justifiedly believing its negation.
To my mind, this results in the group of guards in non-conflicting
bases not justifiedly believing that a man was responsible for an inside
theft of a famous painting at the British Museum. This is because there is
not a single basis of the members’ beliefs that is free of direct and compel
ling counterevidence. Otherwise put, there is no basis that would survive
full disclosure: were all 100 members to fully disclose all of their evidence
and counterevidence, there would be no remaining reason to believe that the
thief is a man. Despite this, since 100 members of G justifiedly believe M,
(GJ1) grants a very high level of justifiedness to G in holding this belief.
It also is not difficult to imagine a variant of non-conflicting
bases—let’s call it variant—that is exactly as the original case is described,
except for two modifications: (i) only 60 of the 100 guards justifiedly believe
that a man was responsible for an inside theft of a famous painting at the
British Museum, and (ii) none of the 100 members possess counterevidence
for any of the relevant bases. According to (GJ1), G has a lower level of
justifiedness in the belief that the thief is a man in variant than in

non-conflicting bases because there is a significantly greater pro
portion of members who justifiedly hold this belief in the latter than there is
in the former. But this seems wrong. In particular, even though there are
fewer members who justifiedly hold the relevant belief in variant, there is
far greater epistemic support for the group’s belief in this case than there
is in non-conflicting bases when the collective entity is viewed as
a whole. Let us call this general objection to (GJ1) the Collective Evidence
Problem.
It might be thought that there is an easy solution to this problem. For
why can’t the counterevidence in non-conflicting bases be under
stood in terms of the group possessing an undefeated defeater for believing
that the thief is a man? Accordingly, why can’t (GJ1) simply be modified to
rule this out as follows:
(GJ2) If a group belief that p is aggregated based on a profile of member

attitudes toward thatp and the individual members’ bases for believing that
p are non-conflicting, then (ceteris paribus) the greater the proportion of
members who justifiedly reject thatp, the greater the group’s level, or grade,
of justifiedness in believing that p, so long as the group does not possess an
undefeated defeater for believing thatp.
If (GJ2) is combined with the claim that G has an undefeated defeater for
believing that the thief is a man, then perhaps the spirit of vertical justifica
tion can be retained while having the resources for denying justified belief
to G in NON-CONFLICTING BASES.
There is, however, a central problem with this strategy; namely, that it is
not at all clear how to understand the group having an undefeated defeater
in non-conflicting bases within the vertical justification framework.
To see this, recall that the paradigmatic instance of vertical group justified
ness is when members’ justifiedness is aggregated without taking into
account the corresponding bases. Thus, a group can end up highly justified
in believing thatp, so long as a significant percentage of members justifiedly
believe that p, regardless of what it is that justifies the members’ beliefs. The
corresponding view of vertical group defeat, then, would permit the group
to have a defeater for believing that p, so long as a significant percentage of
the members have a defeater for believing that p, regardless of what it is that
does the defeating work in question. So, for instance, on this view, a group
might have a defeater for believing that p even when the members have
quite different counterevidence regarding whether p, just as a group might
justifiedly believe that p even when members have quite different reasons
for believing that p. But notice: this is not at all what is found in non
conflicting bases. In this case, not a single member of the group has a
defeater for believing that the thief is a man since not a single member has
counterevidence for his particular basis for this belief. If group justification
and, correspondingly, group defeat is understood in the atomistic aggrega
tive way at the heart of vertical justification—where doxastic states of members
are aggregated independently of their relations to other doxastic states—
then there seems to be no sense in which the group has a defeater for
believing that the thief is a man in non-conflicting bases.
Of course, the proponent of (GJ2) might simply introduce an altogether
different account of group defeat, one that is able to accommodate G’s belief
being defeated in non-conflicting bases. But it should be clear that
whatever account of group defeat is offered such that the groups belief that
the thief is a man ends up defeated, it will be working with a wildly different
conception of group justifiedness than that found in the vertical concep
tion. For as discussed above, there is no sense in which the groups belief is
vertically defeated in non-conflicting bases. Thus, this strategy will
not save the aggregative version of deflationary summativism from the
Collective Evidence Problem.
2.7 The Group Normative Obligations Problem
We have seen that modifications to (GJ) fail to solve the problems afflicting
the aggregative account of group justifiedness. There is one further objec
tion facing this view that I would like to develop, which can be seen by con
sidering the following:
group normative obligations: G is a group whose members con

sist of 3 nurses employed at a nursing home, Nt-N , each of whom justi
fiedly believes that patient O’Brien is not at risk of dying. N is aware that
she forgot to give O’Brien his first medication, but she also justifiedly
believes that this act of negligence alone is not sufficient to put him in dan
ger of death. N2 is aware that she forgot to give O’Brien his second medica
tion, but she also justifiedly believes that this act of negligence alone is not
sufficient to put him at risk of serious health problems. And N3 is aware that
she forgot to give O’Brien his third medication, but she also justifiedly
believes that this act of negligence alone is not sufficient to put his life in
jeopardy At the same time, however, NX~N3 all justifiedly believe that
O’Brien missing all three of his medications would put him at serious risk
of dying.
Moreover, it is an explicit requirement of the collective unit comprising the
positions held by N1-N3 that they always communicate with one another
about the patients for whom they mutually care. Despite this, N1-N3 do not
share their respective acts of negligence with one another, and so each justi
fiedly believes that the other nurses successfully gave O’Brien his medicine.
Thus, through failing to fulfill their responsibilities qua group members,
Nt-N3 lack crucial evidence that they should have had and that would
reveal the epistemic deficiency of their beliefs that O’Brien is not at risk
of dying.
Given the evidence available to N -N individually, each justifiedly believes

that O’Brien is not at risk of dying. Moreover, as individual epistemic agents,
N -N3 are not neglecting any epistemic duties. In particular, subtract their
membership in G and they have no obligation, epistemic or otherwise, to
discuss O’Brien’s care with one another. Indeed, this can even be built
directly into the case—perhaps because of concerns about the unreliability
of gossip, nurses are prohibited from discussing patients’ care, unless they
are members of a unit.
As a group, however, matters are quite different. Given the normative
obligations that N -N3 are bound by as members of their nursing unit, they
have an epistemic duty to consult with one another about O’Briens care.
Were they to do what epistemically they ought to, they would all be aware of
the fact that he failed to receive all three of his medications, and thus they
would each justifiedly believe that O’Brien is at risk of dying. But failing to
possess evidence through neglecting one’s epistemic duties does not get one
off the hook for being responsible for such evidence, either as an individual or
as a group. If one ought to be aware of evidence, this is enough for preventing
epistemic justification, regardless of whether one in fact possesses it. Thus,
while every member of the nursing unit justifiedly believes that O’Brien is at
risk of dying, the group itself does not. Let us call this objection to the
aggregative account the Group Normative Obligations Problem.
This point can be put in terms of defeaters. Recall that a normative
defeater is a belief that a subject ought to have, and that indicates that the
target belief is either false or unreliably formed or sustained. Normative
defeaters function by virtue of being beliefs that a subject should have,

whether or not she does have them. Given this, we might say that even though
Nj-N do not, qua individuals, possess normative defeaters for believing
that O’Brien is not at risk of dying, they do have such defeaters qua group
members. This is because, absent their membership in the group, there is no
basis for saying that they epistemically ought to believe that O’Brien failed
to receive all three of his medications. However, given their group member
ship, there is such a basis for saying this, and such a belief clearly indicates
that their belief that O’Brien is not at risk of dying is false.
It is worth pointing out that though the normative obligations in ques
tion arise by virtue of the nurses’ membership in the nursing unit, they are
nonetheless epistemic rather than merely professional or prudential. To see
this, notice that one’s professional roles often give rise to distinctively
epistemic duties. If one is a medical doctor, for instance, one might be pro
fessionally obligated to consult test results before determining a diagnosis
for one’s patients. This duty arises out of ones role as a physician, but it is surely
epistemic. This is made clear by the fact that it concerns evidence highly
relevant to the belief in question—in this case, the diagnosis—that bears on
whether one has the corresponding knowledge. Similar considerations apply
in group normative obligations: the duty to consult with other
members of the nursing unit regarding the mutual care of O’Brien arises out
of the nurses’ professional roles, but the obligation is epistemic in nature.
In particular, it concerns evidence that the nurses know is highly relevant
to their beliefs about O’Brien’s health that directly affects whether they
know he is not at risk of dying.
Of course, not all distinctively epistemic duties that arise out of profes
sional roles are such that their flouting prevents the possession of know
ledge. For instance, the nursing unit might be required by the administration
to form all of their beliefs about patients via only logical deduction. While
this requirement might ensure a very high degree of epistemic support, it is
far more stringent than what is needed for their beliefs to amount to know
ledge. At the very least, when the epistemic duty in question concerns evi
dence without which the belief in question will be irrationally held, its
flouting will prevent the possession of the corresponding knowledge.
Clearly, this is the case in group normative obligations, where the
nurses continuing to believe that O’Brien is not at risk of dying is irration
ally held when they all know that their co-workers have evidence that bears
directly on this very matter.
Moreover, as was the case with non-conflicting bases, it is not dif
ficult to imagine a variant of group normative obligations—let’s
call it variant2—that is just as the original case is described, except for

two modifications: (i) only 2 of the 3 nurses justifiedly believe that O’Brien
is not at risk of dying, and (ii) there is no requirement of the collective unit
comprising the positions held by N -N3 that they always communicate with
one another about the patients that they mutually care for. According to
(GJ), G has a lower level of justifiedness in the belief that O’Brien is not at
risk of dying in variant2 than in group normative obligations
because there is a greater proportion of members who justifiedly hold this
belief in the latter than there is in the former. But this is wrong. For even
though there are fewer members who justifiedly hold the relevant belief in
variant 2, none of them are neglecting any epistemic duties, as they are in
group normative obligations. This means that the group has a
greater degree of justifiedness in the belief that O’Brien is at risk of dying in
variant2 than in group normative obligations, despite the fact
that there is a greater proportion of members who justifiedly hold this belief
in the latter. The aggregative account of group justifiedness is, again, shown
to be false.
2.8 A Condorcet-Inspired Account of Justified

Group Belief
Before turning to my positive view, there is one more approach to character

izing justified group belief that I would like to consider, one that is inspired
by the Condorcet jury theorem and developed in the work of Christian List.31
Though List presents his account specifically as one of group knowledge, it
can be adapted for our purposes here. In particular, on this view, (1) the
theory of judgment aggregation is followed in holding that a group’s belief
simpliciter is a function of the members’ individual beliefs, but (2) being
justified is understood as a collective property that a group’s belief may or
may not have, depending on whether it satisfies certain “truth-tracking” or
“reliability” conditions at the group level. For instance, it has been shown
that each of a committee members’ individual beliefs might be only very
slightly better than random at “tracking the truth,” thereby meeting only the
very minimal Condorcetian “greater than 1/2” probability-of-correctness
condition, but, nevertheless, a sufficiently large committee of independent
individuals could be very close to perfect in its collective reliability.32 On a
truth-tracking or reliabilist conception of justification, then, the group
31 See List (2005). 32 For an extended discussion of this, see List (2005).
belief that p might be regarded as justified, even though each of the

underlying individual beliefs falls short of the threshold required for
counting as such.
Unlike Goldmans deflationary summativist view, then, justified group
belief on this Condorcet-inspired picture is not merely an aggregate of justi
fied individual beliefs. Rather, a justified group belief may be an aggregate of
individual beliefs, all of which are unjustified, yet where the group belief
meets the relevant truth-tracking or reliability conditions at the collective
level. And unlike inflationary non-summativism, justified group belief is
constructed out of the epistemic features of the individual beliefs of the
group’s members. It is just that the level of justification at the collective level
outstrips what is possessed by any single belief. This presents a view of
justified group belief that does not fit neatly into either the inflationary
non-summativist or the deflationary summativist camp, and also seems to
tell against requiring justified belief at the level of individual members.
For if this view is correct, then individually justified beliefs—of any sort or
quantity—are not necessary for justified group belief.33
Byway of response, it is undeniable that this Condorcet-inspired approach
shows that there can be epistemic value at the level of the group’s belief that
is not present at the level of any of the individual beliefs. But the question
that I want to focus on is whether this value is plausibly regarded as epistemic
justification; and here I want to suggest that, for several reasons, the answer
should be a negative one.
To begin, this view counts as justified group beliefs that have an inappro
priate epistemic grounding, both at the individual and at the collective level.
Lets begin with the former: notice that in order for there to be justified
group belief on this picture, the central requirements are (i) that the group
have a large enough number of independent members who (ii) satisfy the
‘greater than 1/2” probability-of-correctness condition. But then a group
could go about achieving epistemic justifiedness simply by extending mem
bership to more and more people, with no regard whatsoever to the grounds
of the individual beliefs beyond independence. Suppose, for instance, that
we are talking about a scientific research group: it wouldn’t matter whether
the members believe that p on the basis of experimental results or some
biased testimony, via methodical research or wishful thinking. So long as
(i) and (ii) are met, group justifiedness can be, too. Indeed, rather than
33 I am grateful to an anonymous referee for a presentation of this alternative account of

justified group belief.
checking the CVs of potential members and evaluating the epistemic quality
of the relevant individual beliefs, a research group could achieve justified
group belief by simply surveying the beliefs of individuals and recruiting a
large enough number of members who satisfy (i) and (ii). There’s no need
for any members to go to the expense of actually conducting experiments.
This leads to a related, though slightly different, concern: the grounds of
the members’ individual beliefs of members surely matter to the justified
ness of the group, yet this Condorcet-inspired view cannot account for this.
A group with members all of whom performed excellent scientific experi
ments would clearly be better justified in its resulting scientific belief that
p than one where they all held the same belief because of independent
idiosyncratic websites. The former group would, for instance, be quite likely
to have a great deal of p-related beliefs that are justified, to be able to draw
appropriate inferences regarding that p, to have the capacity to explain why
that p is the case, and so on, while the latter group would not. Even if none
of these features is necessary for group justifiedness, they surely have the
capacity to affect the level or grade of epistemic justification present. But if
the truth-tracking or reliability at the collective level of the two groups is
equivalent, then the Condorcet-inspired view counts them as equally jus
tified in their respective beliefs. In fact, if the group’s belief grounded in
idiosyncratic websites is slightly more reliable than the one based on excel
lent scientific experiments, then it would have a greater level of justified
ness, despite the fact that its basis is wildly inappropriate from a scientific
point of view.
So far I have focused on the bases of the individual members, but there
are also similar concerns that could be raised at the level of the group and
its structure. A group that luckily stumbles into reliability at the collective
level through having a large enough group of members who believe that p
and satisfy (i) and (ii) could end up more justified than one that sets up a
structure where, for instance, evidence is vigilantly gathered, shared among
members, and checked multiple times over. Or a group that isolates
members and forces them to obtain information from epistemically ques
tionable sources could end up more justified than one that engages in col
lective deliberation and forms beliefs on the basis of pooled evidence that
has been scrutinized. In this way, the Condorcet-inspired approach lacks
the resources for allowing group justifiedness to be affected by the way in
which truth-tracking or reliability is achieved at the collective level.
There are also considerations involving group action that tell against
this approach. At a minimum, there is a close connection between a group
believing that p and its being epistemically permissible for such a group to
act as if p.34 Without making any commitments here about this connection
being one of either necessity or sufficiency, we can surely say that if one
justifiedly believes that p, then it is generally the case that it is epistemically
permissible for one to act as if p. It is also the case that groups cannot offer
assertions, engage in negotiations, sign contracts, break the law, or perform
any of the other sorts of actions typically attributed to groups without there
being action on the part of some of its members. Of course, this does not
mean that for every group, G, and act, a, G performs a only if at least one
member of G performs a. It may be that one member performs action b, and
another performs action c, and still another performs action d, which, when
taken together, involves G performing a. But it is the case that for every
group, G, and act, a, G performs a only if at least one member of G performs
some act or other that causally contributes to a. Moreover, for many groups,
particular members are granted the authority to serve as proxy agents,
where this means that the actions performed by such individuals count
as actions of the group’s.35 For instance, a spokesperson might be given the
authority to testify on behalf of a corporation, a CEO the authority to pur
chase property for a company, and an administrator the authority to deny a
colleague tenure for the college. Notice, however, that in each of these cases,
justified belief at the level of the individuals serving as proxy agents is crucial
for rendering the group actions in question epistemically permissible. If, for
instance, the administrator believes that the colleague should be denied
tenure merely because, say, a department member with a grudge told him
so, then the denial of tenure is clearly epistemically improper. In particu
lar, the administrator is not properly epistemically positioned to deny the
colleague tenure, given such a poor basis. This is even clearer if we assume
that there is no member of the college who has a basis that is epistemically
better than the administrator’s. But if justified group belief is entirely
severed from the justified beliefs of members, as the Condorcet-inspired
approach does, then such a denial of tenure could turn out to be entirely
permissible.36
34 See, for instance, Fantl and McGrath (2002 and 2009), Stanley (2005), Williamson (2005),
and Hawthorne and Stanley (2008).
35 For a detailed discussion of proxy agency, see Ludwig (2014) and Lackey (2018a and
2018b). See also Chapters 4 and 5 of this book.
36 I develop this line of argument in far more detail, and in response to a broader range of
theories than just the Condorcet-inspired one discussed here, in Lackey (2014b) and in
Chapter 3 of this book.
For all of these reasons, I regard the Condorcet-inspired considerations

as pointing to features of groups that are undoubtedly epistemically valuable,
but that don’t track justified group belief.
2.9 The Group Epistemic Agent Account
We have seen that the paradigmatic versions of both inflationary non-

summativism and deflationary summativism suffer from debilitating
objections. Let us take a step back and see what can be learned from the
problems afflicting the joint acceptance and aggregative accounts of group
justification.
The joint acceptance account treats groups as epistemic entities that can
float freely of the evidential profiles of their individual members. For
instance, as ignoring evidence reveals, even if every member of Philip
Morris possesses massive amounts of scientific evidence revealing the links
between smoking and lung cancer, such a view permits groups to choose
not to accept this evidence and thereby end up justifiedly believing that
smoking does not pose a health hazard. But groups cannot pick and choose
what evidence is available to them—they are constrained by the evidence
possessed by their individual members. This is the central lesson of the
Illegitimate Manipulation of Evidence Problem.
In contrast, the aggregative account altogether avoids concerns associated
with this problem by securing a close dependence of group justifiedness on
member justifiedness. But this is done at the cost of failing to appreciate the
distinctive epistemic issues that arise at the group level. For the proponent
of the aggregative account, group justifiedness is a simple “justified belief
in/justified belief out” matter. Yet we have seen that this model ignores the
complexity of justified belief at the group level, particularly the evidential
relations that exist between members’ beliefs and bases and the epistemic obli
gations that arise via membership in the group.
Otherwise put, suppose we think of the relation between member justi
fiedness and group justifiedness as a function. On the aggregative account,
this is a very simple function; the inputs to the function are the justified
beliefs of individual members, the function merely aggregates them, and
the output is the justified belief of the group. We have seen, however, that
there are at least three ways in which this model is importantly wrong. The
first case we discussed—cconflicting bases—shows that the aggregative
view is incorrect with respect to the inputs to the function that links
individual and group justification. In addition to the justified beliefs of

individuals, we also need to take into account the bases for these individual
beliefs. The second case—non-conflicting bases—shows that this
modification is still not enough to capture a plausible account of group jus
tification. In non-conflicting bases, the force of the objection rests
on the addition of pieces of evidence that are not part of the bases for the
aggregated individual beliefs. The additional pieces of evidence are relevant,
not to the beliefs of the individuals who have them, but to the beliefs of
other individual members of the group. One response to this case is to say
that the range of inputs to the function linking individual and group justi
fication must be expanded yet again: the function should take into account
all of the individuals’ beliefs and their bases. But this is not a plausible
account of group justification. In particular, it does not come close to mod
eling the way in which groups try to ascertain and make use of justification
in determining what they ought to believe. Moreover, the third case—group
normative obligations—shows that even expanding the range of
inputs in this way is not enough to secure an adequate account of group
justification. The upshot of the third case is that the function itself and not
merely the range of inputs to the function, needs to be modified in a sig
nificant way. In particular, what is left out is the impact of normative
expectations and obligations that apply to the individuals who constitute
the group—and that very often will apply to those individuals in virtue of
their membership in the group. This is exactly what we find in group nor
mative obligations, where N -N acquire epistemic obligations pre
cisely because of their membership in the nursing unit.
Moreover, we are now in a position to fully appreciate the ways in which
Goldman’s claim that aggregation transmits justifiedness is inaccurate.37 In
particular, aggregation does not simply transmit justifiedness from individ
ual members to groups since there are a number of ways in which the out
put can end up being far less justified than the aggregation of the inputs. As
we have seen, this can happen when the members’ bases conflict, when their
collective evidence adds up to zero, or when normative obligations that
exist at the group level fail to be satisfied by the relevant members.
Thus, the justified beliefs of groups should be treated neither as states
that can float freely of the evidence possessed by their individual members,
nor as nothing more than the aggregation of the justified beliefs of their
37 See Goldman (2014, p. 20).

members. Instead, I propose that groups be understood as epistemic agents

in their own right, though ones whose justified beliefs are constrained by
the epistemic statuses and normative obligations of their individual mem
bers. Let us call this the Group Epistemic Agent Account of justified group
belief, according to which:
A group, G, justifiedly believes that p if and only if:
(1) A significant percentage of the operative members of G (a) justifiedly

believe that p,38 and (b) are such that adding together the bases of
their justified beliefs thatp yields a belief set that is coherent.
(2) Full disclosure of the evidence relevant39 to the proposition that p,
accompanied by rational deliberation about that evidence among
the members of G in accordance with their individual and group
epistemic normative requirements, would not result in further evi
dence that, when added to the bases of G’s members’ beliefs that p,
yields a total belief set that fails to make sufficiently probable thatp.40
Let us begin with condition (la), where there are three central features that
should be clarified. First, according to the Group Epistemic Agent Account,
a groups justified belief involves the justified beliefs of operative members of
the group. Recall that a condition of this sort was initially motivated by a
38 Silva (2019) challenges this condition through a case in which every member of a jury
arrives at a belief through improper reasoning, but the group itself arrives at the same conclu
sion through proper reasoning, hi such a case, Silva claims that “the group holds a doxastically
justified belief because it properly responds to its evidence, while no member of the group
properly responds to its evidence. So no member of the group is doxastically justified”
(Silva 2019, p. 9). However, there is no reason to posit justified group belief in such a case when
it can easily be explained in terms of other intuitively plausible descriptions, such as a justified
verdict, or a justified official position, or justified collective acceptance, and so on.
39 I am assuming that the evidence relevant to the proposition that p will subsume beliefs
that bear on it, including those that might arise via premise-based aggregation in "doctrinal
paradox” cases. But if this is doubted, the disclosure of relevant beliefs can be built directly into
condition (2), so that it reads as follows:
(2*) Full disclosure of the beliefs and evidence relevant to the proposition that p, accompanied
by rational deliberation among the members of G in accordance with their individual and
group epistemic normative requirements, would not result in further evidence that, when
added to the bases of G’s members’ beliefs thatp, yields a total belief set that fails to make prob
able thatp.
Moreover, notice that this condition focuses on the full disclosure of the evidence that is
relevant to the proposition thatp. It should be noted that relevant does not mean here in prin
ciple relevant but relevant in the circumstances at issue. This should make clear that condition
(2) is not unrealistically strong.
40 To help grasp condition (2) at an intuitive level, it can be understood as being in the same
broad spirit as requirements in the law that appeal to what “a reasonable person would do.”
case involving the custodians at a law firm to show that the joint acceptance
at the heart of inflationary non-summative views must take place between
the right members. Similar considerations apply here. Suppose that all of
the members of the housekeeping staff at Philip Morris have limited evi
dence about the links between smoking and lung cancer such that they
justifiedly believe that smoking does not pose health risks. Even if they turn
out to constitute a majority of the employees at Philip Morris, the epistemic
status of the beliefs of the CEO and board members is what matters to
whether the group justifiedly believes this. This is because they are the
members who have the relevant decision-making authority in the domain
in question.
Second, a significant percentage of operative members needs to justifiedly
believe that p in order for the group to justifiedly believe it. What amounts
to a significant percentage of operative members varies from group to
group—it might be as small as a single dictatorial member, or as large as all
of the members. But simply one or two members justifiedly believing a
proposition in a fully democratic group of 50 is clearly not sufficient for the
group to justifiedly believe this.41
Finally, notice that the group’s belief that p will inherit a strong, positive
epistemic status from the members’ justified beliefs in which it is based.
These member beliefs will, in turn, be justified by whatever features are
required at the individual level, such as that they are produced by reliable
processes, or grounded in adequate evidence, or track the truth. Although
aggregation of these individually justified beliefs is not sufficient for justi
fied group belief, it does lay the foundation for group justification.
Let us now turn to condition (lb). Unlike the deflationary summativist
approach, the Group Epistemic Agent Account requires that adding the
bases of the justified beliefs thatp of the same operative members at issue in
(la) yields a belief set that is coherent. This is what was learned from con
flicting bases: member justifiedness does not transmit smoothly to
group justifiedness, since the bases of the members’ beliefs might be wildly
conflicting. When this happens, such a view faces the Group Justification
Paradox and the Defeater Problem. Otherwise put, it is precisely because
41 As with other threshold notions in epistemology, such as “sufficient” justification for

knowledge, there is room for disagreement over where on the scale the threshold is located for
“significant” percentage and how it comes to be there. Some will argue that contextualism is
helpful here; others will take the threshold to be fixed by practical interests or by implicit social
coordination. As that debate is tangential to the thread of my argument here, I will set it aside.
vertical justifiedness aggregates member beliefs with no regard for their

bases that these problems arise.
The second feature of (lb) that should be noted is that the belief set
resulting from adding together the bases of G’s operative members’ beliefs
that p needs to be coherent.*2 Notice that consistency is not required here.
This is because various philosophers have made a compelling case for the
claim that an individual can be rational despite having an inconsistent set of
beliefs. As was noted earlier, the Preface Paradox describes just this sort
of situation, where it is rational for the author of a book to believe both
the individual claims made in the book, plus the claim that at least one of
them is false. So, consistency cannot be a requirement of individual rational
belief But groups can easily find themselves in situations of this kind, so
similar reasoning leads to the conclusion that consistency is too strong a
requirement of group rational belief, too.42 43
How, then, should we characterize the requirement that the belief set
resulting from aggregation be coherent? As noted in Chapter 1, there are
various ways of understanding this. One option is to understand coherence
in terms of evidential support.44 Alternatively, coherence can be understood
in terms of accuracy-dominance avoidance, in the sense that, for a coherent
set of beliefs, there is no rival belief set that is never worse and sometimes
better than it with respect to overall accuracy.45 Again, I take no stand on
how precisely this concept should be understood, as it is sufficient for my
purposes that there is an intuitive failure of coherence in conflicting
bases, and that there are accounts that can capture this.
Third, notice that the bases that are being added together to yield a
coherent set are those of G’s operative members’ justified beliefs that p. If
operative members believe that not-p, or believe that p for dogmatic or
irrational reasons, then their beliefs and bases are simply irrelevant to the
justificatory status of G’s belief thatp.
Finally, with respect to condition (1) as a whole, it should be emphasized
that its satisfaction is compatible with the bases of some operative members’
42 It should be emphasized that the beliefs in this set will not all be group beliefs. For
instance, some of the beliefs in the set might not be shared by enough of the individual mem
bers to count as the groups beliefs.
43 However, List (2014b) introduced the concept of “consistency of degree /c,” which is
weaker than full consistency by ruling out only “blatant” inconsistencies in an agent’s beliefs
while permitting less blatant ones. Condition (1) might be understood as requiring this weaker
notion of consistency.
44 See Kolodny (2007). 45 See Briggs, Cariani, Easwaran, and Fitelson (2014).
justified beliefs failing to cohere with the bases of others. All that is needed
is that a significant percentage of the operative members of the group satisfy
both (la) and (lb). Thus, the Group Epistemic Agent Account permits justi
fied group belief in cases like the following: 99 out of 100 operative mem
bers of a group have bases for their justified beliefs that p that cohere with
one another, but the basis of the final operative members justified belief
thatp fails to cohere with those of the others.46 The lack of coherence added
by this lone basis does not prevent the group from justifiedly believing that
p, given the coherence to be found among the bases for the justified belief
thatp of a significant percentage of operative members.
It should be clear that the inclusion of (la) in the Group Epistemic
Agent Account altogether avoids the Illegitimate Manipulation of Evidence
Problem afflicting inflationary non-summativist views. One of the features
that make the joint acceptance account susceptible to this objection is that
the evidence available to the group can end up being a matter of choice.
In particular, because groups can often exercise control over what is jointly
accepted, they can manipulate what evidence is, and is not, available to
them as a group. But if the justification of group beliefs is necessarily a mat
ter of the justification of the beliefs of individual members, and the evidence
that is available to individual subjects is not a matter of choice, then there
is no worry that epistemic justification for group beliefs can be achieved
through the illegitimate manipulation of evidence. If, for instance, a significant
percentage of the operative members of Philip Morris justifiedly believe that
smoking causes health problems, then no amount of joint acceptance, or
lack thereof, can make it the case that Philip Morris itself does not justifiedly
believe this, too.
At the same time, it should also be apparent that including (lb) in the
Group Epistemic Agent Account denies justified group belief in con
flicting bases and thereby avoids both the Group Justification Paradox
and the Defeater Problem. In particular, adding together the bases of the
British Museum guards’ beliefs that p yields an incoherent set of beliefs.
This results in G failing to justifiedly believe that someone is planning an
inside theft of a famous painting at the British Museum, despite the fact that
all 100 members of G justifiedly believe this. Since it was the attribution of
justified belief to G here that generated the problems stemming from con
flicting bases in the first place, they simply don’t arise for the Group
Epistemic Agent Account.
46 I am grateful to Sharon Ryan for a comment that led to this point.

While (1) addresses the Illegitimate Manipulation of Evidence Problem,

the Group Justification Paradox, and the Defeater Problem, it is silent when
it comes to the Collective Evidence Problem and the Group Normative
Obligations Problem. This leads us to condition (2) of the Group Epistemic
Agent Account, where there are six features that should be emphasized.
First, notice that this is a negative condition: it requires only that full
disclosure and rational deliberation be such that they would not generate
further evidence that, when added to the bases of G’s members’ beliefs that
p, would yield a total belief set that fails to make probable that p. This is
because condition (1) proceeds by aggregating justified beliefs of operative
members. The bases of these individual beliefs make probable that p. Thus,
if the belief set of these bases is coherent, the set itself should still make
probable that p. What (2) does is to ensure that disclosure and deliberation
do not add evidence such that the resulting expanded set of beliefs that
the group ends up with would no longer make probable that p. There are
various ways to understand probability—for example, as grounded in fre
quencies or propensities, or as an a priori relation—but the details are not
important here.47
Second, condition (2) is a subjunctive conditional. It does not require that
the group members in fact fully disclose all of the relevant evidence and engage
in rational deliberation. This would be far too strong. Instead, it specifies
what needs to be true of the evidence, were the group members to do this.
Third, it is the evidence that actually exists that is relevant to the disclos
ure and deliberation condition—not the evidence that would exist if the
group were to engage in these activities in the counterfactual situation. This
captures the sense in which the groups justification depends on the evidence
they actually have, along with how they ought to think about it.48
47 See, for instance, Goldman (1979 and 1986), Kyburg (2001), and Fumerton (2004).
48 More precisely, let’s distinguish the following: e is the evidence the group members have
in the actual world, e* is e plus what full disclosure and rational deliberation on e would yield,
and e** is e* plus what awareness of what they are doing as they engage in disclosure and
rational deliberation would yield. The last of these, e**, goes beyond what their evidence is and
how they should think about it, and thus is irrelevant to the groups actual justification. This is
relevant to a worry that one might have that a group couldn’t justifiedly believe that it is not
deliberating now, because were they to disclose and deliberate about the relevant evidence, the
process would bring about evidence that they were deliberating. (I am grateful to John
Hawthorne for this objection.) But, as should be clear, what is at issue in this objection is e**,
and not e*, which is what condition (2) picks out.
Otherwise put, awareness of what one is doing in deliberating is typically yielded in deliber
ating, but it follows from the deliberation itself, not from the evidence that is the content of the
deliberation. Here is another instance of this pattern: suppose the existence of a particular
group is tenuous, in the sense that it is likely to be formally dissolved at any moment. Suppose,
Fourth, the rational deliberation at issue in (2) must be in accordance

with the epistemic normative requirements governing both the individual
members and the group as a whole. While being a member of a group does
not absolve one of one’s duties—epistemic or otherwise—it can bring with it
new ones. For instance, as we saw in group normative obligations,
being a member of the nursing unit requires of N2-N3 that they always
communicate with one another about the patients that they mutually care
for, even though this is not expected of them as non-members. According
to (2), then, if any epistemic norms, individual or group, have been violated,
then this would be made clear in the process of disclosure and deliberation,
and hence the group would not have justified group belief.
Notice, however, that condition (2) does not require that the members of
the group in question actually follow the epistemic normative requirements
governing both the individual members and the group as a whole. That this
would be too stringent can be seen by considering the following: suppose
that only in group normative obligations is aware that she for
got to give O’Brien his first medication and that she also justifiedly believes
that this act of negligence alone is not sufficient to put him in danger of
death. Even though N is flouting a group epistemic norm when she fails to
share this information with the other members of the nursing unit, this by
itself is not sufficient for the group to fail to justifiedly believe that O’Brien
is at risk of dying. Condition (2) delivers exactly the right verdict here: full
disclosure and rational deliberation among the members of the nursing unit
would reveal both that Nx failed to fulfill this requirement, but also that
this does not put O’Brien in danger. Thus, this process would not result in
further evidence that, when added to the bases of G’s members’ beliefs that
p, yields a total belief set that fails to make probable that O’Brien is not at
risk of dying.
This bears on a concern that one might have about (2): what if there is a
dogmatic member who steadfastly clings to misleading evidence, but does
not actually share it with the other members of the group? Should the fact
that the belief set in question would fail to make probable that p were she to
further, that the group has evidence on Monday that it exists then, though it hasn’t yet formed
the belief that it exists on Monday. The evidence in question is silent as to whether the group
might exist after Monday. So, deliberation on the import of the evidence will be silent as to
whether the group exists on Tuesday Nevertheless, if the group were to deliberate on Tuesday,
the act of deliberation would provide new evidence that the group exists on Tuesday, but this
wouldn’t be revealed through the content of that deliberation. The condition in question is
specifying the content of what the deliberation would lead to; it does not require that an act of
deliberation ever occur.
share it result in the group actually failing to justifiedly believe that p? By

way of response, notice that the mere possession of counterevidence by a
member does not, by itself, result in the group failing condition (2) of the
Group Epistemic Agent Account; instead, it has to be such that it would survive
full disclosure of all of the relevant evidence and rational deliberation by the
members of G, where the latter is done in accordance with the governing
epistemic normative requirements. Counterevidence that is possessed entirely
because of dogmatism or some other epistemic vice would presumably be
rejected or dismissed via this process. Moreover, in those cases where a
member has counterevidence that is not in fact shared but is such that it
would survive this sort of process of disclosure and deliberation, then
condition (2) seems to provide the correct verdict. For instance, suppose
that an operative member of a scientific research group has evidence that
undermines the conclusion that a drug is effective in treating a particular
land of cancer, but does not disclose it to the other members. Were she to
share it, however, all of the other members of the research group would
accept it, thereby yielding a belief set that fails to make probable that
the drug is effective in treating cancer. Here, the unshared counterevidence
possessed by the operative member prevents the group from justifiedly
believing that the drug is effective in treating cancer. But this also seems to
be the right result, one supported by the fact that we would surely hold the
group responsible for the harmful consequences of treating cancer patients
with the drug were we to learn that one of the operative members had com
pelling evidence challenging its effectiveness.
A related issue that arises here is this: suppose there is a skeptic in a given
group who doesn’t disclose her skeptical worries to the other members.
Isn’t it the case that for just about any proposition and any belief set, adding
a skeptical argument to it would result in a belief set that fails to make
probable that p? Thus, doesn’t it turn out on the Group Epistemic Agent
Account that the mere presence of a skeptic in a group results in the group
itself failing to justifiedly hold any beliefs?49 By way of response, it should be
emphasized that the skeptical arguments would have to be such that they
would survive full disclosure of all of the relevant evidence and rational
deliberation by the members of the group in accordance with all of the indi
vidual and group epistemic norms. This means that whether the addition of
the skeptical doubts would in fact result in a belief set that fails to make
49 I am grateful to Nick Leonard for this objection.

probable a given proposition depends on the evidential force of the skeptical

arguments themselves. But many, if not most, people—philosophers and
otherwise—continue to regard themselves as having knowledge even in the
face of skepticism. So, there is no reason to think that the same wouldn’t be
the case with respect to most groups. On the other hand, if skepticism is
correct, then none of us—individuals or groups—have justified belief. But
this would be in virtue of the fact that skepticism is correct and not in virtue
of the presence of a skeptical group member.
Fifth, all of the group members, not just the operative ones, are relevant
to the counterfactual disclosure of evidence and rational deliberation in (2).
This is because both operative and non-operative members can possess
relevant counterevidence for believing that p, or can flout epistemic norms,
that bear significantly on the justificatory status of a groups belief. For
instance, suppose that the nursing unit comprising Nx-N is a smaller part
of a larger group—the nursing home staff—that includes the custodians for
the building. Even if the ten custodians at the nursing home in group
normative obligations are not operative members of the group and
thus do not contribute positively to justified group belief when O’Brien’s
health is concerned, their having seen Nj-N3 all fail to give the patient his
medications might still be relevant counterevidence to whether the group
justifiedly believes that O’Brien is not at risk of dying. Moreover, their failing
to communicate this information to the head of the nursing unit consisting
of N -N3 might also be in violation of general epistemic norms embraced by
the nursing home.50
50 It might be noticed that there is an asymmetry in the Group Epistemic Agent Account:
while only operative members can contribute positively to the group’s justified belief, any
member can bear negatively on the groups epistemic status. This shouldn’t be surprising, how
ever, since this asymmetry is mirrored at the individual level. For instance, just about every
externalist about epistemic justification accepts that negative reasons can defeat knowledge,
even if positive reasons are not necessary for knowledge. Similarly, justification for believing is
harder to come by than justification for doubting—I may, for example, doubt that a car is reli
able on the basis of a used car salesman’s raising questions about the car’s condition, but I
wouldn’t believe that another car is reliable on the basis of this same man’s testimony that it is.
Still further, consider Harman’s (1973) well-known newspaper case: Jill reads in a newspaper
that the President has been assassinated and, though the story is true, it has been suppressed by
the government. As a result, all of the other newspapers and television stations are reporting
that the President is fine and that the assassin actually killed his bodyguard. Harman argues
that Jill does not know that the President was assassinated in such a case. But the point that is
of interest here is that, if this conclusion is correct, then evidence that one does not possess can
undermine one’s knowledge. But the same is never said about acquiring knowledge.
Finally, it should be emphasized that condition (2) does not entail (lb).
To see this, suppose that the bases of the operative members’ justified
beliefs that p are incoherent from the start, but full disclosure and rational
deliberation would not turn up new evidence that, when added to these
bases, yields a total belief set that fails to make probable that p. In such a
case, the group’s belief that p will satisfy condition (2), and, let us suppose,
(la) as well, but surely the initial incoherent bases render the belief unjus
tified. This verdict of unjustified group belief is precisely what condition
(lb) ensures.
Moreover, condition (lb) ensures that the basis for group justification is
widely distributed among the group’s operative members. Even in cases
where the operative members who justifiedly believe that p do not all share
the same base for that belief, there is still a significant percentage of them
who justifiedly believe thatp with bases that could in principle be shared. In
that sense, they exhibit the cohesiveness that is characteristic of genuine
group phenomena.
With these points in mind, let’s see how (2) handles the Collective
Evidence Problem. Recall that this objection is generated by non
conflicting bases, where the group belief in question is that a man
was responsible for an inside theft of a famous painting at the museum. In
this case, none of the bases of the British Museum guards’ individual beliefs
conflict, but each subgroup of 20 guards has counterevidence for the basis
of the justified beliefs of a different subgroup. When viewed as a collective
whole, then, the evidential basis of the group is zero. This is because there is
not a single basis of the members’ beliefs that is free of counterevidence. But
notice: if the 100 guards were to engage in full disclosure and rational delib
eration, all of the counterevidence for the bases would emerge. For instance,
each of the first 20 guards, Mx-M , would disclose that they have evidence
that the thief’s goatee is fake, and thus all of the members would then realize
that this undermines the basis of belief for M21-M40> Were all five subgroups
to do this, it is clear that there would be no surviving evidence for believing
that a man was responsible for the inside theft. Thus, full disclosure and
rational deliberation among the guards in non-conflicting bases
would clearly produce further evidence that, when added to the bases of the
members’ relevant beliefs, yields a total belief set that fails to make probable
the proposition in question. According to the Group Epistemic Agent Account,
then, the group of British Museum guards does not justifiedly believe that a
man was responsible for the inside theft—despite the fact that all 100 guards
justifiedly believe this—which is precisely the desired verdict.
Let us now see how (2) deals with the Group Normative Obligations
Problem. Recall that in group normative obligations, each of Nt-N3
justifiedly believes that O’Brien is not at risk of dying, but also fails to fulfill
her group epistemic duty of sharing with the other nurses that she forgot to
give him his medication. In such a case, were the nurses to fully disclose and
rationally deliberate about all of their relevant information, and were this
evidence to be added to the original bases, the resulting belief set would
clearly fail to make probable that O’Brien is not at risk of dying. For such a
set would include both the belief that Nx-N3 all forgot to give O’Brien his
medication and the belief that his missing all three of his medications would
put him at serious risk of dying. Once again, then, the Group Epistemic
Agent Account delivers the correct verdict: the nursing unit’s belief fails
condition (2) and thus the group does not justifiedly believe that O’Brien is
not at risk of dying, despite the fact that all 3 nurses justifiedly believe this as
individuals.51
51 Silva (2019) argues for replacing my view with what he calls Evidentialist Responsibilism
for Groups (ERG), according to which:
A group, G, justifiedly believes that P on the basis of evidence E iff: (1) E is a sufficient reason to
believe P, and the total evidence possessed by enough of the operative members of G does not
include further evidence, E*, such that E and E* together are not a sufficient reason to believe P,
and (2) G is epistemically responsible in believing P on the basis of E.
Group Responsibilist Condition:
A group, G, is epistemically responsible in believing P on the basis of E iff (a) enough of the
operative members of G satisfy their G-relevant epistemic duties, and (b) G properly bases its
belief on E.
Space constraints prevent me from discussing Silvas very interesting view in detail, but let me
just note a couple of concerns with his account.
First, according to the ERG, some operative members of a group can have defeaters while
the group itself remains justified on this view. This is because the total evidence possessed by
“enough” of the operative members cannot include further evidence, E*, that functions as a
defeater. Presumably, then, there can be “some” operative members who have such counterevi
dence. But why would justified group belief be compatible with any operative members having
evidence that is in direct conflict with the evidence that other operative members possess?
Second, the ERG requires that the relevant operative members of G satisfy their G-relevant
epistemic duties, But what if the G-relevant epistemic duties deviate in important ways from
general epistemic duties? For instance, as we’ve noted, juries exclude hearsay evidence even
when this exclusion not only fails to have a general epistemic advantage, but often has specific
epistemic disadvantages, by, for instance, leading us farther away from the truth. Other groups
might have similarly epistemically disadvantageous group duties. A group, for instance, might
exclude the relevance of scientific testimony from non-approved sources, where the criteria
for approval is explicitly bound up with financial motivations. Thus, the group satisfies its
G-relevant epistemic duties, but such duties are objectively deeply epistemically problematic.
Why would we say that the group is justified?
2.10 Central Objection to the Group

Epistemic Agent Account
I would like to here consider and respond to a central objection that might
be raised to the Group Epistemic Agent Account; namely, that it succumbs
to a version of the IMEP. In particular, if group justification depends on the
justification of the beliefs of the group members, then isn’t it possible for the
epistemic status of group beliefs to be determined by the deliberate manipu
lation of the group membership? For instance, suppose that a group aims
to arrive at a particular justified group belief through the accepting and
removal of certain members. To this end, suppose that Philip Morris hires
only new employees who actually justifiedly believe that the scientific evidence
regarding the ill effects of smoking is unreliable, and they fire all those who
believe such evidence to be reliable. In such a case, isn’t it possible for Philip
Morris to end up justifiedly believing that the scientific evidence regarding
the ill effects of smoking is unreliable precisely because they illegitimately
manipulated the justified beliefs of the group members?52 And isn’t this just
a slightly different version of the IMEP?
By way of response, the first point to notice is that, unlike the version of
the IMEP afflicting the JAA, every account of group justification succumbs
to this one. For instance, at the inflationary end of the spectrum, Philip
Morris might hire only new employees who will jointly accept that the sci
entific evidence regarding the ill effects of smoking is unreliable, and fire all
those who will not, thereby resulting in Philip Morris justifiedly believing
that smoking is not dangerous according to the JAA. At the deflationary end
of the spectrum, Philip Morris might hire only new employees who actually
justifiedly believe that the scientific evidence regarding the ill effects of smok
ing is unreliable, and fire all those who believe such evidence to be reliable,
thereby resulting in Philip Morris justifiedly believing that smoking is not
dangerous according to deflationary summativism. What this reveals is that
it is simply part of the nature of collective justification that a change in a
group’s membership can change the justificatory status of the group’s beliefs.
The second point I should like to make is that there is an important
asymmetry between the version of the IMEP afflicting the JAA and the one
purportedly facing the Group Epistemic Agent Account; namely, that the
illegitimate manipulation of evidence easily results in the group’s belief
52 I am grateful to Stew Cohen, Juan Comesana, and Brian Miller for this objection.
being justified in the former, but not in the latter. With respect to views
where the very reasons available to the group are determined via joint
acceptance, the group can illegitimately restrict the evidence available to
them and thereby end up with justified beliefs that are intuitively unjusti
fied. But when there is a summative constraint on group justification, this
also brings with it a broadly summativist conception of defeaters. Thus, if a
group aims to arrive at a particular justified group belief through the accept
ing and removal of certain members, then at least some members of the
group are aware of this. And if at least some members of the group are privy
to the fact that the evidence is being manipulated in this way, then they have
a defeater for accepting the proposition in question, thereby preventing the
group from justifiedly holding the target belief.
But what, it might be asked, if a person outside of the group aims to arrive
at a particular justified group belief through the accepting and removal of
certain members? If this person is not a member, then the evidence she has
will not count as a defeater for the group. So, couldn’t the group end up
with a justified belief through the manipulation of evidence by this out
side person?
However, the mere fact that evidence can be illegitimately manipulated,
with the end result being a justified belief, is not at all surprising. This hap
pens at both the individual and at the group level. For instance, if I deliber
ately undergo a memory-removal procedure, then obviously the justification
of my memorial beliefs will be seriously altered. More realistically, I might
deliberately choose to watch only BBC news because my friend works there,
my students might hide their consulted sources because they don’t want me
to know that they plagiarized, or I might always consult brunettes when
asking for directions out of habit. In such cases, the evidence available to me
is being restricted—either by me or by others—but this does not necessarily
prevent me from having justified beliefs in the relevant domains. If, say, the
BBC news or brunettes are reliable source of information, then their testi
mony may be able to provide the epistemic grounding needed for me to
have knowledge. Or, if I am wholly ignorant of what sources my students
relied upon, then they cannot provide evidence that either grounds or defeats
my belief that the work is original. This is true with respect to groups, too.
The members of a company may decide to only rely on CNN because their
CEO owns stock in it, or they may never look in the drawers of co-workers
because it is against the company’s ruled to do so, or they may always
depend on the testimony of their own lawyers when obtaining legal counsel
out of habit. Again, this restriction of the available evidence does not, by
itself, prevent the group in question from having justified beliefs, since the
sources might still be reliable or provide adequate evidence.
Applying this to the case at issue, if an outside person of a group aims to
arrive at a particular justified group belief through the accepting and
removal of certain members, then this seems no different from an outside
person deliberately restricting the evidence available to a group. And if, as
we have seen above, the latter can result in a justified belief, I see no reason
why the former cannot.
Otherwise put, in all of the cases where the attribution of justified belief
is epistemically unproblematic, the manipulation of evidence at issue is
what we might call indirect— it involves restricting one’s access to evidence.
This is importantly different from manipulating evidence in a direct way,
which involves ignoring evidence of which one is already aware or fabricating
evidence that doesn’t otherwise exist. A paradigmatic example of the former
is choosing to not read a newspaper; a paradigmatic example of the latter
is reading the newspaper and then deliberately choosing to ignore what
one just learned. Thus, the version of the IMEP afflicting the JAA concerns
the direct manipulation of evidence, while the one at issue with respect to
the Group Epistemic Agent Account involves only the indirect manipulation
of evidence.
2.11 Conclusion
We have seen that the Group Epistemic Agent Account handles all of the
problems afflicting the two dominant approaches in the literature to
understanding the justification of group beliefs. While inflationary non-
summativists focus entirely on what a group does—that is, whether its
members engage in joint acceptance or not—deflationary summativists focus
exclusively on what a group has—that is, whether its members’ beliefs are
individually justified. The Group Epistemic Agent Account, in contrast,
incorporates both components. In particular, groups are understood as
epistemic agents on this view, ones that have evidential and normative con
straints that arise only at the group level, such as a sensitivity to the rela
tions among the evidence possessed by group members and the epistemic
obligations that arise via membership in the group. These constraints bear
significantly on whether groups have justified belief. At the same time,
however, group justifiedness on the Group Epistemic Agent Account is still

largely a matter of member justifiedness, where the latter is understood as
involving both beliefs and their bases. The result is a view that neither
inflates nor deflates group epistemology, but instead recognizes that a
group’s justified beliefs are constrained by, but are not ultimately reducible
to, members’ justified beliefs.
Group Knowledge
There are two quite influential views of group knowledge that are inflationary
and non-summative in nature, and that pose challenges to the account of
justified group belief developed in Chapter 2. The first is often referred to as
“social knowledge,” and it is developed and defended in most detail by
Alexander Bird. A paradigmatic instance of social knowing is taken to be
the so-called knowledge possessed by the scientific community, where no
single individual knows a given proposition, but the information plays a
particular functional role in the community. The second is “collective
knowledge,” which occupies an important place in United States law.
According to the “collective knowledge doctrine,” knowledge may be
imputed to a group by aggregating bits of information had by its individual
members. If these accounts of social knowledge and collective knowledge
are correct, then my view of justified group belief is false, particularly the
requirement that some of the operative members of a group have the rele
vant justified beliefs themselves. So, in this chapter, I will take a close look at
these two inflationary conceptions of group knowledge. I will argue that
both accounts fly in the face of fundamental features of knowledge, and
thus should be rejected as accounts of group knowledge.
3.1 Social Knowledge
One way that a group is said to know that p without a single of its members
knowing thatp is if information is distributed across a collective entity in a
“compartmentalized” way. An oft-cited instance of this is Edwin Hutchins’s
example of the crew of a large ship safely navigating the way to port.1 Each
crew member is responsible for tracking and recording the location of a dif
ferent landmark, which is then entered into a system that determines the
ship’s position and course. In such a case, the ship’s behavior as it safely
travels into the port is clearly well-informed and deliberate, leading to the
1 See Hutchins (1995).
DOI: 10.1093/OSO/9780199656608.003.0004
conclusion that there is collective knowledge present, More precisely, it is

said that the crew as a whole knows, for instance, that they are traveling
north at 12 miles per hour, or that the ship itself knows this, even though no
single crew member does. This is taken to show that knowledge is socially
extended in an important sense.2
Recently, an even more radical conception of socially extended know-
ledge has been defended by Alexander Bird, according to which knowledge
can be possessed, not only by structured groups with a unified goal—like
the crew of a ship—but also by large and unstructured collective entities
with diverse aims. He calls this phenomenon “social knowing” or “social
knowledge,” paradigmatic instances of which are “North Korea knows how
to build an atomic bomb,” or “The growth of scientific knowledge has been
exponential since the scientific revolution.” In such cases, a collective entity
is said to know that p despite the fact that not a single individual member is
even aware of thatp, thereby showing that social knowledge does not super
vene on the mental states of individuals.
In the first part of this chapter, I will argue that knowledge is not socially
extended in this radical sense.3 My argument is twofold: first, I show that
endorsing “social knowing” or “social knowledge” leads to serious epis
temological problems and, second, I suggest that the work done by ascrib
ing social knowledge to collective entities can instead be done by describing
such entities as being in a position to know.
To begin, it is undeniable that we often attribute knowledge to large,
unstructured social groups. We talk about the U.S. knowing that Osama bin
Laden was killed, the scientific community knowing that climate change is
having a serious impact on wildlife, and liberals knowing that Fox News is
biased. These are all instances of what Bird calls organic groups. Unlike
established or structured groups, which derive their cohesion from joint
acceptance or external rules, organic groups are held together by organic
solidarity, which “involves bonds that arise out of difference, primarily the
inter-dependence brought about by the division of labour. The key feature
of the division of labour is that individuals and organizations depend
on others who have different skills and capacities” (Bird 2010, p. 37).
2 Indeed, some go further and argue that Hutchins’s example, “rather than that of a jury or a
board of directors facing a special decision, should serve as a paradigm of collective know
ledge” (Klausen 2015, p. 823).
3 While I will here focus on Bird’s radical view, in large part because it is developed in such
fine detail, my arguments apply to weaker inflationary non-summativist views of group know
ledge as well.
GROUP KNOWLEDGE 113
For instance, the division of labor in science is made clear through the way
in which different subfields and specialties bind scientists together. Bird
offers the following to illustrate this:
...a palaeobiologist is investigating the relationship between certain

extinct animals. A significant part of the relevant evidence concerns the
age of the rocks in which the fossils of the animals were found, and thus
depends on the work of geologists. The geologists, in dating the rocks,
depend in large measure on techniques that concern the radioactivity of
rock samples, and thereby depend on theories and equipment developed
by physicists. (Bird 2010, p. 38)
The sort of interdependence found between scientists is a paradigm of the

land of division of labor that is central to organic groups, and such groups
are, according to Bird, the bearers of social knowledge.
Unlike every other conception of group knowledge, however, social
knowledge is said to not supervene on the mental states of individuals.
Notice the force of this claim: it is not saying that a group can know that p
without a single individual member of the group knowing thatp. As we saw
earlier with respect to belief and justification, this is a widely accepted thesis
in the collective epistemology literature, and allows for weaker require
ments involving the members of groups, such as joint acceptance. But social
knowledge permits an organic group to know that p without individual
belief that p, joint acceptance that p, commitment thatp; indeed, without a
single individual person—group member or not—even being aware of thatp.
This conclusion is motivated by cases like the following:
accessible information: Dr. N. is working in mainstream science,

but in a field that currently attracts only a little interest. He makes a discov
ery, writes it up and sends his paper to the Journal ofX-ology, which pub
lishes the paper after the normal peer-review process. A few years later, at
time t, Dr. N. has died. All the referees of the paper for the journal and its
editor have also died or forgotten all about the paper. The same is true of the
small handful of people who read the paper when it appeared. A few years
later yet, Professor O. is engaged in research that needs to draw on results in
Dr. N.’s field. She carries out a search in the indexes and comes across
Dr. N.’s discovery in the Journal ofX-ology. She cites Dr. N.’s work in her own
widely-read research and because of its importance to the new field, Dr. N.’s
paper is now read and cited by many more scientists. (Bird 2010, p. 32)
Let T1 be the time at which Dr. N.’s scientific discovery, d, is first published,
T2 be the time at which Dr. N. and all of the readers of the paper are dead,
and T3 be the time at which Professor O engages with Dr. N’s work and
makes it such that it is widely read and cited. According to Bird, there is a
collective subject—namely, the scientific community or “wider science”—
that knows that d throughout this entire process. He writes:
Was Dr. N.’s discovery part of scientific knowledge? I argue that it was so
throughout the period in question. There is no doubt that it was at the end
and also at the beginning. By publishing in a well-known, indexed journal,
Dr. N. added to the corpus of scientific knowledge in the way that many
hundreds of scientists do each month. Now consider the intermediate
time t. As regards its status as a contribution to scientific knowledge, it
seems irrelevant that Dr. N. and others who had read the original paper
had died or forgotten about it. What is relevant is that the discovery was in
the public domain, available, through the normal channels, to anyone,
such as Professor O., who needed it. (Bird 2010, p. 32)
According to Bird, then, the scientific community knows that d at T2

despite the fact that not a single individual is even aware of that d at this
time. This is taken to support the negative claim that social knowledge does
not supervene on the mental states of individuals. It is also taken to estab
lish the positive thesis that social knowledge is fundamentally a matter of
the accessibility of the information in question to the members of the col
lective entity who need it.
But surely accessibility cannot be the full story for social knowledge, This
is where Bird’s thesis about the “functional role of knowing” enters. Taking
individual cognitive faculties as the model, Bird argues that only social
structures with the following properties are candidates for being social
knowers:
(i) They have characteristic outputs that are propositional in nature

(propositionality);
(ii) They have characteristic mechanisms whose function is to ensure or
promote the chances that the outputs in (i) are true (truth-filtering);
(iii) The outputs in (i) are the inputs for (a) social actions or for (b) social
cognitive structures (including the very same structure [the
structure that produces the output]) (function of outputs) (Bird 2010,
pp. 42-3)
GROUP KNOWLEDGE 115
Scientific communities clearly satisfy (i)-(iii). The characteristic output of the

scientific community is the journal article, which is clearly propositional.
There are mechanisms within science, such as the peer-review process, to
promote the chances that the outputs in journal articles are true. And the
research results published in scientific journals are the direct inputs for
social action, such as the use of new drugs for diseases or the application of
novel technologies in businesses.
Combining these considerations, the following conception of social
knowledge emerges:
SK: Social Knowledge: A social structure, S, socially knows that p if and

only if (1) that p is true,4 (2) S satisfies (i)-(iii), and (3) the information that
p is accessible5 to the members of S who need it.6
SK is able to capture not only the knowledge purportedly had by the scien
tific community in cases such as accessible information, but also
our ordinary attributions of knowledge to collective entities like the U.S.
Despite these virtues, I will argue in what follows that social knowledge is
not knowledge after all, as accepting SK leads to unacceptable epistemo
logical consequences.
3.2 Social Knowledge and Action
The first point to notice is that no matter how inflationary collective entities
and their states are deemed, groups cannot float entirely freely from their
individual members. Indeed, even those who argue that groups “have minds
of their own”7 recognize that there is undoubtedly an intimate connection
4 Bird writes, “My view is unashamedly veritistic, and is an extension of traditional analytic
epistemology and its conviction that knowledge entails truth” (Bird 2010, p. 24).
5 There are questions that may be raised about the notion of accessibility operative here, but
they lie beyond the scope of this chapter.
6 Despite the fact that Bird argues throughout his (2010) that accessibility is what is key for
social knowing (e.g., “What is important [for social knowing] is that the information should be
accessible to those who need it” (2010, p. 48), he weakens this requirement near the end of his
paper. He writes: “...the fundamental point is functional integration—the knowledge plays a
social role.... [I]t is not the accessibility of the knowledge that is essential to its being social
knowledge; rather it is the capacity of the knowledge to play a social role... in virtue of the
structure and organization of the group; accessibility is the principal means by which that is
achieved” (2010, p. 48). This is an even weaker account than what is offered in SK. Since the
arguments that follow will all show that SK is too weak, I will ignore this slightly modified
version.
7 See Pettit (2003).
between groups and their individual members, as is made clear in the

following passage from List and Pettit:
The things a group agent does are clearly determined by the things its
members do; they cannot emerge independently. In particular, no group
agent can form intentional attitudes without these being determined, in
one way or other, by certain contributions of its members, and no group
agent can act without one or more of its members acting.
(List and Pettit 2011, p. 64).
Such a connection between groups and their members is particularly vivid

when it comes to action. A group cannot offer assertions, engage in negoti
ations, sign contracts, break the law, or perform any of the other sorts of
actions typically attributed to groups without there being action on the part
of some of its members. Of course, this does not mean that for every group,
G, and act, a, G performs a only if at least one member of G performs a. It
may be that one member performs action b> and another performs action c,
and still another performs action d, which, when taken together, involves G
performing a. For instance, one nursing home worker might forget to turn a
patient over in order to prevent bedsores, another might forget to wash this
same patient’s bedsores, and a third might forget to give the patient her anti
biotics. Each of these actions might be necessary, though not sufficient, for
the patient’s death, and thus the collective entity might be convicted of neg
ligent homicide, even though not a single one of the workers is herself guilty
of this action. Nevertheless, even though the group committed an act of
negligent homicide without any individual committing this very same act,
the group could not have performed the act without the members perform
ing relevant acts, such as failing to turn the patient over and wash her. This
supports the following weaker relationship between groups and their mem
bers when actions are concerned:
GMAP: Group/Member Action Principle: For every group, G, and act, a,

G performs a only if at least one member of G performs some act or other
that causally contributes to cz.8
8 I should make explicit that I will discuss various forms of “proxy agency” in the next two
chapters where an agent has authority to perform an action on behalf of a group. Such an agent
may or may not be a member of the group itself, but I do not regard these cases as violating
GMAP. In all cases of proxy agency, there is some action made by a member of the group that
causally contributes to the group's action, whether this is the granting of authority to the proxy
agent, the acceptance of inherited authority, the absence of objections (which is an omission
that causally contributes to the group’s action), and so on.
GROUP KNOWLEDGE 117
What the GMAP makes clear is that while group action cannot occur
independently of its members, it can go beyond what any of them do
individually
The second point to note is the intimate connection that exists between
knowledge and action. A widely accepted view of this connection is that a
necessary condition on knowing that p is that it is epistemically appropriate
to use the proposition that p in practical rationality. For instance, Jeremy
Fantl and Matthew McGrath argue that “S [knows] that p only if S is rational
to act as ifp” (Fantl and McGrath 2002, p. 78).9 According to Fantl and
McGrath, then, rationally acting as if p is a necessary condition on knowing
that p. Some argue for an even stronger relationship between knowledge
and practical rationality. John Hawthorne and Jason Stanley write, “Where
one’s choice is p-dependent, it is appropriate to treat the proposition that p
as a reason for acting iff you know that p” (Hawthorne and Stanley 2008,
p. 578).10 Similarly, Timothy Williamson maintains that the “epistemic
standard of appropriateness” for practical reasoning can be stated as
follows: “One knows q iff q is an appropriate premise for one’s practical
reasoning” (Williamson 2005, p. 231).11 Since the sufficiency of S’s knowing
that p for S to rationally act as if p is logically equivalent to the necessity of
S’s rationally acting as if p for S’s knowing that p, both Hawthorne and
Stanley and Williamson are in agreement with Fantl and McGrath, but they
also think that knowing thatp is necessary for it being epistemically appro
priate to use the proposition that p in practical reasoning. A similar view is
found in Stanley (2005), where he writes: “To say that an action is only
based on a belief is to criticize that action for not living up to an expected
norm; to say that an action is based on knowledge is to declare that the
action has met the expected norm” (Stanley 2005, p. 10). Here, Stanley sug
gests that knowledge is sufficient not only for being properly epistemically
positioned to rely on p in practical reasoning, but also for being so posi
tioned to act on p. While the subtle differences between these accounts are
9 The explicit formulation of this quotation refers to justified belief, rather than to know
ledge. But immediately following this passage, Fantl and McGrath write, “. . .it might seem that
we are imposing an unduly severe restriction on justification and therefore on knowledge”
(Fantl and McGrath 2002, p. 79). It is clear, then, that Fantl and McGrath intend for this condi
tion to apply to both justification and knowledge. See also Fantl and McGrath (2009).
10 Hawthorne and Stanley restrict their conditions to “p-dependent choices” since p may
simply be irrelevant to a given action.
11 A contextualist version of this principle is: “A first-person present-tense ascription of
‘know’ with respect to a proposition is true in a context iff that proposition is an appropriate
premise for practical reasoning in that context” (Williamson 2005, p. 227).
interesting, I will focus here on their similarities. In particular, they all

endorse at least the following:
KAP: Knowledge/Action Principle: S knows thatp only if S is epistemically

rational to act as if p or, equivalently, S is epistemically rational to act as if p
if S knows thatp.12
Now, it should be noted that the KAP holds that knowledge is sufficient for
rendering action epistemically appropriate, but there may be other senses of
propriety in which knowledge fails to be so sufficient. For instance, I may
clearly know that my drunk colleague is making a fool of himself at a
departmental party, but it may nonetheless be inappropriate for me to act as
if this is the case by confronting him about his behavior. It may be impru
dent because it would strain our friendship; or it may be impolite because it
would be utterly embarrassing to him; or it may simply be pointless because
he won’t remember my actions the next day anyway. Thus, my confronting
my friend may be inappropriate in all of these ways, while nonetheless being
epistemically proper and thus in keeping with the KAP.
The KAP is defended on a number of different grounds that apply at both
the individual and the group level. First, it has intuitive appeal. If I decide to
leave for the airport an hour later than was expected, my knowing that the
relevant flight was delayed seems sufficient to render such a conclusion
epistemically permissible. If my choice is questioned, appealing to my
knowledge adequately meets the challenge, while offering anything less—
such as my suspecting that the flight is delayed, or being justified in believ
ing that it is—does not. Similarly, if BP uses dispersants to clean up the oil
spill in the Gulf of Mexico, the company knowing them to be both effective
and safe seems adequate to render the action epistemically permissible.
If BP’s action is challenged, appealing to knowledge sufficiently meets the
challenge, while offering anything less—such as suspecting that the dispers
ants are effective and safe—does not.
Moreover, the KAP has significant theoretical power. For instance, it is
often noted that while it is epistemically inappropriate to rely on the propos
ition that one’s lottery ticket will lose in one’s practical reasoning and rele
vant actions if one has merely probabilistic evidence for this conclusion, it is
epistemically permissible to so rely on this proposition once one has learned
12 I should note that I argue in Lackey (2007) against a principle similar to KAP, but the
weaker condition I defend would support my arguments here against SK just as well.
GROUP KNOWLEDGE 119
the results of the lottery.13 The KAP, combined with the thesis that one
possesses knowledge that ones lottery ticket will lose in the latter, but not
the former, case can easily explain this data. Similarly, it is frequently
observed that when low standards are in play within a contextualist frame
work, such as in DeRose’s case of self-attributing knowledge that the bank is
open when it is not especially important that he deposit his paycheck, it
seems epistemically appropriate to rely on the proposition in question in
practical reasoning and action.14 In contrast, when high standards are in
play, such as when DeRose has just written a very large and important check
for which he needs to ensure that adequate funds will be available in his
bank account, it seems inappropriate to so rely on this proposition.15
Once again, the KAP, combined with the view that one can self-attribute
knowledge in the former, but not the latter, case can account for this data
with ease.
Finally, the KAP is said to explain what makes knowledge distinctively
important or valuable. According to Fantl and McGrath, “If you know that
p, then p is warranted enough to justify you in <£-ing, for any </>” (Fantl and
McGrath 2009, p. 66). Thus, for any instance of practical reasoning or
action, </>, one’s knowing that p is sufficient for epistemically justifying one
in </>-ing. Such a principle, Fantl and McGrath argue, “secures the distinctive
importance of knowledge” (Fantl and McGrath 2009, p. 182).16 In a similar
spirit, Hawthorne claims that “...the importance of the concept of know
ledge consists, in large part, in such [a] connection... as [that between
knowledge and action]; in turn, it seems likely that any view that severs
such [a connection] will be highly disruptive to our intuitive sense of the
epistemic landscape” (Hawthorne 2004, p. 31). And Stanley maintains that
rejecting the connection between knowledge and action “devalues the role
of knowledge in our ordinary conceptual scheme” (Stanley 2005, p. 10).
With the GMAP and the KAP in mind, let us return to the paradigmatic
instance of social knowledge: the scientific community’s knowing that d at
T2, despite the fact that no individual is aware of that d at this time.
According to the KAP, if G knows that d, then it is epistemically rational for
G to act as if d. Let us make this vivid by supposing that d is the discovery of
13 See, for instance, Hawthorne (2004). 14 See DeRose (2002).

15 See, again, DeRose (2002).
16 More precisely, they write: “We have not of course considered all possible knowledge-free
accounts of knowledge-level justification. But our discussion gives us reason to think that, at
least given fallibilism, there is no such account. If, indeed, there isn’t, it looks like KJ secures
the distinctive importance of knowledge” (Fantl and McGrath 2009, p. 182).
an enzyme that plays a role in the development of cancer cells. Given the
GMAP, a group can act only through its individual members, so G’s actions
would be through individuals, not a single one of whom needs to even be
aware that d. For the sake of clarity, let’s add to our envisaged scenario that
Dr. P., who is known worldwide as the leading researcher focusing on
cutting-edge cancer treatments, is speaking at a conference about the current
state of affairs in the scientific community, and questions about d arise. In
this context, we can thus imagine that Dr. P. is acting as a spokesperson of
sorts for the scientific community.17 When SK is combined with the GMAP
and the KAP in this way, the result is that it is epistemically rational for G,
through the actions of Dr. P. to assert that d in lectures and published work,
to approve cancer drugs that depend on d, to conduct further experiments
for cancer treatment that rely on d, to apply for grants that take d for
granted, and so on, But does this seem right?
If Dr. P. and indeed every other living member of the scientific commu
nity, has no evidence whatsoever that the enzyme in question in fact plays a
role in the development of cancer cells, then in what sense would it be epi
stemically rational for the scientific community to assert that d or approve
cancer drugs that depend on d? If the community did so happen to perform
these actions, it seems that they wouldn’t be related to the knowledge in
question, but, instead, would be entirely a matter of luck. Indeed, the com
plete absence of evidence upon which any of these actions would be based
makes them not only epistemically irrational and impermissible, but also
reckless and irresponsible. Imagine, for instance, that Dr. P. is pressed about
how precisely the enzyme plays a role in the development of cancer cells, or
whether there are any risks involved in treating cancer though targeting this
enzyme, or how robust the evidence is supporting this discovery, or whether
there are some cancers rather than others that are linked to the enzyme. He,
on behalf of the scientific community, would be utterly silent on all of these
questions. This shows that it is epistemically impermissible for the scientific
community to act on the discovery of an enzyme that plays a role in the
development of cancer cells in any of the ways that are typical of knowledge.
The upshot of these considerations, then, is that if the GMAP and the KAP
are true, SK is false: a group cannot know that p at a time when no individ
ual is aware thatp at that time.
Now it should be noted that this argument is not at odds with Bird’s
characterization of a social structure, since scientific communities generally
17 I will have a lot more to say about the role of spokespersons in groups in later chapters.
GROUP KNOWLEDGE 121
satisfy condition (iii). In particular, the published findings in scientific jour

nals are often the inputs of social action. My claim is thus not that scientific
communities never engage in social action but, rather, that when there is an
instance of mere social knowing—such that the output in question is access
ible, but not accessed, by any individual—it is not epistemically rational for
the community to act on it.
A natural response for Bird to make here is the following: a group can
rationally act on its social knowledge thatp in the absence of any individual
awareness that p when its action is the result of a goal-directed system that
is appropriately responsive to the input that p. Consider, for instance, the
following:
distributed information: The 47 members of the UN Population

Commission are collecting data for a report that they will issue, Charting
the Progress of Populations. Each member works independently and enters
the information that he or she collects into a computer that, in turn, pro
cesses the data and provides various results as outputs. One such result is
that the birth rate of Latinos is on the rise in the U.S., a conclusion of which
not a single member of the UN Population Commission, or anyone else, is
aware.18 This conclusion is then automatically sent to the New York Times,
which writes, ‘According to the UN Population Commission, the birth rate
of Latinos in the U.S. is on the rise.”
Here Bird might argue that the UN Population Commission knows that the
birth rate of Latinos in the U.S. is on the rise and that it is rational for it to
act on this knowledge by, for instance, asserting that this is the case to the
New York Times, despite the fact that there is no individual awareness of this
fact at the time of the assertion. In particular, the members of the
Commission are simply cogs in a very sophisticated system that takes the
information from the members as inputs and produces actions on behalf of
the group as outputs. And this can happen whether there is individual
awareness or not.
By way of response to this point, there are two points that I would like
to make. First, it is not clear that the output described in Distributed
Information is rightly regarded as the UN Population Commission's action.
Indeed, it is not clear that it is an action at all, let alone the UN Population
Commission’s. Is the supermarket door acting when it responds to the
18 This is a significantly modified version of a case found in Tollefsen (2007).

sensors being activated by a shopper? Is the elevator acting when it takes me

to the sixth floor because I pressed the corresponding button? Presumably
not. But then what is the salient difference between these cases and the
information about the birth rate of Latinos being sent to the New York Times?
To answer this question, let us take a step back and look at the connec
tion between belief and action. One classic role of belief is that, together
with desire, it rationalizes action.19 Thus, if you want to offer an explanation
of my approaching the coffee pot in the kitchen, you can cite my desire for
coffee and my belief that it can be found in the pot in the kitchen. But now
consider the relationship between the UN Population Commission’s pur
ported belief that the birth rate of Latinos is on the rise and its actions.20
Imagine, for instance, that the UN Population Commission has convened a
meeting with all 47 of its members to vote on a question that is directly tied
to whether this is the case, say, whether additional funding should be pro
vided for programs involving this ethnic group. If there is not a single mem
ber of the Commission that is aware of the results of their research, then
surely the group would vote against the extra funds for Latino programs in
the U.S. But then how can this be reconciled with the UN Population pur
portedly believing, and indeed knowing, that the birth rate of Latinos is on
the rise? Moreover, if the Commission did end up voting in favor of the
extra funding, it certainly would not be because its belief is rationalizing its
action in the relevant sense. In particular, there would be absolutely no con
nection between the output of the computer and the voting of the members
in such a case. Given this, there is compelling evidence that the UN
Population Commission does not believe, and hence does not know, that
the birth rate of Latinos is on the rise in distributed information.21
Thus, this strategy is not going to save the connection between social
knowledge and action. Of course, the proponent of SK can simply deny the
KAP, either in general or specifically with respect to groups. This move,
however, comes at a steep cost. For notice that it is not merely that social
knowledge will be disconnected from action in some rare or contrived
cases; rather, it will be disconnected from action in all cases. Whenever
there is an instance of mere social knowledge, and a group is said to know
that p without this state supervening on the mental states of any individuals,
the KAP would be violated. However, even if one doubts the truth of the
19 For the classic discussion of this view, see Davidson (2001).

20 I am assuming that if the UN Population Commission knows that the birth rate of
Latinos in the U.S. is on the rise, it believes this, too.
21 For additional arguments supporting this conclusion, approached from the perspective
of holding corporations morally responsible, see Velasquez (2003).
GROUP KNOWLEDGE 123
KAP, it is nearly universally accepted that knowledge bears some close

relationship with action.22 Yet social knowledge severs even the weakest
such connection since it is entirely divorced from action. This puts serious
pressure on what the value or significance of social knowing is. For if it can
not figure into explaining or rationalizing group actions, then why should
we care about it?
Moreover, those who deny the KAP typically substitute it with a weaker
epistemic norm. But such a move isn’t available to the proponent of SK since
the same sort of problem arises. For instance, consider a Justified Belief/
Action Principle: S is epistemically rational to act as if p if S justifiedly
believes thatp. Now if G knows that d, then, assuming justification is neces
sary for social knowledge, it is epistemically rational for G to act as if d.
Since, according to the GMAP, a group can act only through its individual
members, G’s actions would be through individuals, not a single one of
whom is even aware that d. So, if G did act as if d, the action would be a
matter entirely of luck. Thus, if the GMAP and a Justified Belief/Action
Principle are true, then the falsity of SK follows again. What this shows is
that the reason motivating the proponent of SK to reject the KAP leads to
the rejection of even significantly weaker epistemic norms. This is obviously
an unwelcome result, as surely some norm connecting an epistemic state
with action is correct.
3.3 Social Knowledge and Defeaters
Even if the proponent of SK were to bite the bullet about divorcing social
knowledge from group action and group responsibility, there is a further
problem facing this view. To see this, consider the following addition to the
original Case of Dr. N.:
addition: Suppose that because of their ignorance of Dr. N.’s published

paper, many members of the scientific community come to believe that not-
d at T2. Indeed, suppose that at scientific conferences and workshops, there
is often explicit collective agreement among the participants that not-d.
Because of this, the members of the scientific community act on not-d by,
for instance, asserting that not-d in lectures and published work, approving
cancer drugs that depend on not-d, conducting further experiments for
22 Indeed, I have argued against the KAP in my (2010) because I think it is too strong, but I
nonetheless accept that there is a very tight connection between knowledge and action.
cancer treatment that rely on not-d, applying for grants that take not-d for
granted, and so on.
The first point to notice here is that on every available account of group
belief, the scientific community believes that not-d at T2.23 In particular,
whether group belief is understood in terms of joint acceptance of the
proposition by the members of the collective, in terms of belief by the indi
vidual members of the group, or in functional terms, the scientific commu
nity counts as believing that not-d as the scenario is described.
The second point that is here relevant is that, like individuals, groups can
have defeaters. It may be recalled from Chapter 2 that there are two central
kinds of defeaters that are typically taken to be incompatible with justifica
tion and, therewith, knowledge at the individual level. First, there are
psychological defeaters, which can be either rebutting or undercutting, A
psychological defeater is a doubt or belief that is had by S, and indicates that
S’s belief that p is either false (i.e., rebutting) or unreliably formed or sus
tained (i.e., undercutting). Second, there are normative defeaters, which can
also be either rebutting or undercutting. A normative defeater is a doubt or
belief that S ought to have, and indicates that S’s belief that p is either false
(i.e., rebutting) or unreliably formed or sustained (i.e., undercutting).
Recall, further, that a defeater may itself be either defeated or undefeated,
and that when one has a defeater for one’s belief that p that is not itself
defeated, one has an undefeated defeater.
Applying these considerations to groups, it is fairly straightforward to
understand how collective entities could have at least some psychological
defeaters: to the extent that we understand what it means for a group to
believe that p, we also understand what it means for a group to believe that
p but also to believe that q, where q indicates that G s belief that p is either
false or unreliably formed or sustained. Indeed, in addition we find pre
cisely this sort of scenario: the scientific community believes that d because
it purportedly knows that d, but it also believes that not-d. The belief that
not-d, then, is a classic instance of a rebutting psychological defeater of the
group’s belief thatp that is not itself defeated.
Now, there are two different conclusions that might be drawn regarding
this case, neither of which is attractive for the proponent of SK. On the one
hand, it might be argued that the scientific community has a rebutting
23 For different accounts of group belief, see Chapter 1.

GROUP KNOWLEDGE 125
psychological defeater at T2, and thus the social knowledge in question has
been defeated. There are at least two problems with this response. First, in
the absence of an argument, it seems epistemically arbitrary to maintain
that the mental states of individual members of the group can contribute
negatively to social knowing, but not positively. For notice that according to
the SK, social knowing that p is determined entirely by the satisfaction of
conditions (l)-(3), which require that an appropriate social structure truly
believe that p, where that p is available to the members of the collective
entity who need it. There is, then, nothing about the mental states of either
the individual members or of the scientific community itself that contrib
utes positively to the knowledge that d at T2. So then why would their men
tal states be relevant negatively?
While I regard this first problem as sufficient for ruling out the conclu
sion that the scientific community has a rebutting psychological defeater at
T2, it is worth mentioning a second, though less serious, worry: if the scen
ario described in addition includes a defeater, then there will be less
social knowledge than might have been thought. To see this, suppose, for
instance, that there is a discovery published in an obscure journal that
definitively establishes that dinosaur extinction is the result of extremely
large-scale volcanic activity. According to the SK, this can count as an
instance of social knowledge even if the author of the published paper and
all those who were aware of it die. At this later time, however, it seems
plausible and indeed probable that a significant number of the members of
the scientific community might nonetheless believe that such extinction is
the result of a meteor impact. Thus, what might have initially seemed to be
an instance of social knowledge turns out to instead be defeated. What is
significant for our purposes here is that this scenario does not seem anom
alous or unrealistic. In many cases of purportedly social knowledge, it is
likely that there will be a significant contingency of relevant individuals
who hold conflicting beliefs that function as either rebutting or undercut
ting psychological defeaters. Moreover, among those who do not disbelieve
or have doubts about the matter in question, there almost certainly will be
many who should disbelieve or have doubts about it, given the other beliefs
that they hold. This threatens to shrink the instances of social knowledge to
a relatively small number, thereby questioning the significance of such a
phenomenon to our epistemic lives.
Given these problems with granting that the scientific community has a
rebutting psychological defeater at T2, it might be argued, on the other
hand, that the social knowledge in question has not been defeated at this
time. In particular, it might be claimed that in a situation such as that found

in addition, a group can know thatp even though the group believes that
not-p. There are two different ways this might be done. First, the proponent
of SK might support this verdict by saying that it is possible for a group to
both believe that p and believe that not-p while not having a rebutting psy
chological defeater. The problem with this strategy, however, is that it is
exempting groups from the norms governing rationality and thus removing
them from the realm of the rational altogether. For believing both that p
and that not-p is the height of irrationality, and rebutting defeaters are pre
cisely what rule out this combination of states from being epistemically
permissible. But if groups are not subject to standard norms of rationality,
then there seem to be excellent grounds for ruling them out as knowers.
Otherwise put, if a subject is not deemed irrational for believing both thatp
and that not-p, then this is excellent evidence that the subject is not a
knower at all but, rather, is a mere receptacle of information.
The second strategy for supporting the verdict that the social knowledge
in question has not been defeated at T2 is to say that it is possible for a
group to both know that p and believe that not-p because belief is not a
necessary condition on knowledge. In this way, the conclusion that the sci
entific community believes both that p and that not-p would be avoided
since there would be no group belief thatp despite there being social know
ledge thatp. But then the proponent of SK would have to explain why grant
ing the existence of this kind of social knowledge is more compelling than
the view that belief is a necessary condition on knowledge. It seems doubt
ful that this could be done, not only because of the prima facie implausibil
ity of SK, but also because there is an excellent reason why belief is nearly
universally taken to be necessary for knowledge; namely, there needs to be a
connection between the knower and the proposition known to distinguish
knowing that p from merely being in a position to know that p. Moreover,
notice that the proponent of this strategy cannot even argue for something
weaker, such as group acceptance that p, to replace group belief that p since
social knowledge can be altogether free of the mental states of the individual
members.24 There is, then, even further reason to reject that SK is providing
an account of a collective entity knowing thatp.25
24 Kallestrup (2016) endorses Bird’s arguments and is aware of the objections raised in this
chapter, but responds that his “discussion is premised on the possibility of satisfactory answers.”
As should be clear, Kallestrup’s response here is unsatisfactory, as he is acknowledging that
there are objections to his view but makes no attempt to address them.
25 Bryce Huebner argues against countenancing as knowledge phenomena like Bird’s notion
of social knowing on the following grounds: “If there is no way for the justification of a claim to
be located, checked or reproduced, and if no one is really accountable for having made it, it is
GROUP KNOWLEDGE 127
3o4 Knowing, Being in a Position to Know, and Should

Have Known
In addition to the specific problems discussed thus far with respect to social
knowledge, there is the further question why we should attribute knowledge
to collective entities in such cases in the first place. With respect to individ
uals, there is a clear difference between knowing that p and being in a pos
ition to know thatp, which itself is grounded, at least in part, in the difference
between information that has been accessed and information that is merely
accessible. For instance, if I have an unopened letter on my desk that con
tains a confession from my friend to a murder, we wouldn’t say that I know
that my friend committed the crime prior to my opening it and reading its
contents.26 Instead, we would say that I am in a position to know this.
Indeed, it would not only be bizarre for me to assert that my friend commit
ted the murder or to act on this by reporting her to the police if the infor
mation is merely accessible, but not accessed; it would also be reckless and
irresponsible. When the police ask why I’m reporting my friend, for
instance, I would have absolutely nothing to offer by way of support.
But why would the situation be any different when groups are concerned?
If the discovery of an enzyme that plays a role in the development of cancer
cells is published in a journal article that is accessible to, but not accessed by,
any living scientist, why wouldn’t we provide the same verdict as we did in
the individual case: the scientific community is in a position to know this
discovery, but it doesn’t know it? Not only is this the more intuitive descrip
tion, it also accords well with the disconnect that was earlier discussed
between social knowledge and action. In particular, there is such a disconnect
hard to see why this should count as knowledge at all” (Huebner 2014, p. 214). See also Kukla
(2012). By way of response, Deborah Tollefsen argues that “what seems to be motivating
[Huebners] skepticism is an epistemic internalism that requires for knowledge the ability to
give a justification or to have access to reasons. No one person seems to be able to provide a
justification or has access to reasons, and so no one, including the group, can be held respon
sible. But why not take an externalist approach and think of group knowledge as the result of a
reliable process? There are various ways to preserve the notion of epistemic responsibility
within an externalist theory of knowledge. It may be that large-scale scientific collaboration is,
in its current form, not using reliable processes, but this is an empirical question—one that a
science of distributed cognition might one day answer” (Tollefsen 2014). (See Klausen (2015)
for an endorsement of Tollefseris response.) As should be clear, my arguments in this chapter
appeal to general features of knowledge, such as its close connection with epistemically per
missible action and its incompatibility with defeaters, and so even if Tollefsen’s response works
against Huebers (and Kukla’s) view, focusing on epistemic externalism will not help here.
26 I am assuming, of course, that I have no other relevant evidence about my friend commit
ting the crime.
precisely because this phenomenon involves a group being in a position to

know, rather than knowing, and there isn’t an intimate connection between
the former and action as there is with the latter.
Moreover, the mere fact that social knowledge isn’t knowledge after all
doesn’t mean that we can never hold groups responsible for the information in
question. Let’s consider the individual case, first. Suppose that not only is the
confession from my friend sitting unopened on my desk, but also that my friend
explicitly told me to read it before heading over to the police station for ques
tioning. Here it might be appropriate to hold me responsible for the content
of the letter since I should have known that my friend committed the crime.27
Similar considerations apply with respect to groups. Perhaps it is not only
that the discovery of an enzyme that plays a role in the development of can
cer cells is published in a journal article that is accessible to any living scien
tist, but also that practicing oncologists have an epistemic responsibility to
have read the article. As in the individual case, it might be appropriate to
hold the oncology community responsible for this discovery since it should
have known about it.
Thus, not only does social knowledge lead to unacceptable epistemo
logical consequences, but the role that it would play can be better filled by
other states, such as being in a position to know and should have known.
We can conclude, then, that knowledge is not socially extended in the rad
ical sense found in SK, and thus does not pose a problem for the view of
justified group belief developed in Chapter 2.
3.5 Collective Knowledge
The second view of group knowledge, which is not only highly influential
but also at odds with my own account of justified group belief, is found with
the application of the “collective knowledge doctrine” in U.S. law. In the
case of the United States v. Bank of New England, for instance, the Bank of
New England was charged and convicted of thirty-one violations of the
Currency and Foreign Transaction Reporting Act. The details of the case are
as follows: from May 1983 through June 1984, James McDonough visited
the Prudential branch of the Bank of New England on thirty-one separate
27 While I understand “should have known” in terms of normative defeat, see Goldberg
(2017) for another view.
GROUP KNOWLEDGE 129
occasions to withdraw money from a corporate account. On one such

occasion, McDonough presented the bank teller with two checks made
payable to cash in the amounts of $8500 and $5000. According to the
Reporting Act, a Currency Transaction Report (CTR) must be filed when
ever a cash withdrawal is made that exceeds $10,000 and it is a violation to
willfully fail to file such a report. The teller on duty was unaware of the
Reporting Act, while the teller’s supervisor was aware of the Act, but did not
know that the customer’s two deposits had to be aggregated for purposes of
the reporting requirement. The bank’s project coordinator, who was
working in the bank’s main office, knew that the law required aggrega
tion, but had no knowledge that the transaction in question occurred”
(Ragozino 1995, p. 433). None of the three employees of the bank indi
vidually committed a criminal violation of the Act, then, because none,
individually, willfully failed to file a CTR. However, according to the “col
lective knowledge doctrine,” which resulted from this case, the knowledge
of the individual employees can be added or aggregated and then prop
erly attributed to the bank itself. Indeed, according to the trial court’s
instructions to the jury considering the case: “if Employee A knows one
facet of the currency reporting requirement, B knows another facet of it,
and C a third facet of it, the bank knows them all” (Hagemann and
Grinstein 1997, p. 214). Given this, the knowledge of the three individual
employees—that is, the teller’s knowledge that two deposits exceeding
$10,000 had been made, the head teller’s knowledge of the reporting
requirement, and the coordinator’s knowledge that multiple deposits
must be aggregated—can be combined and then imputed to the bank.
Hence, it was concluded that the Bank of New England satisfied the
knowledge requirement needed for establishing mens reay thereby lead
ing to a guilty verdict with respect to the violations of the Currency and
Foreign Transaction Reporting Act.
In order to assess this notion of collective knowledge in greater detail,
let’s consider a structurally similar case, which will allow us to determine
whether the doctrine in question is generally true. Note that if it is not, this
is reason to think there is an alternative explanation for why we are inclined
to hold the Bank of New England accountable for its violation of the report
ing obligation. Suppose that three police officers all work for the same unit
of the Chicago Police Department (CPD). Officer A knows (1): that a seven
year-old child, Jimmy Smith, has been reported missing from the Rogers
Park neighborhood of the city since this morning. He knows this because it
was communicated to him by his superior. Officer B knows (2): that Jimmy
Smith was wearing a Frida Kahlo t-shirt this morning because he lives next
door to the Smith family, and he remembers commenting on how he loves
Frida Kahlo’s work when he saw Jimmy walk out of the house. And Officer
C saw a seven-year-old wearing a Frida Kahlo t-shirt walking with an adult
man while he was patrolling a park in Edgewater, the neighborhood just
south of Rogers Park. Officer A knows only (1), but not (2) or (3); Officer B
knows only (2), but not (1) or (3); and Officer C knows only (3), but not
(1) or (2). According to the collective knowledge doctrine, the knowledge
of the individual police officers can be properly attributed to the Chicago
Police Department as a group.28 Thus, the CPD knows (1), (2), and (3), even
though no single police officer knows this.
But let’s examine this conclusion a bit closer. As should be clear, if any
one—individual or group—knows all three of these facts, then the knowing
agent should approach the seven-year-old in the park to determine whether
he is Jimmy Smith. This is especially true if the knowledge in question is
had by a police officer or unit charged with finding the missing child. Not
only is this supported by the Knowledge/Action Principle discussed at
length in sections 3.2-3.4, it is also intuitive. If a police unit knows that a
child has just gone missing, and it also knows that a child fitting a very spe
cific description that is known to be true of the missing boy is now walking
through a park in a neighborhood close to where he was last seen, it
undoubtedly should check to see if the child is Jimmy Smith.29 Imagine, for
instance, that the child is in fact Jimmy Smith, and that he is walking with
his abductor in the park, and his parents learn that a police officer who is
part of a unit that purportedly knew (1), (2), and (3) failed to intervene.
They would, quite rightly, be filled with outrage and despair but, perhaps
even most of all, confusion. How, they might ask, could you know (1), (2),
and (3) and yet allow our son to walk past you in the hands of his abductor?
Notice, though, that none of this is true of the scenario as described
above. It is not at all puzzling why the police unit, and Officer C in particu
lar, did not approach the young boy in the park. The unit could quite easily
explain to the parents that while each officer had bits of relevant knowledge,
they did not communicate effectively, and so Officer C had no idea that a
28 At the very least, the knowledge can be imputed to the unit of the Chicago Police
Department that the three officers in question belong to, but nothing in what follows turns on
this distinction.
29 I take it that not many young children in Illinois wear Frida Kahlo t-shirts.
GROUP KNOWLEDGE 131
child was even missing from Rogers Park when he saw the boy in Edgewater.
Indeed, it would be deeply problematic for him to approach what is from
his perspective an ordinary child walking in the park with an adult man,
with no evidence that there is a problem. If Officer C did stop Jimmy Smith,
what would he be able to say to his abductor upon being asked why he was
approaching them? Nothing at all relevant, it seems. And if he did succeed
in preventing Jimmy Smith from being abducted, this would not be in any
way the result of something creditable to him or the police unit. It would be
pure luck. But action that is guided by knowledge is not the result of luck in
this way, and so there is very good reason to deny that the police unit in fact
has knowledge, as the collective knowledge doctrine says.
This problem is similar to the one discussed in section 3.4 regarding
social knowing. Collective knowledge, as it is understood in U.S. law,
divorces knowledge from the intimate connection it has with action. As we
saw, knowledge is said to be both necessary and sufficient for epistemically
permissible action. If, for instance, one knows that p, then one is properly
epistemically positioned to act as if p. But collective knowledge that p does
not make one properly epistemically positioned to act as ifp. In the absence
of Officers A, B, and C communicating their individual pieces of knowledge
with one another, any action taken by the police unit as a collective or as
individual officers to question Jimmy Smith or his abductor would be
unwarranted. Indeed, it would be no different than an unrelated group of
officers stopping to question a random man and child walking in a park.
Not only does this raise epistemic concerns, as I have been arguing here, but
there are also ethical and legal problems with people who are going about
their lives being stopped by law enforcement for no reason whatsoever.
Since the connection that knowledge has with action is one of its distinctive
features, and is even said to be what makes knowledge valuable, we have
reason for concluding that collective knowledge is not knowledge after all.
The proponent of the collective knowledge doctrine may respond by
pointing to the case of United States v. Whitfield (2011), as it involves deter
mining the extent of the application of the collective knowledge doctrine in
U.S. law. Camden police officers in three marked police vehicles were
patrolling an area of the city known for violence and drug activity involving
crack cocaine. One police officer observed a hand-to-hand exchange involv
ing the defendant in an area of the city known for drug activity while two
different officers apprehended the defendant without having communicated
with the first one. The defendant challenged the legality of his seizure
because the officer who apprehended him did not witness the hand-to-hand
exchange. However, citing the collective knowledge doctrine, the Third

Circuit of the United States Court of Appeals upheld the seizure on the
grounds that “the knowledge of one law enforcement officer is imputed to
the officer who actually conducted the seizure, search, or arrest.” In particu
lar, the Court reasoned that it is unnecessary for one officer, who is working
with his fellow officers as a “unified and tight-knit team,” to communicate to
the other officers all of the information relevant to the seizure in question!
It would make little sense to decline to apply the collective knowledge

doctrine in a fast-paced, dynamic situation such as we have before us, in
which the officers worked together as a unified and tight-knit team;
indeed, it would be impractical to expect an officer in such a situation to
communicate to the other officers every fact that could be pertinent in a
subsequent reasonable suspicion analysis. Applying the collective know
ledge doctrine here, there is little question that there was reasonable suspi
cion to seize Whitfield.
Thus, the Third Circuit found that collective knowledge of the group of
officers as a whole justified the seizure of the defendant in question.
The important dimension of the Third Circuit’s reasoning here is that it is
the fact that the officers worked together as a “unified and tight-knit team”
that purportedly justifies the application of the collective knowledge doc
trine. But in what sense did the officers work as a team? In case law, the col
lective knowledge doctrine has two approaches that should be distinguished
in answering this question. According to the vertical approach) one law
enforcement officer who possesses probable cause may instruct another
officer to act without communicating the requisite knowledge in question.
For instance, in United States v. Hensley (1985), an informant told a police
officer in St. Bernard, Ohio that Hensley had driven the getaway car from an
armed robbery in St. Bernard six days earlier. The officer put out a “war
ranted flyer” to nearby police departments, which described Hensley and
the crime for which he was being sought, and asked that he be picked up
and held if seen. The police department in Covington, Kentucky read the
flyer to its officers and, on this basis, the Covington police pulled Hensley
over upon spotting him. The United States Supreme Court, relying on an
earlier case Whiteley v. Warden (1971), argued:
[L]anguage in Whiteley suggests that, had the sheriff who issued the radio
bulletin possessed probable cause for arrest, then the [arresting] police
GROUP KNOWLEDGE 133
could have properly arrested the defendant even though they were
unaware of the specific facts that established probable cause. Thus
Whiteley supports the proposition that, when evidence is uncovered dur
ing a search incident to an arrest in reliance merely on a flyer or bulletin,
its admissibility turns on whether the officers who issued the flyer pos
sessed probable cause to make the arrest. It does not turn on whether
those relying on the flyer were themselves aware of the specific facts which
led their colleagues to seek their assistance. In an era when criminal sus
pects are increasingly mobile and increasingly likely to flee across jurisdic
tional boundaries, this rule is a matter of common sense: it minimizes the
volume of information concerning suspects that must be transmitted to
other jurisdictions and enables police in one jurisdiction to act promptly
in reliance on information from another jurisdiction.
Thus, the Supreme Court reasons here that when an officer who issues a
warranted flyer possesses the requisite knowledge for probable cause, and
instructs other officers to act on this, a corresponding arrest is proper even
when the arresting officer lacks the knowledge in question. Further insight
into the reasoning underlying this can be seen by looking to Commonwealth
of Pennsylvania v. Yong, where the Pennsylvania Supreme Court argues that:
“Read jointly, Whiteley and Hensley instruct that the collective knowledge
doctrine serves an agency function. When a police officer instructs or
requests another officer to make an arrest, the arresting officer stands in the
shoes of the instructing officer and shares in his or her knowledge.”
This vertical approach to the collective knowledge doctrine is not at all at
odds with the account of justified group belief outlined in the previous
chapter. There are at least two ways of understanding this. First, the police
departments working together to apprehend a suspect might be construed
as forming a unified group for this purpose, where a significant percentage
of the operative members possess the requisite knowledge for probable
cause. In particular, the officers who put out the flyer can be seen as the
operative members of the group who have the relevant justified belief, and
so the mere fact that the arresting officer lacks it does not pose a problem.
Moreover, notice that there is no sense whatsoever in which the arrest of the
suspects in these sorts of cases is random or lucky. Instead, there is a very
close connection between the knowledge and action—it just occurs through
the group member(s) who has it asking another to act on it. A second, and
related, understanding of the vertical approach to the collective knowledge
doctrine is where the arresting officer (or officers) is understand as a proxy
agent30 for the knowing officer (or officers). On this reading, the arresting
officer’s actions constitute the actions of the knowing officer, and so there is
no gap between knowledge and action. Since the arresting officers actions
just are the knowing officer’s actions, the knowledge in question is guiding
the detainment and arrest. This is made possible at least in part through the
proxy agent knowingly acting with the delegated authority of a proxy. The
arresting officer’s knowledge of the possession of probable cause is part of
what allows his action to be guided by that knowledge, even in the absence
of his own individual beliefs as to what that probable cause is. Once again,
the account of justified group belief defended in the previous chapter sup
ports this reading. The knowing officer is an operative member who indi
vidually possesses the relevant justified belief that serves as the foundation
for the group’s knowledge, and the arresting officer is functioning as a proxy
agent for him.
On both of these readings of the vertical approach to the collective
knowledge doctrine, the sense in which a group of officers is working as a
“unified and tight-knit team” is clear. There are direct channels of commu
nication between individuals or subgroups that provide excellent evidence
to the arresting officers that other members of the group have knowledge of
probable cause, even if detailed information about this is not conveyed. This
is fairly standard when working as collectives. Members of groups often fol
low norms that allow one another to trust that each is doing his or her share
of the work. This enables streamlined communication that still secures an
appropriate connection between epistemic states and action without requir
ing the conveying of detailed information that would make group work far
less efficient and effective. Importantly, these norms also allow the proper
apportioning of blame when individual members do not meet their pre
scribed obligations. For example, if the superior officer has misread the bul
letin from a neighboring jurisdiction, the blame for making an unwarranted
arrest would fall on his shoulders rather than on those of the arresting
officer.
There is, however, a second, horizontal approach to the collective know
ledge doctrine in U.S. law that is “broader” and at work in United States v.
Whitfield, On such an approach, “the probable cause assessment is not
focused on a single officer’s knowledge; rather, probable cause is assessed
by aggregating the knowledge of two or more law enforcement officials
30 As already mentioned, I will discuss proxy agency in far more detail in the final two
chapters.
GROUP KNOWLEDGE 135
working together.” (Commonwealth ofPennsylvania v. Yong) In Commonwealth

of Pennsylvania v. Yong (2015), the majority opinion states that courts that
rely on the horizontal approach “have ignored the original aim of the
rule” by “eliminating the requirement that officers actually communicate
with each other” Moreover, the majority reasoned that “an expansive
interpretation of the collective knowledge doctrine does not comport
with the fundamental requirement that warrantless arrests be supported
by probable cause” What we see here is that the horizontal approach to
the collective knowledge doctrine is far from uncontroversial in
U.S. courts. Indeed, the mere aggregation of the knowledge of individual
members of a group is often explicitly rejected as sufficient for the knowledge
required for probable cause. This is the case, not only because the courts
have raised doubts about groups possessing collective knowledge in the
absence of any relevant communication whatsoever, but also because of
the potential dangers of such attributions.
In Commonwealth of Pennsylvania v. Yong, concerns very similar to those
raised in earlier sections of this chapter are articulated with respect to the
horizontal approach: “under any approach that permits aggregation of
unspoken information or justifies actions taken absent direction from a
person with the necessary level of suspicion, there remain serious concerns
for protecting citizens from unconstitutional intrusions.” The majority con
tinues by quoting from United States v. Massenburg (2011, emphasis added):
No case from the Supreme Court...has ever expanded the collective

knowledge doctrine beyond the context of information or instructions
communicated (“vertically”) to acting officers. Some of our sister courts
have authorized “horizontal” aggregation of uncommunicated informa
tion. See United States v. Ramirez, 473 E3d 1026, 1032-33 (9th Cir. 2007)
(collecting cases)....
The rationale behind the Supreme Court’s collective-knowledge doctrine
is, as the Court noted in Hensley, a “matter of common sense: [the rule]
minimizes the volume of information concerning suspects that must be
transmitted to other jurisdictions [or officers] and enables police... to act
promptly in reliance on information from another jurisdiction [or
officer].” Hensley, U.S. 469 U.S. at 231. Thus, law enforcement efficiency
and responsiveness would be increased[.]...
The Governments proposed aggregation rule serves no such ends. Because
it jettisons the present requirement of communication between an
instructing and an acting officer, officers would have no way of knowing

before a search or seizure whether the aggregation rule would make it legal,
or even how likely that is. The officer deciding whether or not to perform a
given search [or seizure] will simply know that she lacks cause; in ordin
ary circumstances, she will have no way of estimating the likelihood that
her fellow officers hold enough uncommunicated information to justify
the search. And as an officer will never know ex ante when the aggregation
rule might apply, the rule does not allow for useful shortcuts when an
officer knows an action to be legal, as Hensley did, Perhaps an officer who
knows she lacks cause for a search will be more likely to roll the dice and
conduct a search anyway, in the hopes that uncommunicated information
existed. But as this would create an incentive for officers to conduct
searches and seizures they believe are likely illegal, it would be directly
contrary to the purposes of longstanding Fourth Amendment jurispru
dence. (Massenburg, 654 E3d at 494.)
Note especially the points emphasized in this lengthy quote. The Fourth
Circuit is unambiguously stating that acting on "collective knowledge”
understood merely horizontally—where individual bits of knowledge are
simply aggregated and attributed to the group in the absence of any relevant
communication—severs the connection between knowledge and action.
The best-case scenario from an epistemic point of view is that the arresting
officer will be acting on an estimate of how likely it is that there is probable
cause rather than on knowledge of probable cause. But estimates of this sort
clearly do not amount to knowledge in any reasonable sense. The worst-case
scenario is that the arresting officer will simply gamble with respect to the
lives of others and conduct a search with the hope that uncommunicated
information exists for probable cause. Again, this seems to be a far cry from
collective knowledge.
Building on this view from the Fourth Circuit, the Pennsylvania Supreme
Court says:
In light of these concerns, we cannot acquiesce to the Commonwealths

request to broadly interpret the collective knowledge doctrine and adopt
an unrestricted horizontal application.... we will not endorse an approach
that has the potential of encouraging police without the requisite level of
suspicion to infringe on a person’s freedom of movement in the hopes that
his or her fellow officers possess such level of suspicion. See Massenburg,
654 F.3d at 494.
GROUP KNOWLEDGE 137
As should be clear, I entirely agree with the Pennsylvania Supreme Court

here; indeed, I think the Court is offering an argument very similar to the
one against social knowing from the previous sections. The Court is
unequivocally saying that the mere aggregation of individual bits of know
ledge—without any relevant communication—just isn’t collective know
ledge, Because knowledge warrants the corresponding action, calling this
knowledge illegitimately sanctions searches and seizures that infringe on
people’s rights. This is not only epistemically problematic, it is also
morally and legally wrong. Otherwise put, knowledge has a very close
connection with action; it is, for instance, typically sufficient for
epistemically permissible action. But searching someone on the basis of
horizontally understood collective knowledge is impermissible. So, we
have reason to conclude that collective knowledge, so understood, isn’t
knowledge after all.
Moreover, as I argued in the previous sections of this chapter, the mere
fact that a group of police officers does not have the knowledge in question
does not eliminate attributions of responsibility. Perhaps the group is such
that they should have communicated with one another the information
necessary for probable cause and thus should have possessed the relevant
knowledge. We can certainly criticize both the individual officers and the
collective as a whole for failing to do what they ought to have done. Or per
haps the officers are in a position to know the relevant facts, and so they are
subject to disapproval for failing to do further epistemic work. However the
details are fleshed out, the main point is that the absence of an attribution of
knowledge does not in any way eliminate the attribution of responsibility.
3.6 Conclusion
We have seen, then, that two of the more serious challenges to the account
of justified group belief defended in Chapter 2 can be avoided. Both social
knowledge and collective knowledge sever the crucial connection between
knowledge and action, and open the door to serious abuses, not only epi
stemically, but morally and legally as well. Bits of information that are
merely accessible to group members, or individual instances of knowledge
that are aggregated with no communication, do not amount to group
knowledge in any robust sense. Let’s now turn in the remaining chapters to
some of the things that groups can do, particularly as they relate to their
states of believing and justifiedly believing.
Group Assertion
Groups make assertions all the time. It is nearly a daily occurrence to hear a
university announcing a new initiative, a police department denying a
charge of brutality, or a company reporting information about its financial
value. Yet despite the frequency with which we take it at face value that
groups do offer assertions, there is a shocking paucity of philosophical work
on this topic. This chapter aims to at least begin the process of filling this
gap in the literature.
As we have seen in the previous chapters, there are, broadly speaking,
two different approaches to understanding collective phenomena in
general. On the one hand, there is a deflationary approach, according to
which explaining such phenomena does not require new theoretical
resources; rather, we can simply rely on our grasp of the same phenomena
at the individual level. This is because the states and acts of groups just are
the states and acts of individual members “summed up.” For instance, a
deflationary view of group assertion holds that a group asserts a proposition
just in case some of the members of the group assert the proposition. Which
members are here relevant, and under what circumstances their assertions
count as the groups, need to be fleshed out, but the core idea is clear: group
assertion is reducible to the assertions of individual members. Thus, to the
extent that we understand individual assertion, we have the central
resources for explaining group assertion.
According to an inflationary approach, on the other hand, collective
phenomena cannot involve the mere summing up of the same phenomena
at the individual level. This is because there can be states and acts of groups
where there is no corresponding state or act of a group member. For
instance, an inflationary view of group assertion holds that a group can
assert a proposition even when not a single member of the group does. In a
very important sense, then, collective phenomena are over and above any
phenomena at the individual level.
In this chapter, I will develop and defend an inflationary approach to
what I will argue is the core kind of group assertion: authority-based. I will
show that groups can offer assertions even when no members of the group
The Epistemology of Groups. Jennifer Lackey, Oxford University Press (2021), © Jennifer Lackey.
DOI: 10,1093/oso/9780199656608,003,0005
GROUP ASSERTION 139
do, and thus that it is the group itself that is doing the asserting? This
conclusion is further supported, I will argue, by the fact that it is the group,
rather than any individual, that is subject to the norm or norms governing
assertion. Finally, I will explain why I regard assertion as being unlike the
collective phenomena discussed in the previous three chapters in demanding
that it be understood in straightforwardly inflationary terms.
4.1 Two Kinds of Group Assertion
Let’s begin by highlighting some key dimensions of group assertion. Very

roughly, there are two ways in which a group might assert that p: first, a
group may assert that p through all of its members reasonably intending to
convey that p together in virtue of coordinated individual acts. Let us call
this coordinated group assertion. An instance of this sort of group assertion
is where the members of a tour group stranded on a desert island work
together to form the words “We Need Help” in the sand. All of the members
coordinate individual acts of communication that together convey the view
of the group as a whole. Another example of coordinated group assertion is
where all of the members of a research team collectively draft an article
together, such as through Google Docs. If such members work collabora
tively to literally compose, say, a single sentence—much like the members of
the tour group put together the message in the sand—then this is the asser
tion of the research team. This should be distinguished from a case where
each member of a group writes different parts that together make up a sin
gle article, which is how some co-authored or collaborative work is done.
Whereas the former is an instance of coordinated group assertion, the latter
is merely a collection of individual assertions.
While coordinated group assertion is surely important, the far more
common kind of group assertion, and the one that I think has not been fully
appreciated, is that offered through an authorized spokesperson(s). I am
understanding the notion of a spokesperson(s) as subsuming any set of
1 To avoid confusion, I should note that I argued in Lackey (2014a) on behalf of a deflationary
account of group testimony. But while my topic here is on what we might call the metaphysics
of group assertion or group testimony (where I here use “assertion” and “testimony”
interchangeably)—i.e., what is it for a group to assert or testify—the account in Lackey (2014a)
takes up the epistemology of group assertion or group testimony—i.e., how do we acquire justi
fied belief or knowledge via group assertion. Thus, my view is inflationary in a metaphysical
sense, but deflationary epistemologically.
individuals that is distinct from a group as a whole and that speaks on a

groups behalf with the proper authority. A spokesperson might be a member
of a group, such as when the chair of a philosophy department has the
authority to speak on behalf of the department when hiring decisions are at
issue. Alternatively, a spokesperson might not be a member at all, such as
when a lawyer is hired to speak on a philosophy departments behalf where
pending litigation is concerned. The point that I wish to emphasize here,
however, is that the standard way in which a group asserts is through
an authorized spokesperson(s).2 Whenever a group asserts through an
authorized set of individuals that is smaller than the group as a whole, let us
call this an authority-based group assertion. This kind of group assertion will
be the central focus of this chapter.
How should we understand a spokespersons having the requisite author
ity to speak on a groups behalf? This question has been largely absent from
work on collective phenomena, but a notable exception is found in Kirk
Ludwigs (2014). Since his is the only extended discussion of this issue in the
literature, it is worth considering in some detail.
4.2 Having the Authority to Be a Spokesperson
Ludwig’s central aim is to provide an account of what he calls proxy agency,

where “one person or subgroup’s doing something counts as or constitutes or
is recognized as (tantamount to) another person or group’s doing some
thing” (Ludwig 2014, p. 76). According to Ludwig, being a spokesperson for
a group is a paradigmatic instance of being a proxy agent and is also what
John Searle calls a status function.3 “The two core ideas in the concept of a
status function are that some object or thing or person has a certain social
function, a function in certain social transactions, and that it has that func
tion in virtue of its having acquired a certain status among a relevant group
of people by way of their attitude toward it” (Ludwig 2014, p. 87). A stand
ard example used to clarify the concept of a status function is money: twenty
dollar bills are simply pieces of paper unless the relevant group members,
such as buyers and consumers, agree to their having a particular social
2 It is also possible for a group to be structured so that every member is authorized as a

spokesperson for the group. I’m grateful to an anonymous reviewer that led to the inclusion of
this point.
3 See Searle (1995).
GROUP ASSERTION 141
status. Similarly, Ludwig argues that individuals making assertions are

spokespersons only when their status as such is granted by the relevant
members of the community, which here includes all members of the group
and audience in question. In particular, “a status function is a property an
object has in virtue of people so regarding it.. .that enables it to play a cer
tain role in a social transaction. This has to include all who participate in
the social transaction. The announcing group’s authorization of an individ
ual plays its role only in the context of an action plan that specifies its func
tion relative to a collective action by a larger group” (Ludwig 2014, p. 89).
Let us call Ludwigs view here the status function model of being a spokes
person. One of the upshots of this approach is that the actions of proxy
agents do not themselves constitute the actions of groups. This is because
“in one way or another, group action through proxy agency calls upon every
member of the group to contribute” (Ludwig 2014, p. 100). Thus, when a
spokesperson asserts, this is the culmination of the activities of all of the
group members, such as their granting authority to the spokesperson to
speak on their behalf or their agreeing to the institutional arrangements
that provide such authority.
There are, however, at least two central problems with this status function
conception of spokespersons. The first is that a groups asserting does not
depend on audience recognition. More precisely, groups can assert through a
spokesperson not only when the audience members fail to regard the
speaker as playing the role of spokesperson for the group, but also when
they reject both her status and her corresponding assertion. Ludwig expli
citly denies this possibility He writes, “... if the group identifies a possible
mechanism for group announcement, but doesn’t communicate it to the
audience, or if the audience doesn’t find it acceptable, and so refuses to pay
attention, then the group fails to achieve its aim. This would be analogous to
someone declaring a certain object was to be the royal seal without getting
others to go along with it” (Ludwig 2014, p. 93). But suppose, for instance,
that the police chief of a nearly universally sexist community has the sole
power to designate a spokesperson to represent the department with regard
to a highly publicized murder investigation, and for the first time in history,
he appoints a woman to this role. Let’s call her Jane. Suppose further that
the community finds the appointment of a woman utterly unacceptable and
so refuses to listen to, or accept, anything that Jane asserts. Has Jane asserted
on behalf of the police department? To my mind, the answer is clearly yes.
What is needed for a spokesperson to assert on behalf of a group is that she
has the authority to do so, regardless of whether this is recognized by the
audience members.4 If, for instance, it is written into the police departments
policies and procedures that the chief has the sole power to appoint the
spokesperson, and he so appoints Jane, then she has the authority to speak
for the department even if the sexist community rejects that she is playing
this role and ignores everything that she says.
This verdict is paralleled in the individual case: those who are ignored do
not fail to be asserters or fail to assert; instead, they are the victims of testi
monial injustice5 or their assertion fails to achieve its desired aim of uptake.6
If, for instance, a woman asserts that she does not want to have sex with her
partner, she is asserting this, even if her assertion is ignored or refused.
Indeed, even if her partner is such that he does not accept her status as an
asserter of refusals of sex in general, she is still an asserter. This is because
she has the authority to refuse unwanted sexual advances even when her
authority is ignored or rebuffed. In this way, I disagree with Ludwig’s claim
that a spokesperson asserting on behalf of a group without audience accept
ance “would be analogous to someone declaring a certain object was to be
the royal seal without getting others to go along with it.” It would instead be
analogous to an individual being the victim of testimonial injustice.7
Consider, also, some consequences of requiring audience recognition or
acceptance in order for a spokesperson to assert on behalf of a group.
4 This view is supported by our general practices involving spokespersons. For instance, in a
recent CNN article addressing whether President Trump called for an end to Robert Muellers
investigation into collusion with Russia, it is reported that Trump’s attorney, John Dowd, said
that "he was speaking on his own behalf, although he had earlier told the Daily Beast, which
first reported the statement, that he was speaking on behalf of the President. Dowd’s comment
wasn’t authorized by the President, a person close to...Trump told CNN” (https://www.cnn.
com/2018/03/17/politics/john-dowd-mueller-russia-investigation/index.html, accessed March
18, 2018).
5 See Pricker (2007).
6 Some read Austin (1962) as requiring uptake in order for illocutionary speech acts to be
successful. See, for instance, Langton (2009). For objections to the uptake requirement, see
Antony (2011). Fricker (2012) applies this reading of Austin to testimony, writing “Without my
uptake, whatever you may succeed in doing with your words, it won’t be quite testifying”
(Fricker 2012, p. 254). Even if this is a correct reading of Austin, there are at least three worries
with applying it to testimony or assertion. (For our purposes here, we can treat testimony and
assertion interchangeably.) First, there is not a single view in the literature of what it is to testify
that supports the uptake requirement. (See, for instance, Coady (1992), Fricker (1995), Audi
(1997), Graham (1997), Elgin (2002), and Lackey (2008).) Second, this view has the conse
quence that one does not testify in a private diary that is never read, in a courtroom when one
is not believed, and so on. Third, if one takes someone to be lying and thus there is no uptake,
then there is, on this view, no assertion. If asserting is a necessary condition on lying, then we
get the result that the known liar cannot lie. For all of these reasons, uptake should not be taken
to be necessary for testifying or asserting.
7 Just to be clear, the parallel is as follows: just as individuals can assert in the absence of
audience recognition, so, too, can groups assert via spokespersons without such recognition.
GROUP ASSERTION 143
Women would systematically be denied the ability to serve as a spokesperson

in sexist communities, Black people would be unable to do so in racist
societies, and so on. Moreover, suppose that Jane asserts on behalf of the
police department at T1 when the community is sexist, but then years later
at T2 the community has changed and is now accepting of her role as the
group’s spokesperson. Her statement would go from failing to be the group’s
assertion at T1 to being the group’s at T2 and thus it would be incapable of
functioning as evidence of the group’s view at T1 but not at T2. These
conclusions all seem problematic, as they conflate one’s asserting with ones
asserting being properly appreciated.
The second problem with Ludwig’s status function view is that whether a
spokesperson asserts on behalf of a group does not require that the members of
the group accept or recognize the authority of the spokesperson-, the spokes
person simply has to have the authority. Suppose that all of the fellow police
officers of the sexist department above refuse to accept the policy that per
mits a woman to be the authorized spokesperson for the department. Thus,
even though Jane is appointed as the spokesperson for the department, her
statement would not be the group’s assertion on Ludwigs view because all
members of the group need to accept the institutional arrangements that
provide such authority. But if the policies and procedures of the police
department do not require agreement or consensus in order for authority to
be given to Jane to serve as the spokesperson, then the members being
disgruntled or unhappy doesn’t make it the case that she is not asserting on
behalf of the group.
By way of response to this sort of worry, Ludwig argues that simply by
virtue of agreeing to be a member of a group, one thereby accepts the
policies and procedures of the group. He writes: “since meeting the mem
bership condition requires endorsing the division of roles and responsibilities
(that is partly what defines the role of membership), anyone who joins such
a group explicitly endorses its arrangements,8 in accepting membership,
and in that act then contributes constitutively to the authorization of its
various roles” (Ludwig 2014, p. 97). But there is a dilemma facing this
response: either group membership does not require the acceptance of the
8 Explicit endorsement of a groups policies and procedures is a very strong requirement for
group membership. Ludwig argues that this requirement is true only of “genuine organiza
tions,” where “members choose to join and, hence, agree to the conditions of membership,
which includes an endorsement of the institutional arrangements” (Ludwig 2014, p. 97). As I
will argue later, however, I think that members can join groups without such an endorsement.
policies and procedures of the group or the notion of acceptance operative

here is vacuous.
To see this, consider, first, a sabotaging member of a group: suppose that a
police officer in the sexist police department becomes convinced of the
moral wrongness of the sexism of his group and (i) rejects all of the sexist
policies and procedures of the department, and (ii) actively works to under
mine them. Surely this officer is still a member of the police department in
question, yet it is not at all clear how he accepts the groups policies and
procedures. Acceptance is typically understood as being such that it would
manifest itself in one’s actions. So, for example, one might be said to accept,
even if one does not believe, that smoking is safe if one would assert that
smoking is safe, act as if smoking is safe, defend the safety of smoking, and
so on.9 This is because acceptance often results from a consideration of one’s
goals, such as the financial aim of making smoking appealing. But the sabo
taging members actions support his rejecting rather than accepting the
institutional arrangements of the police department.
Of course, Ludwig might say that the very joining of the police depart
ment by the sabotaging member brings with it an acceptance of its policies
and procedures, even if this wouldn’t be manifested in any of his actions.
But the sense of acceptance that must be operative here is so thin that it is
vacuous. To make this even clearer, consider a sabotaging joiner of a group:
suppose that a new police recruit joins the department precisely because he
rejects all of the sexist policies and procedures of the department and wishes
to actively work to undermine them. To my mind, sabotaging joiner is
clearly a member of the police department—he has the same authority and
benefits as all of the other members of the department, receives a paycheck
from the department, and so on. There is, however, absolutely no sense in
which he accepts the policies and procedures of the police department.
Indeed, he joins the group with the sole aim of undermining the institu
tional structure because of its sexist nature. To say that all of this is still
compatible with the sabotaging joiner accepting this institutional structure
is to render the notion of acceptance here vacuous.
Thus, I reject the status function model of spokespersons—spokespersons
are not like money or royal seals, which require agreement or recognition
by the members of the social transactions in order to be what they are.
9 For detailed discussions about this difference between acceptance and belief, see, for
instance, van Fraassen (1980), Stalnaker (1984), Cohen (1989 and 1992), Wray (2001), and
Hakli (2007 and 2011).
GROUP ASSERTION 145
Instead, my view might be called pluralist: there are many mechanisms for
securing the relevant land of authority needed for being a spokesperson.
One of the more common ways is where there is agreement, and authority is
acquired through members of a group explicitly or implicitly granting it to a
spokesperson. For instance, a philosophy department might vote to elect
the chair as its spokesperson on matters related to job searches, or the mem
bers might grant this authority when they accept employment at an institu
tion where this is part of the chairs duties. Or members of a group might
sign a legal contract that grants authority to a lawyer to speak on their
behalf on matters related to the litigation in question. But the granting of
authority by the members of a group is not the only way in which it might
be acquired.
Another way is through tradition or inheritance, such as when a member
of a monarchy has the authority to speak on behalf of his or her nation on,
say, matters of national security. Even if members of the nation explicitly
reject the monarch’s authority and actively seek to distance themselves from
the expressed views, the authority might exist nonetheless. Moreover, unlike
heads of state who are voted into office, citizens of a monarchy might have
no say in who is speaking on their behalf, and if they acquired their citizen
ship through birthright, there might be no sense in which they ever accepted
the relevant institutional structure.10
Still another way in which such authority might be acquired is through
non-objection. Suppose that a collection of protesters informally gathers
outside the deans office at a university to object to the recent firing of a
tenured faculty member. When the media shows up on the first day,
suppose that one of the protesters—call her Mary—states, “We object to the
faculty member’s employment being terminated without due process.” On
this first day, Mary’s statement is an instance of an individual offering her
own view of what a collective entity believes. In other words, the assertion is
Mary’s, not the group’s. But suppose that the protesters continue to meet
and no one objects to Mary reporting their views to the media. At some
point, Mary acquires the authority to speak on behalf of the group through
10 Ludwig argues that citizenship is a hybrid status, where “operative members” are those
“who have accepted membership” and thus when we say that a hybrid institutional group has
done something qua institution, this “entails that (and only that) its operative members have
all contributed, whether or not it has non-operative members as well” (Ludwig 2014, p. 99).
But why would those who obtained citizenship through birthright not be operative? Doesn’t
this subgroup make up the bulk of most nations? Moreover, since it is highly questionable
whether accepting membership is necessary for group membership, it would be best to not
build this into ones account of group agency.
the absence of objections from the members, thus rendering her statements
those of the protesters.11
It may also be worth leaving open the possibility that having the author
ity to be a spokesperson can be moral or fundamental. Just as I have the
authority to refuse sexual advances, regardless of whether this authority has
ever been recognized or appreciated, perhaps parents have the authority to
assert on behalf of their very young children, even if they live in a society
where this has always been denied.
These are simply some examples of how authority can be acquired, but
there are certainly others, such as through seizure or coercion. The central
point to note here, though, is that the having of authority to be a spokes
person need not be granted or accepted by either the members of the group
or the audience in question, and it is the having of authority to speak on a
group’s behalf that in large part determines whether the assertion in ques
tion is an individual’s or a group’s.12
In addition to being pluralist, the conception of authority operative here
is de facto or descriptive rather than normative, and thus the authority in
question need not be morally or politically legitimate. Consider a case
where the authority in question is acquired in some sense illegitimately:
suppose, for instance, that in the case of the protesters discussed above, the
members do not object to one of them speaking on their behalf because
they are oppressed or bullied by him. Or suppose that some revolutionaries
seize authority from a political figure to speak on behalf of a subset of the
citizenry. Is the relevant group asserting in these sorts of cases?
The short answer to this question is yes. The mere fact that authority is
acquired in, say, a morally illegitimate way does not mean that the person in
question doesn’t have it. As I said above, the authority at issue here is de
facto or descriptive authority, not normative. A group of rebels might seize
authority from the president of a country so as to oppress the members of
an ethnic minority. Even if this seizure of authority is morally illegitimate,
the rebels might still come to have the authority to speak on behalf of the
country. The same is true in the individual case: a highly aggressive business
executive might become the president of a corporation through immoral
11 Ludwig (2014) might deny that this is a case of a spokesperson asserting on behalf of a
group since he claims that only genuine organizations can authorize proxy agents. But this isn’t
plausible. Unstructured, informal groups can evolve to have spokespersons without any clear
act of "joining” or of agreeing to the conditions of membership.
12 I will say what else is needed to distinguish individual from group assertion in what
follows.
GROUP ASSERTION 147
dealings, but this doesn’t prevent him from having the authority to serve as
the corporations spokesperson. Or suppose that a woman feels so dominated
by her husband that she never objects when he speaks on her behalf.
Through this systematic non-objection, the husband might come to have
the authority to be his wife’s spokesperson on a range of issues, even if the
process whereby this is achieved is morally illegitimate. Of course, there are
limits to this, which make clear the difference between authority and power.
Some psychological trauma may be so severe that the absence of objection
is due to the inability to object, and so it might not be possible to acquire
authority through non-objection in such cases, despite having power. But
the central point that I want to emphasize here is that illegitimately acquired
authority can be authority nonetheless.
There are, however, a couple of objections about this notion of authority
that should be considered. First, suppose that a king has been taken to have
the authority to speak on behalf of the citizenry without anyone realizing
that in fact the laws of the monarchy grant this authority to the queen. Who
has been the spokesperson for the nation, the king or the queen?
On my view, this would simply be described as a conflict of authority. The
king has authority to speak on behalf of the citizens through non-objection,
and the queen has authority to speak on their behalf through the law. And
such a conflict would have to be resolved through, for example, negotiation,
in order to determine who has the final authority. But this is unique neither
to my view nor to group assertion. Suppose that unbeknownst to us, my
husband and I each hire a different lawyer to represent me in a suit. The first
lawyer says on my behalf that I want to settle while the second one says on
my behalf that I don’t. Which one is my actual spokesperson? Again, there
is a conflict of authority that needs to be resolved here before it can be
determined what my assertion is.
Second, suppose that a king has the legal authority to speak on behalf of
his citizens, but there is widespread discontent in his nation about the exist
ence of the monarchy. No one acknowledges his authority and no one takes
him to be speaking for the nation. Is he still asserting on their behalf?13
If the king is still regarded as the king, and with this role comes, say,
the legal authority to speak on behalf of the citizens, then, yes, the king
continues to assert on behalf of his nation despite their discontent. It is,
however, possible that the widespread discontent among the citizens brings
13 I am grateful to Michael Bratman for this question.

about social changes that do undermine the king’s having this authority.
Perhaps there is so much opposition that it becomes an open question
whether the nation still has a king, or whether one of the Icing’s roles is to be
the spokesperson for the people. In these cases, it would be indeterminate
whether the king is asserting for the nation. If there is radical social change
and the monarchy is dismantled or the king is stripped of much of his
authority, then he would no longer be asserting on behalf of the nation. It
might also be the case that the citizens forge new groups—such as a revolu
tionary or opposition party. While the king might still reign over the nation
and thereby speak on its behalf, there might be different spokespersons for
these opposition groups.
We have seen, then, that having the authority to be a spokesperson can be
grounded in a multitude of features, where agreement or recognition by the
members of the social transactions is merely one such option. I now want to
turn to another aspect of being a spokesperson that is worth highlighting.
4.3 The Autonomy of Spokespersons
In addition to it being the case that the standard way in which a group
asserts is through an authorized spokesperson(s), another central point that
I wish to emphasize is that most spokespersons have a certain degree of
autonomy or independence. A spokesperson, at the very least, is not merely
a parrot or a mouthpiece with a script, repeating verbatim what she has
been told by the members of the group. But even more strongly, a spokes
person often asserts on behalf of a group without consulting the group or its
members regarding the specific content of the proffered statement. This is at
least in part because spokespersons are frequently required to speak for
their clients “on the spot,” to respond to new questions and concerns by
extrapolating from the information that they already have. Moreover,
spokespersons sometimes have expertise that goes beyond what the repre
sented group and its members have. A lawyer, for instance, need not consult
with her client each time she speaks on its behalf since at least some of what
she states concerns legal matters over which her client might be wholly
ignorant.
Combining the central features of authority-based assertion thus far
highlighted—namely, that the standard way in which a group asserts is
through an authorized spokesperson(s), and that most spokespersons have
a certain degree of autonomy—results in the possibility that a group can
GROUP ASSERTION 149
assert a proposition about which it and its individual members are wholly
unaware. Here is an example:
autonomous spokesperson: Philip Morris hires spokesperson S—who
is not a member of the group—to represent the company’s views to the
public.14 Philip Morris explicitly tells S that the company’s official view is
that smoking is safe, no matter what. At a recent press conference, S, in her
role as the official spokesperson for Philip Morris, is asked whether
smoking causes disease X. No member of Philip Morris has ever heard of
disease X, nor do they have any beliefs about its safety, but S responds on
Philip Morris’s behalf that smoking does not cause disease X.
In autonomous spokesperson, Philip Morris asserts that smoking does
not cause disease X while no member of the company has ever even heard
of disease X. This is because S has the authority to autonomously speak on
behalf of Philip Morris where the safety of smoking is concerned, even
when this goes beyond matters that S has explicitly discussed with Philip
Morris’s members. Any adequate account of group assertion, then, needs to
accommodate this distinctive feature of the way that groups assert.15
4.4 Coordinated and Authority-Based Group Assertion
With these considerations in mind, I propose the following accounts of

coordinated group assertion (CGA) and authority-based group assertion
(ABGA), respectively:
CGA: A group G asserts that p in the coordinated way if and only if the
members of G coordinate individual acts, a}>,. .an, so that they all reasonably
intend to convey thatp together in virtue of these acts.
ABGA: A group G asserts that p in the authority-based way if and only if
thatp belongs to a domain d, and a spokesperson(s) S (i) reasonably intends
14 One might ask the following: if Philip Morris hires an outside spokesperson, S, to repre
sent the company’s view, does this thereby make S a member of the group in question? The
answer here is clearly no. If the Supreme Court hires an outside clerk to assist with legal
research, this does not thereby make the clerk a member of the Supreme Court. If Northwestern
University hires Bulley and Andrews Construction Firm to renovate one of the academic
buildings, this does not make the construction workers members of Northwestern. Bringing a
suit against the firm, for instance, is not to thereby bring suit against Northwestern.
15 While individuals might also grant authority to another to speak on their behalf, such as
when a lawyer represents an individual client, group assertion is distinctive in that this is the
standard way in which groups assert.
to convey the information that p in virtue of the communicable content of

an individual act (or individual acts) of communication,16 (ii) has the
authority to convey the information in d, and (iii) acts in this way in virtue
of S’s authority as a representative of G.17
According to the CGA, coordinated group assertion simply involves indi

vidual acts—such as placing rocks in the sand or words in a document—
that are coordinated, and so there is not much to add to what has been said
about individual acts. I will, therefore, spend the remainder of the chapter
focusing on authority-based assertion. And here there are a number of fea
tures to note.
First, condition (i) of the ABGA is modeled on the account of individual
testimony that I have developed elsewhere.18 In particular, the focus is on
acts of communication so as to allow for assertions that do not involve state
ments, such as nods, pointing, and other gestures. Moreover, to avoid coun
tenancing as assertions acts of communication where the intention is to
convey the information that p in virtue of features about the assertion—
such as my intending to convey the information that I have a soprano voice
by asserting this in a soprano voice19—it is required that the speakers rea
sonably intend to convey the information that p at least in part in virtue of
the act’s communicable content. Still further, the intention in question
16 I should note that in Lackey (2006 and 2008), condition (i) is presented as being both
necessary and sufficient for an individual to testify (or assert). However, to distinguish what a
spokesperson does in testifying or asserting on behalf of someone else, rather than on behalf of
herself, my account of individual testimony (assertion) should explicitly specify this. Thus, it
should read:
S testifies (asserts) that p by making an act of communication a if any only if S rea
sonably intends to convey on behalf of herself the information that p (in part) in
virtue of as communicable content.
I am grateful to Marija Jankovic for a question that led to the inclusion of this note.
17 One might wonder whether there is a third kind of group assertion, what we might call
distributed group assertion. Suppose, for instance, that there are three members of a committee,
each of whom uploads information to an automated system. Ml submits that p, M2 submits
that q, and M3 submits that r. The system then aggregates the information and issues a public
report that the committees view is that s, even though no member of the group is aware of this
aggregated result. Is this group assertion? Strictly speaking, the answer is no, as there is simply
no one who intends to convey the information that s. When we learn that 5 from the automated
output, were learning from the system, not from the group. This is supported by the fact that if
it were the group’s assertion, then the committee could learn from its own assertion. For
instance, when the output that s is issued and the committee learns this by reading the report,
the committee itself could come to learn that s from its own assertion. Given this, distributed
group assertion is assertion in only an extended sense.
18 See Lackey (2006 and 2008).
19 This is a slightly modified example from Audi (1997).
GROUP ASSERTION 151
needs to be a reasonable one. A group does not assert that its name is Philip
Morris—even if it intends to convey this information—through winking at
the public. This is because, in the absence of prior agreement that a certain
sequence of winks will be understood as conveying Philip Morris’s name,
this intention is not a reasonable one.
Second, according to the AB GA, a group can assert thatp even when not
a single member of the group either intends to convey the information that
p or asserts thatp, thereby permitting groups to have autonomous spokes
persons who assert on their behalf “on the spot.” At the same time, the
ABGA does not allow such spokespersons to assert on a groups behalf on
any topic whatsoever. Both of these results follow from condition (ii), which
requires that a spokesperson(s) have the authority to convey some or all of
the propositions in a domain of which thatp is a member. So, for instance, a
spokesperson might have the authority to speak on Philip Morris’s behalf
with respect to matters that concern the safety of smoking, but not about
questions concerning the company’s finances. This enables my view to
deliver the correct verdict that Philip Morris is asserting that smoking does
not cause disease X in autonomous spokesperson.
Moreover, notice that condition (iii) of the ABGA requires that S assert
on G’s behalf in virtue of S’s authority as a representative of G. For instance,
suppose that Philip Morris’s spokesperson tells his wife while on vacation
that the company disregarded valid scientific evidence about the dangers of
smoking. In such a case, he might be personally asserting to his wife about
this fact, but he is not doing so on behalf of Philip Morris. This is because
even if he has the authority to convey this information on behalf of Philip
Morris, he is not doing so in virtue of this authority; instead, he is doing so
in virtue of his role as a husband to his spouse. Condition (iii) thus rules out
such individual assertions from counting as a group assertion, even if one of
the members in fact has the authority to speak on behalf of the group.
In order to better understand both conditions (ii) and (iii) of the ABGA,
I would like to draw an important distinction between what we might call a
rogue spokesperson and a bad spokesperson.
On the one hand, a rogue spokesperson is one who asserts that p on
behalf of G either without having the authority to do so or without doing so
in virtue of this authority. There are at least two different ways in which a
spokesperson can be rogue. First, S might assert that p on behalf of G, where
thatp is not part of the domain in which S has authority to represent G. For
instance, Philip Morris’s spokesperson might assert that the company’s
favorite movie is Citizen Kane or that the company does not support gay
152 THE EPISTEMOLOGY OE GROUPS
marriage, despite having the authority only to speak on behalf of the com
pany when the safety of smoking is at issue. Here, the content of the state
ment in question lies outside of the scope of S’s authority in speaking on
behalf of G and thus condition (ii) of the ABGA fails to be satisfied. Second,
S might assert that p on behalf of G, where S’s asserting that p does not aim
to reflect the view G intends for S to assert on its behalf. For instance, Philip
Morris might have a bumbling spokesperson who aims to be a whistle
blower and expose the company’s deceptive practices, but because of her
bumbling ways, ends up inadvertently asserting precisely what G wishes.20
In such a case, even though the spokesperson might in fact assert that
smoking is safe, and even though this might accurately represent what
Philip Morris wishes S to report on its behalf, S is speaking for herself as a
whistleblower when she makes this assertion, not for the company. Given
this, while S might have the authority to speak on behalf of Philip Morris
when the safety of smoking is concerned, S does not assert that smoking is
safe in virtue of her authority as a representative of G, thereby failing to sat
isfy condition (iii) of the ABGA. Thus, when a rogue spokesperson, S,
asserts that p on behalf of G, the assertion in question is S’s, not Gs, either
because S does not have the authority to assert that p on behalf of G or
because she fails to do so in virtue of her authority as a representative of G.
On the other hand, a bad spokesperson is one who asserts that p on
behalf of G and has the authority to do so, but nonetheless fails through
incompetence or negligence to say what G intends for S to assert on its
behalf. One way this might happen is if the spokesperson is simply very bad
at drawing the relevant inferences that follow from G’s other beliefs. For
instance, when S is asked whether smoking causes disease X, S might
answer affirmatively because S fails to realize that Philip Morris intends for
S to respond negatively to this question on its behalf, even though this is the
obvious inference from all of the company’s other views on the matter.
Another way a spokesperson might be bad is through failing to pay close
enough attention to the details of G’s views. For instance, when S is asked
whether Philip Morris agrees with scientists that smoking causes emphy
sema, S might answer affirmatively because S failed to listen carefully to the
discussions at the company’s board meetings. In the former case, S is an
incompetent spokesperson and in the latter case, S is a negligent spokesperson,
but in both cases S is a bad spokesperson who is asserting on behalf of
20 I am grateful to Anne Baril for this example.

GROUP ASSERTION 153
Philip Morris. This is because S not only has the authority to assert thatp on
behalf of G, but S also does so in virtue of her authority as a representative
of G—she just does so badly. Thus, when a bad spokesperson, S, asserts that
p on behalf of G, the assertion in question is Gs, not S’s.
The difference between a rogue and a bad spokesperson might be made
more vivid by considering the likely consequences of their respective state
ments. While a rogue’s assertion might be disavowed or otherwise denied by
the group in question and the spokesperson might be fired, a bad spokes
person might be forced to retract the assertion on behalf of the group and
be reprimanded or trained. A rough analogy on the individual side might
be the difference between an unfortunate statement offered while under the
control of hypnosis versus one made while drunk: in the former case, one
didn’t assert anything at all, and thus can completely disavow it, while in the
latter case, one did offer an assertion and thus needs to retract it, and per
haps apologize, the next morning.
One might worry here that my view has the unattractive consequence
that a group asserts that p even when every member of the group protests
that the spokesperson in question made a serious mistake in asserting thatp
on their behalf. While this is indeed true of my view when the spokesperson
is merely bad, rather than rogue, I regard this as the correct result. Consider
a spokesperson for an individual: if I hire a sloppy or mediocre attorney to
defend me in a lawsuit, I might end up asserting through the attorney that,
for instance, I’ll accept a settlement offer, despite this not being what I
ultimately wanted. The same is true of action more broadly—if I grant
authority to a financial advisor or a stockbroker to make financial transac
tions on my behalf, I might end up selling one of my stocks despite my
vehement opposition to this after the fact. This is why we should choose our
spokespersons, and our representatives more broadly, very wisely.
Notice that on this view, a rogue spokesperson and a bad spokesperson
might offer assertions with the very same content in identical circum
stances, yet one might be S’s assertion while the other is G’s. SI might assert
that smoking causes emphysema because she aims to be a whistleblower
while S2 might assert that smoking causes emphysema because she fails to
draw obvious inferences from Philip Morris’s other views on the matter.
When SI and S2 both offer their assertions on behalf of the company in
response to the same question at a single press conference, Si’s assertion is
her own while S2’s is Philip Morris’s.
Finally, it should be noted that it is precisely conditions such as (ii) and
(iii) that distinguish an individual asserting about the beliefs of a group from
a group asserting. Suppose, for instance, that a member of Philip Morris,

who has no authority to speak on its behalf, asserts that the groups view is
that smoking is safe. Even if this member has access to what the groups
view is and purports to be speaking on its behalf, this is not group assertion;
instead, it is an individual asserting about the group’s view. According to the
ABGA, this is because the member is not a spokesperson that has the
authority to convey information about the safety of smoking on behalf of
Philip Morris. Of course, as mentioned earlier, in some cases, a member
might try to offer a group assertion and, to the extent that she succeeds, she
might in part create her own authority as the group’s spokesperson. But
until this happens, she is speaking for herself, not the group.
4.5 Two Other Accounts
There are two other views of group assertion in the literature.21 The first is
Miranda Fricker’s variant of a joint acceptance account, according to which
we should “... construe a group testifier as constituted, at least in part, by
way of a joint commitment to trustworthiness as to whether p (or whatever
range of p-like questions might delineate the body’s expertise, formal remit,
or informal range of responsibility)” (Pricker 2012, pp. 271-2, original
emphasis). Fricker here takes the joint commitment to trustworthiness to
be constitutive of a group being a testifier, and thus it seems to follow that
no group could offer assertions in the absence of such a commitment. But
this is a puzzling requirement, for it seems to confuse being a testifier sim-
pliciter with being an epistemically good testifier. Surely an individual can
testify about all sorts of matters without any commitment at all to trust
worthiness; liars and those engaged in other forms of deception do precisely
this. What we would say about them is that they are not epistemically good
testifiers, but that they are testifiers nonetheless. The same is true of groups.
Groups whose members do not jointly commit to trustworthiness—such as
certain deceptive corporations and governments—are testifiers, even
though they are not epistemically reliable ones. Indeed, if we adopt Fricker’s
account of group testimony, not only would we be hard-pressed to account
21 Both of these views are presented as accounts of group testimony, but they can be under
stood as accounts of group assertion for our purposes. I will thus use “testimony” and “asser
tion” interchangeably here.
GROUP ASSERTION 155
for the lies of groups that generally eschew trustworthiness, we would also
thereby have difficulty holding them responsible for such deception.
Fricker appeals to Edward Craigs distinction in his (1990) between being
a testifier or informant and being a source of information and argues that she
is offering an account of only the former. Footprints in the sand, for
instance, might be a source of information, as are photographs, but neither
is a testifier. Persons can function this way, too—I might infer that you are
nervous from the hesitancy with which you deliver your testimony, even
without your asserting that you are nervous. But even granting such a dis
tinction, surely not every epistemically bad testifier turns out to be a mere
source of information. When Philip Morris says that smoking is safe, the
corporation is a group testifier if anything is, despite the fact that it is clearly
an epistemically bad one here. So, my central criticism is unaffected by
Craigs distinction. Later in her paper, Fricker goes on to say, “At any rate,
my main claim can be the weaker one: that any group partly constituted by
way of a joint commitment to trustworthiness (regarding some relevant
range of questions) is pre-eminently suited to enter into the second-personal
relations of trust that characterize testimony” (Fricker 2012, p. 272). Even
this weaker notion, however, is problematic, since, again, second-personal
relations of trust clearly do not characterize testimony simpliciter—at best,
they characterize epistemically good testimony.
The second account of group testimony, or assertion, in the literature is
offered by Deborah Tollefsen, where she argues as follows:
A group G testifies that p by making an act of communication a if and

only if:
1. (In part) in virtue of a’s communicable content G reasonably intends to

convey the information thatp.
2. The information that p is conveyed by either (i) a spokesperson S or
(ii) a written document.
3. If (i), G does not object to S’s uttering p on its behalf and if G intends
for any specific individual(s) to utter p, it intends for S to utter p and S
believes that he or she knows this.
4. If (i), S utters p for the reasons in 3.
5. If (ii), G does not object to the way in which p is conveyed in writing.
6. G conveys the information that p in the right social and normative
context.
7. In conveying the information that p in the right social and normative

context, G is taken to have given its assurance thatp is true. (Tollefsen 2009,
pp. 12-13)
There are issues to be raised with every condition of this account. Let s begin
with (1), which Tollefsen adapts from the account of individual testimony
found in Lackey (2006 and 2008). The problem with applying it here is that
it is not clear that it can accommodate the autonomy of spokespersons and,
therewith, the kind of group testimony found in autonomous spokes
person. In particular, while Philip Morris testifies that smoking does not
cause disease X, not a single member of the group reasonably intends to
convey this information in virtue of S s making the act of communication a.
So, unless a group can intend to do something that no individual member
intends to do, (1) is a problem for group testimony.22
Regarding condition (2), there are three problems. First, a groups con
veying the information that p through either a spokesperson or a written
document is an unnecessary disjunction, as a spokesperson can clearly
communicate on behalf of a group in both verbal and written form. So (i)
subsumes (ii). Second, there can be more than one spokesperson who con
veys the information thatp. A subgroup of individuals, for instance, might
be called upon to communicate a company’s view, and thus all of these
members would function as relevant spokespersons, (i) should, therefore,
be modified accordingly Finally, (2) lacks the resources for accommodating
instances of coordinated group testimony.
Turning to condition (3), worries arise regarding both parts. First, it
necessitates that G does not object to S’s utteringp on its behalf. As I empha
sized in the text, however, spokespersons often have some autonomy with
respect to speaking on their clients’ behalf, and so they do not present their
statements to the group for prior approval before they are offered. (3), then,
cannot require for every instance of group testimony given via a spokes
person that the group does not object to S’s uttering thatp on its behalfprior
to the utterance. But nor can it be necessary that the group does not object
to the spokesperson’s uttering that p on its behalf during or after the utter
ance. If it did, it would make whether a group in fact testified depend on
something that could possibly come years after the statement was offered
22 I should note that I am not saying that a group cannot intend to so something that no
individual member of the group intends to do. But if ones account of group testimony is going
to rely on a thesis this substantive, then it should be defended.
GROUP ASSERTION 157
since a group could object to a spokespersons testimony long after it was

offered. Even more importantly this would permit groups to deny having
testified when clearly they did. Suppose, for instance, that in autono
mous spokesperson, Philip Morris attempts to avoid legal and moral
responsibility for smoking-related health problems by denying having testi
fied to its safety simply because (3) wasn’t satisfied. Not only does it seem
that the company testified despite the failure of this condition, but groups
also shouldn’t be able to get off the normative hook so easily. Since a spokes
person can clearly convey information on behalf of a group in writing, simi
lar problems apply to condition (5).
The second part of (3) requires that, if G intends for any specific
individual(s) to utter p, it intends for S to utter p and S believes that he or
she knows this. But consider this: suppose that Philip Morris has two offi
cial spokespersons, Maria for Mondays, Wednesdays, and Fridays and
Terrence for Tuesdays and Thursdays. Suppose, further, that the group,
knowing that it is Monday, intends for Maria to state on its behalf that
smoking is safe, but it turns out that she called in sick and was replaced at
the last minute by Terrence. When he reports to the public on behalf of
Philip Morris that smoking is safe, this is no less the company’s testimony
than if Maria had done so, despite the fact that the group doesn’t intend for
Terrence to do so.
There is also a problem with (4), which requires that if the information
that p is conveyed via S, S utters that p for the reasons in (3). Recall that the
reasons in (3) are that the group doesn’t object to S’s uttering that p on its
behalf and intends for S to utter that p. Once again, however, this condition
is too strong and ignores the different roles that are often given to spokes
persons. For instance, a spokesperson might be hired specifically to piece
together the information gathered at a group’s meeting into a legally sound
view and then report it to the public. In such a case, the spokesperson utters
that p, not because the group intends for S to do so, but because the group
intends for S to use its autonomy and legal expertise to present the best ver
sion of the group’s view from a legal point of view. It is, however, still the
group’s testimony
Condition (6), which requires that G convey the information thatp in the
right social and normative context, is included to rule out as group testi
mony statements such as those offered by Philip Morris’s spokesperson to
his wife while on vacation. But not only is a clear, substantive characteriza
tion of which social and normative contexts are “right” difficult to come by,
we have seen that the same result can be achieved by requiring that the
spokesperson in question have the authority to speak on behalf of the group

being represented.
Finally, there are problems with condition (7), which requires that G be
taken to have given its assurance that p is true. I have elsewhere argued23
extensively against what is known as the assurance view of individual testi
mony, and my objections apply straightforwardly to the group case. So, I
will briefly mention only the following: such a condition makes the act of
testifying depend on the recipients reception of it, but this has counterintuitive
results. Suppose, for instance, that a corporation is called to testify against a
partner company and the jurors are skeptical of the spokesperson’s trust
worthiness because of the conflict of interest. In such a case, the jurors
might not take the company to have given its assurance that p is true, but
surely it has testified.24
Ihere are, therefore, significant problems facing both of the existing
accounts of group assertion, none of which apply to the view defended in
this chapter.
4.6 Group Assertion Is Not Reducible to

Individual Assertion
If what I have argued is correct, a spokesperson asserting on behalf of a

group in the right sort of way can be constitutive of group assertion, and
thus this phenomenon must be understood in inflationary terms since a
group may assert that p even when no member of the group asserts that p.
In autonomous spokesperson, for example, Philip Morris asserts
that smoking does not cause disease X despite the fact that no member of
Philip Morris asserts this or is even aware that there is such a disease as
23 See Lackey (2008).

24 I should note that Tollefsen’s account of group testimony is adapted from Justin Hughes’s
account of group speech acts, according to which:
For a group, G, speaker, S, and utterance, x, G utters x if and only if:
1.There exists a group, G, this group has an illocutionary intention, and x conveys that
illocutionary intention.
2.S believes that he or she knows the illocutionary intention of G and that X conveys this
illocutionary intention.
3.G does not object to S uttering x on its behalf and if G intends for any specific individual(s)
to utter x, it intends for S to utter x. S believes that he or she knows this.
4.2 and 3 are the reasons S utters x. (Hughes 1984, p. 388)
My arguments here apply, mutatis mutandis, to Hughes’s account.
GROUP ASSERTION 159
X. Given this, the only one who could be doing the asserting here is the
group itself.
But one might wonder how substantive this conclusion is. For even
though group assertion is not reducible to the assertion(s) of individual
members of the group, isn’t it still reducible to individual assertion(s)? In
particular, isn’t the group’s assertion in autonomous spokesperson
reducible to the spokespersons assertion? If so, the mere fact that the
spokesperson is not a member of the group doesn’t seem to reveal anything
deeply important about the nature of group assertion. Indeed, the extent to
which group assertion demands an inflationary treatment seems to be a
minor quibble regarding whether the reductive base needs to be made up of
group members or not. The heart of the view, however, seems clearly
deflationary.
This understanding of the view of group assertion that I’ve defended in
this chapter is, I think, deeply mistaken. In a nutshell, my response to this
worry is this: when spokespersons are speaking on behalf of groups that
they represent, they are not themselves asserting anything at all, a conclu
sion that is clearly supported by noticing that what they say does not have
any of the paradigmatic features of assertion. Let us begin with what is
arguably the most decisive consideration here: assertion is governed by an
epistemic norm, but what spokespersons say is not. For instance, it has been
widely argued that knowledge is the norm of assertion—that one should
assert that p if and only if one knows that p.25 While such a view is not
immune to objections, most of the critics simply replace it with a weaker
epistemic norm, such as justified belief, or reasonable to believe, and so
on.26 But now notice: there is no sense whatsoever in which spokespersons
are governed by such norms. Consider a chair serving as the spokesperson
for her department in a conversation with the administration about future
hiring plans. Under no circumstances should she assert thatp to the admin
istration only if she knows, or justifiedly believes, or has reason to believe
thatp. All focus on belief, either directly or indirectly, and whether the chair
believes something is entirely irrelevant to the norms she should follow as a
spokesperson. More precisely, the dominant norm governing spokespersons
is to assert what best reflects the view of the group she is representing. Because
25 See Unger (1975), Williamson (1996 and 2000), Adler (2002), DeRose (2002), Reynolds
(2002), Hawthorne (2004), and Fricker (2006). Cohen (2004) says that he is “not unsympa
thetic” to the view.
26 See, for instance, Douven (2006), Lackey (2007), and MacKinnon (2013).
a spokesperson can be doing everything that she ought to even if reporting

on behalf of a group a proposition that she personally has absolutely no
basis for, and indeed evidence against, believing, there is simply no epi
stemic norm of assertion governing spokespersons. Since someone is clearly
asserting something in cases such as autonomous spokesperson, and
this asserter is subject to the norm(s) governing assertion, the natural con
clusion to draw is that it is the group itself. Thus, we have a group asserting
that p when no individual at all is asserting thatp.27 Such a view is nowhere
in the ballpark of a deflationary view.
This conclusion is further supported by considering other features of
assertion. In addition to being governed by an epistemic norm, Sanford
Goldberg highlights the following in his recent book on assertion:28
1. Conveyed Self-Representation: “Many writers describe assertions as

involving the speaker’s representing herself as knowing, or at least hav
ing evidence for, what she has asserted.” (Goldberg 2015, p. 7)
2. Sincerity: “Another feature of assertion which, though not unrelated
to assertion’s epistemic significance, nevertheless deserves to be called
out separately, has to do with assertion’s relation to belief. Simply put,
when they are performed sincerely, assertions express or manifest
one’s beliefs.” (Goldberg 2015, p. 8)
3. Entitlements and responsibilities: “Suppose that you believe some
thing on the basis of Jones’ say-so, and then are queried regarding the
grounds of your belief... .we might say that Jones’ assertion author
ized or entitled you to do so; and when you do ‘pass the buck’ to her
in this way, Jones then has the responsibility to address the challenge
herself. It would thus appear that in asserting that p, the speaker
authorizes the hearer to defer any legitimate challenge to the truth of
the claim to her, and generates the responsibility for taking up that
challenge.” (Goldberg 2015, p. 8)
We see, again, that while these features are true of the group’s assertion, they
are not true of the spokesperson’s report. Beginning with 1, when a spokes
person reports that p on behalf of a group, there is no sense in which she
27 Just as an attorney might bring a lawsuit on behalf of her client, without being a party to
the suit herself, so, too, a spokesperson might assert on behalf of another without thereby
asserting herself.
28 Goldberg (2015).
GROUP ASSERTION 161
represents herself as knowing, or having evidence for believing, that p.

Instead, it is the party she is speaking for that is being represented as having
the appropriate epistemic relationship to that p. This is related to 2, since
even if one wishes to reject that there is an epistemic norm governing asser
tion, it surely is true that when assertions are performed sincerely, they gen
erally express or manifest the asserter’s beliefs. Again, however, this is not at
all the case with respect to the spokesperson’s reports, where sincerity would
be wildly out of place. Indeed, a spokesperson who aimed to be sincere, and
report what she herself believed, would be subject to significant criticism
and censure by the party she is representing. Finally, if it is appropriate to
“pass the buck” to anyone in cases of authority-based assertion, surely it
should be passed to the group rather than to the spokesperson. The group is
the one who espouses the view in question, and the group is the one that
bears the responsibility for the assertion—not the group’s messenger.
Indeed, it may even be the case that the spokesperson knows very little
about why the group holds the view that it does, and would rightly need to
check with the party she is representing before responding to any objec
tions. It should thus be clear that spokespersons are not asserting anything
in cases of authority-based group assertion; they are simply the means by
which groups offer assertions.
One issue that I would like to address is this: I argued in the earlier chap
ters of this book on behalf of views of collective phenomena, such as group
belief and justified group belief, that are more deflationary29 than is my
account here of group assertion.30 Yet in this chapter I am defending an
account of group assertion that is robustly inflationary. Is there an explan
ation of this asymmetry that is not ad hoc?
Yes, and here it is: when, and only when, it is possible to grant authority
to another agent or agent-like entity to do something on one’s behalf does it
follow that inflationism is true. So, for instance, I can give authority to my
lawyer to speak on my behalf, to lie on my behalf, to bullshit on my behalf,
and to act on my behalf. In all of these cases, then, it will be possible for my
actions to be constituted by the actions of another, even when I myself am
entirely ignorant of the matter. Thus, accounts of all of these phenomena
will be inflationary in nature. In contrast, I cannot grant authority to
29 I should note that I have never defended a view that is entirely deflationary. Rather I have
argued for views that have as a condition that some of the individual members of the group
instantiate the phenomenon in question.
30 See Lackey (2016) and Chapters 1, 2, and 3 of this book.
another to believe on my behalf, or to desire on my behalf, or to justifiedly

believe on my behalf, or to know on my behalf. To be sure, I can defer to
others in such cases. When asked, “Where do you want to have dinner,” I
can respond by saying, “Wherever my daughter wants to go” What this
means is that I’m giving authority to my daughter to decide where we’re
going to eat our next meal. I’m deferring to her desires, but nothing she
does is constitutive of my mental states.
It may now be asked, however, whether there is tension in my overall
view. For, on the one hand, I am saying that states such as group belief or
group knowledge require member belief or member knowledge, and yet, on
the other hand, I am also saying that group assertion does not even require
by a single member of the group in question awareness of the proposition
asserted. Thus, a group can assert thatp in the complete absence of belief or
knowledge that p. But then isn’t there a conflict between my appealing to,
say, an epistemic norm of assertion—which has some connection to belief—
to motivate the extent to which I’m an inflationary theorist, while also
denying that group belief is necessary for group assertion?
By way of response, let me first emphasize that I am here providing an
account of group assertion, not epistemically permissible assertion. Given
this, many of my examples, such as autonomous spokesperson, are
cases of a group offering an assertion, though not necessarily in an epistem
ically appropriate fashion. For instance, while it is true that Philip Morris
asserts that smoking does not cause disease X, it is clearly not epistemically
proper to do so, as there is no basis at all for believing such a claim.
Nevertheless, given my account of authority-based group assertion,
surely it is plausible to think that there will be some group assertions on my
view that are epistemically permissible despite the fact that the group itself
fails to possess the knowledge in question. But I don’t regard this as prob
lematic, as I have elsewhere argued extensively that assertion at the individ
ual level can be epistemically proper in the absence of knowledge, and even
in the absence of belief. This leads to my embracing what I call a Reasonable
to Believe Norm of assertion.31 So it would simply follow that group asser
tion is like individual assertion in requiring neither knowledge nor belief;
indeed, the considerations in this paper can be viewed as providing even
further arguments against views such as the Knowledge Norm of assertion.
Of course, given the combination of theses I hold, it would have to be shown

GROUP ASSERTION 163
either (i) that groups properly assert that p when, and only when, it is
reasonable for groups to believe that p, or (ii) that the norms governing
assertion differ at the individual and at the group levels. I favor option (i),
but arguing in favor of it lies beyond the scope of this chapter. The point
that I wish to emphasize here is that my inflationary view of group assertion
is compatible with my appealing to general features of assertion to support
such a view.
It is also worth noting that my view of the nature of group assertion has
significant consequences for the epistemology of group testimony. In par
ticular, if, as I have argued, the statement in autonomous spokes
person is an instance of a group asserting, then a widely accepted view in
the epistemology of individual testimony—the transmission view32—cannot
be true of group testimony. According to the transmission view, knowledge
is transmitted via testimony and thus if H knows that p on the basis of S’s
testimony that p, then S must know that p. But now consider a modified
version of autonomous spokesperson: suppose that it is true that
smoking does not cause disease X and the public comes to learn this on the
basis of Philip Morris’s spokesperson stating that this is so. In such a case,
the public knows that smoking does not cause disease X on the basis of
Philip Morris’s testimony, but there is no sense whatsoever in which Philip
Morris knows that smoking does not cause disease X, as Philip Morris
doesn’t even have the concept of disease X. So even if the transmission view
is true of individual testimony, it cannot apply at the level of groups.
4.7 Conclusion
In this chapter, I’ve provided the framework for an account of group asser
tion. On my view, there are two kinds of group assertion, coordinated and
authority-based, with authority-based group assertion being the core
notion. I’ve argued against a deflationary view, according to which a group’s
asserting is understood in terms of individual assertions, by showing that a
group can assert a proposition even when no individual does. I’ve also
argued on behalf of an inflationary view, according to which it is the group
32 Proponents of different versions of the transmission view include Welbourne

(1979, 1981, 1986, and 1994), Hardwig (1985 and 1991), Ross (1986), Burge (1993 and 1997),
Plantinga (1993), McDowell (1994), Williamson (1996 and 2000), Audi (1997,1998, and 2006),
Owens (2000 and 2006), Reynolds (2002), Faulkner (2006), and Schmitt (2006). For objections
to this view, see Lackey (2008).
itself that asserts, a conclusion supported by the fact that paradigmatic

features of assertion apply only at the level of the group. A central virtue of
my account is that it appreciates the important relationship that exists
between most groups and their spokespersons, as well as the consequences
that follow from this relationship. My view, thus, provides the framework
for distinguishing when responsibility for an assertion lies at the collective
level, and when it should be shouldered by an individual simply speaking
for herself.
Group Lies
We often talk about groups lying. For instance, a Reuters headline regarding
a lawsuit brought against BP by the U.S. government claims, “BP lied about
[the] size of U.S. Gulf oil spill, lawyers tell trial.”1 In particular, the plaintiffs
argue that immediately after the 2010 spill, internal company e-mails reveal
that BP publicly reported that only 5000 barrels of oil were leaking into the
ocean per day with the deliberate intention to be deceptive, even though the
company believed that this report was false and, in fact, knew that up to
100,000 barrels per day could have been leaking. This case is not unusual: a
cursory review of recent news pulls up stories about the lies of Facebook,
Google, the Trump Administration, and various drug companies.
Moreover, there are often enormously significant consequences that
follow from group lies, for both the bars and those to whom they lied. If BP
lied about how many barrels of oil were leaking into the ocean, this could be
the difference between its being fined $17.6 billion instead of $4.5 billion. If
the Trump Administration lied about Russian interference in the 2016
election, then this not only warrants the impeachment of Donald Trump,
but also calls into question the legitimacy of our democratic processes. If a
pharmaceutical company lies about the potentially harmful side effects of a
highly lucrative drug to treat cancer, then this could result in its bearing
responsibility for the health problems and death of countless patients.
Despite the prevalence of group lies and their often far-reaching effects,
there has never before been a philosophical treatment of group lies.2 This
chapter begins the process of filling this surprising gap in the literature by
focusing on the question of what a group lie is. After providing an account
of how to understand individual lies, I will consider, first, whether group
lies can be understood in terms of the lies of the groups members and, sec
ond, whether group lies can be characterized in terms of joint agreement by
1 http://www.reuters.com/article/2013/09/30/us-bp-trial-idUSBRE98T13U20130930,
accessed July 20, 2019.
2 Lackey (2018b) is the only exception.
The Epistemology of Groups. Jennifer Lackey, Oxford University Press (2021), © Jennifer Lackey.
DOI: 10.1093/OSO/9780199656608.003.0006
the group’s members to lie. After showing both views to be misguided, I

offer my own account of group lying, according to which it crucially
involves the group offering a statement. In particular, because what a group
says can come apart from what its individual members say, I argue that a
group might lie when no individual member lies, and a group might fail to
lie even though every individual member does. Thus, my view provides a
framework for not only understanding what a group lie is, but also for hold
ing groups responsible for their broader linguistic behavior.
5.1 Individual Lies
A natural place to turn in trying to understand the nature of a group lie is

with accounts of what is involved in an individual lying. The traditional
view of lying, with roots dating back at least to the work of Augustine in De
mendacio, holds that this phenomenon involves two central components:
stating what one does not believe oneself and doing so with the intention to
deceive. More precisely:
LIE-T: A lies to B if and only if (1) A states thatp to B, (2) A believes thatp
is false, and (3) A intends to deceive B by stating thatp.3
LIE-T remained the generally accepted view of the nature of lying until
somewhat recently, with condition (3) coming under repeated attack. The
form of this challenge has been to produce clear instances of lying where
there is no intention on the part of the speaker to deceive the hearer, thereby
showing that (3) is not a necessary condition for lying. To this end, there are
three central kinds of lies that are used as counterexamples: (i) bald-faced
lies, (ii) knowledge-lies, and (iii) coercion-lies.4 Each of (i) through (iii) is
taken to show quite decisively that the traditional conception of lying has
been radically misguided. In particular, it is concluded not only that lying
does not require the intention to deceive, but also that deception is not at all
a part of what it is to lie.5 Thus, LIE-T has been replaced with a variety of
3 Proponents of various versions of the traditional view include Isenberg (1964), Chisholm
and Feehan (1977), Williams (2002), and Mahon (2008).
4 See Sorensen (2007), Fallis (2009), and Carson (2010).
5 Such a move is made explicitly by Fallis: shortly after presenting counterexamples to the
condition that lying requires the intention to deceive, he concludes, “These cases show that
lying is not always about deception” (Fallis 2009, p. 43).
GROUP LIES 167
competing accounts, none of which even makes mention of deception. If

correct, this radical shift in our conception of lying has significant implica
tions beyond the obvious ones involved in understanding the nature of this
phenomenon. For one natural criticism that we might have of the liar is that
she is engaged in intentional deception, where such deceit carries the weight
of the prima facie moral wrongness of such acts. Divorcing lying from
deception, however, also divorces it from this explanation of its prima facie
moral wrongness.
My own view is that the tides have turned too quickly in the literature on
lying. For while it is indeed true that (i)-(iii) are lies and that there is no
intention on the part of the speaker to deceive the hearer in such cases, this
does not warrant severing the connection between lying and deception
altogether. Thus, I replace LIE-T with the following:
LIE-L: A lies to B if and only if (1) A states that p to B,6 (2) A believes that
p is false, and (3) A intends to be deceptive to B in stating thatp.
I will show in what follows not only that LIE-L can capture all three of the
kinds of lies that LIE-T cannot (i.e., (i)-(iii)), but also that non-deception
accounts of lying wrongly count as lies classic cases of what I have elsewhere
called selfless assertions. This reveals that, contrary to the currently wide
spread approach in philosophy, lying is indeed tied to deception as a matter
of necessity.
5.2 Counterexamples to the Traditional View of Lying
Let’s begin with the first land of counterexample to LIE-T: bald-faced lies.
A bald-faced He is an undisguised He,7 one where a speaker states thatp where
she believes that p is false and it is common knowledge that what is being
stated does not reflect what the speaker actually believes. For instance, sup
pose that a student is caught flagrantly cheating on an exam for the fourth
time this term, all of the conclusive evidence for which is passed on to the
Dean of Academic Affairs. Both the student and the Dean know that he
cheated on the exam, and they each know that the other knows this, but
the student is also aware of the fact that the Dean punishes students for
6 For an analysis of what it means for one to state that p, see Chisholm and Feehan (1977).
7 Quoted from The American Heritage Dictionary in Sorensen (2010).
academic dishonesty only when there is a confession. Given this, when the
student is called to the Dean’s office, he states, “I did not cheat on the exam.”8
This is a classic bald-faced lie: the speaker states a proposition that he
believes is false and both the speaker and the hearer know that this is the
case and know that the other knows this. There is, then, no intention on the
part of the speaker to deceive the hearer. In particular, in stating that he
did not cheat on the exam, the student does not intend to bring about any
false beliefs in the Dean, either about his cheating on the exam or about
his beliefs regarding this event. Indeed, he may even wish for the Dean to
believe that he did cheat on the exam, just to relish in the Dean’s spineless
ness. Nonetheless, the student is clearly lying. This shows that condition (3)
of LIE-T is false.9
The second kind of counterexample to LIE-T involves what Sorensen
(2010) calls “knowledge-lies.” “An assertion thatp is a knowledge-lie exactly
if intended to prevent the addressee from knowing that p is untrue but is
not intended to deceive the addressee into believing [that] p” (Sorensen
2010, p. 610). For instance:
In Spartacus (Universal Pictures, 1960), the victorious Roman general,

Marcus Licinius Crassus, asks the recaptured slaves to identify Spartacus
in exchange for leniency. Spartacus... rises to spare his comrades crucifix
ion. However, the slave on his right, Antoninus, springs to his feet and
declares, “I am Spartacus!” Then the slave on Spartacus’ left also stands
and declares “I am Spartacus!”, then another slave, and another until the
whole army of slaves is on their feet shouting, “I am Spartacus!”
(Sorensen 2010, p. 608)
Each slave in this case is offering a knowledge-lie; however, with the excep
tion of Antoninus, none intends to deceive Crassus into believing that he is
actually Spartacus. For once the second slave claims this identity, it is clear
that he is instead aiming to prevent Crassus from learning who Spartacus is.
Given that each slave seems to be lying, condition (3) of LIE-T is again
shown to be false.
8 This is a slightly modified version of an example found in Carson (2010).

9 Though he calls them “cynical assertions,” Kenyon (2003) also discusses bald-faced lies.
However, because he assumes the truth of LIE-T, he concludes that such assertions are not lies.
This seems problematic, not only because bald-faced lies are called lies in our ordinary talk,
but also because our corresponding actions support this talk, e.g., we would charge one with
perjury for offering a bald-faced lie on the stand, we would regard someone as a liar who
repeatedly made such assertions, and so on.
GROUP LIES 169
The third kind of counterexample to LIE-T involves what we might call

coercion-lies. A coercion-lie occurs when a speaker believes that p is false,
states that p, and does so, not with the intention to deceive, but because she
is coerced or frightened into doing so. For instance, suppose that an inno
cent bystander witnesses the murder of a gang member by someone from a
rival gang, but is threatened with death if she testifies against the murderer.
Because of this, the bystander states on the stand at trial, “I did not witness
the defendant murder the victim in question.”10 Here the intention of the
bystander is not to deceive the court into believing that she did not witness
the murder; instead, her aim is to avoid retaliation from the defendant’s fel
low gang members. Indeed, she may even desperately wish for the court to
believe that she did witness the crime. That the court ends up being deceived
by her statement is simply an unintended consequence of the action needed
to achieve the aim of self-preservation. Despite this, the bystander clearly
lies on the stand, evidenced at least in part by the fact that she could be
found guilty of perjury. The intention to deceive is again shown not to be
necessary for lying.
5.3 Non-Deception Accounts of Lying
The combination of these three types of counterexamples provides a for

midable challenge to the traditional view of lying and has prompted a flurry
of alternative views. The three most prominent ones, offered by Fallis
(2009), Carson (2010), and Sorensen (2007), respectively, are:
LIE-F: A lies to B if and only if (1) A states that p to B, (2) A believes that p
is false, and (3)
A believes that she makes this statement in a context where the following
norm of conversation is in effect: Do not make statements that you believe to
be false. (Fallis 2009, p. 34)
LIE-C: A lies to B if and only if (1) A states thatp to B, (2) A believes thatp is
false or probably false (or, alternatively, A does not believe that p is true),
and (3) A intends to warrant the truth of thatp to B. (Carson 2010, p. 37)11
10 This is a modified version of a case found in Carson (2010).

11 Carson also includes a condition requiring the actual falsity of the proposition that is
being stated. But to my mind, the necessity of this condition is decisively refuted by a case that
he himself discusses (Carson 2010, p. 16). On this point, then, I am in agreement with
Augustine when he writes, “...a person is to be judged as lying or not lying according to the
LIE-S: A lies to B if and only if (1) A asserts that p to B, and (2) A does
not believe that p. (Sorensen 2007, p. 256)
LIE-F, LIE-C, and LIE-S are virtually identical in the first two conditions,
with the latter accounts simply allowing classic cases of bullshit, where the
speaker does not believe that p is false but also does not believe that p is
true, to count as lies.12 They also all share the common feature of completely
divorcing lies from deception. But merely stating what one believes to be
false is not sufficient for lying, since speakers frequently say what they
believe is false when being ironic, joking, or reciting lines in a play and yet
are not lying when they do so. For this reason, each proposal adds a further
component to capture only those statements that are genuine lies. Fallis
does so through requiring that the speaker believes that she is offering her
statement in a context where the following norm of conversation is in effect:
Do not make statements that you believe to be false. Since such a conversa
tional norm is not believed to be in effect by a speaker who is being ironic,
humorous, or acting, LIE-F successfully rules them out as instances of lying.
So, too, do LIE-C and LIE-S, the former because Carson understands
intending to warrant the truth of a proposition as being a promise or guar
antee that what one says is true13 and the latter because speakers typically
do not offer flat-out assertions in cases of irony, jokes, and acting.14
Given that all three accounts divorce lying from the intention to deceive,
they can also capture bald-faced lies, knowledge-lies, and coercion-lies,
unlike LIE-T. When denying cheating to the Dean, the student certainly
does not believe that the context is an ironic or humorous one, so he believes
that the conversational norm is in effect: Do not make statements that you
intention of his own mind, not according to the truth or falsity of the matter itself” (Augustine
1952 [395], p. 55). Broncano-Berrocal (2013) objects to my account of lying on the grounds
that I fail to acknowledge that lies are necessarily false but, as should be clear, I regard this as a
virtue of, rather than a problem with, my view.
12 See Frankfurt (2005) for a discussion of bullshit.

13 According to Carson, one warrants the truth of a statement when one makes a statement
in a context where “one promises or guarantees, either explicitly or implicitly, that what one
says is true” (Carson 2010, p. 26). Moreover, whether one warrants the truth of a statement is
independent of what one intends or believes.
14 According to Sorensen, the only condition on assertion is that it must have “narrow
plausibility,” where this is understood as follows: “someone who only had access to the asser
tion might believe it.” “Wide plausibility,” in contrast, is "credibility relative to one’s total evi
dence” (Sorensen 2007, p. 255). Moreover, “[m]uch of what we say does not constitute
assertion. We signal a lack of assertive force by clear falsity (as with metaphor) or by implausi
bility” (Sorensen 2007, p. 256).
GROUP LIES 171
believe to be false; he intends to warrant the truth of what he says since he

wishes to go on the record as denying having cheated on the exam; and he
offers a flat-out assertion. The same is true of both the slaves claiming to be
Spartacus and the innocent bystander denying having witnessed the mur
der in question. Neither the slaves nor the bystander thinks that there is
anything about the contexts in which their statements are being offered that
would prevent this usual conversational norm from being in effect; they all
are inviting their hearers to trust them, even if the invitations are empty;
and their statements are flat-out assertions.
5.4 Back to Deception
Despite the virtues of these non-deception views of lying, I will now argue
that they are misguided, as the divorce between lies and deception is an
unhappy one.
The first point to notice is that there is a range of ways of being deceptive.
Perhaps the most obvious is the one that is the focus of proponents of non
deception accounts of lying, where the aim is to bring about false beliefs in
the victim of the deceit. But another, less explicit, form of deception is
where the aim is to conceal information. According to the Oxford English
Dictionary) deceit is “the action or practice of deceiving someone by con
cealing or misrepresenting the truth.” And Carson, despite endorsing a
non-deception account of lying, claims that “[t]o conceal information is to
do things to hide information from someone—to prevent someone from
discovering it. Often, concealing information constitutes deception or
attempted deception” (Carson 2010, p. 57).15 Given this, I propose the fol
lowing distinction, which will suffice for our purposes even if it does not
fully capture all of the ways of being deceptive:
Deceit: A deceives B with respect to whether p if and only if A aims to

bring about a false belief in B regarding whether p.
Deception: A is deceptive to B with respect to whether p if A aims to
conceal information from B regarding whether p.
15 Carson (2010) is interested in both lying and deception, but is clear that he regards the
latter as not necessary for the former.
Concealing information regarding whether p can be understood broadly

here, so that it subsumes, among other phenomena, concealing evidence
regarding whether p. Moreover, notice that concealing information is
importantly different from withholding information. To withhold informa
tion is to fail to provide it, rather than to hide or keep it secret. If I am trying
to find a home for my challenging puppy, I withhold information about her
lack of being housebroken if you don’t ask me anything about it and I do not
mention it. But if I frantically discard all of the training pads lying through
out my house before you come over, then I am concealing the information
that she is not trained.16 Finally, notice that concealing information is suffi
cient, though not necessary, for being deceptive; thus, it is merely one
instance of a more general phenomenon. Obviously, another way of being
deceptive is to be deceitful, where one’s aim is to bring about a false belief in
one’s hearer.
With this distinction in mind, let us return to the three counterexamples
to LIE-T. In the case of the student’s bald-faced lie to the Dean, while he
does not intend to deceive the Dean into falsely believing that he did not
cheat, he does intend to conceal crucial evidence from the Dean that is
needed for punishment from the university—namely, an admission of
wrongdoing. Without the evidence that would be provided by a confession,
the Dean is paralyzed to take action against the student, and so conceal
ment of the student’s knowledge of what actually happened is the central
aim of his statement.17 According to our distinction above, then, the student
does not intend to deceive the Dean, but he does intend to be decep
tive to him.
In the case of the slaves’ knowledge-lies to Crassus, while it is clear that
there is no intention to deceive Crassus into believing that they are all
Spartacus, there is the intention to conceal the true identity of Spartacus.
Each slave, besides Antoninus, aims to conceal the information that that
person—that is, Spartacus—really is Spartacus and, in so doing, intends to
be deceptive without deceiving.
16 For further discussion of the distinction between withholding and concealing informa
tion, see Carson (2010,pp. 56-7).
17 Fallis (2014) objects to my treatment of bald-faced lies as follows: “it is clear that the stu
dent does not aim to conceal his confession. There is no confession to be concealed since he has
not confessed” (p. 90). My claim is not that the student is concealing his confession, but that he
is concealing evidence which would be conveyed through his confession—namely, his know
ledge or memory of what actually took place.
GROUP LIES 173
Finally, in the case of the bystander’s coercion-lie to the court, while she
does not intend for the court to believe that she did not witness the defend
ant murder the victim in question, she does aim to conceal the eyewitness
testimony that can be used for a conviction. Otherwise put, the bystander is
not aiming to prevent the court from convicting the defendant, but she is
aiming to prevent the court from convicting the defendant on the basis of
her testimony. Again, there is deception without the intention to deceive.
It is worth pointing out that one can be deceptive in the relevant sense,
even if the information that one is aiming to conceal is common knowledge.
In other words, ignorance of that which is being concealed is not necessary
in order to be the victim of deception. To see this, notice that deception
requires that A aims to conceal information from B, and A can certainly
aim to do this even if A is ultimately, perhaps even inevitably, unsuccessful
in achieving this. I can aim to win a marathon even if I know that I will
ultimately fail to achieve this goal. In this sense, even if “conceal” is a suc
cess term—that is, if A conceals x from B, A succeeds in hiding x from
B—“aiming to conceal” is surely not—that is, A can aim to conceal x from B
even if A fails to succeed in hiding x from B. It is, therefore, not available to
the proponent of a non-deception view of lying to reject my analyses of the
above cases by arguing that, because there is common knowledge of that
which is being lied about,18 there is no concealment of information and,
accordingly, no deception.
Thus, none of the counterexamples facing LIE-T succeeds in showing
that the broader notion of deception is not necessary for lying.19 Moreover,
that there is such a necessary relationship between lying and deception can
be supported by considering a case of what I have elsewhere called selfless
18 It should be noted that there is the relevant common knowledge only in the case of the
bald-faced lie and the coercion lie. In the knowledge-lie, Crassus obviously does not know the
identity of Spartacus.
19 Staffel (2011) challenges that Sorensens knowledge-lies are counterexamples to LIE-T by
claiming that he assumes that deception occurs only when someone is brought to flat-out
believe a false proposition. She argues, however, that “[t]his notion of deception is implausibly
narrow, because it overlooks the possibility of deceiving someone by merely making her more
confident in a falsehood” (Staffel 2011, p. 301). While I agree with Staffel both that the concep
tion of deception that is assumed in the arguments against LIE-T is too narrow and that a
speaker can deceive a hearer by making her more confident in a falsehood, I am obviously
interested in a different notion of deception in this paper. Also, Staffel grants that there are
“atypical” cases in which knowledge-lies fail to deceive, but this is not true when my broader
notion of deception is at work. It should also be noted that this point—that one can deceive
another by making her more confident in a false belief—was already made by Krishna (1961)
and Chisholm and Feehan (1977). (Thanks to Don Fallis for the references.)
assertion20 There are three central components to this phenomenon: first a

subject, for purely non-epistemic reasons, does not believe that p; second,
despite this lack of belief, the subject is aware thatp is very well supported
by all of the available evidence; and, third, because of this, the subject asserts
thatp without believing that p. Here is an instance of selfless assertion:
creationist teacher: Stella is a devoutly Christian fourth-grade

teacher, and her religious beliefs are grounded in a personal relationship
with God that she takes herself to have had since she was a very young
child. This relationship grounds her belief in the truth of creationism and,
accordingly, a belief in the falsity of evolutionary theory. Despite this, Stella
fully recognizes that there is an overwhelming amount of scientific evidence
against both of these beliefs. Indeed, she readily admits that she is not bas
ing her own commitment to creationism on evidence at all but, rather, on
the personal faith that she has in an all-powerful creator. Because of this,
Stella thinks that her religious beliefs are irrelevant to her duties as a
teacher; accordingly, she regards her obligation as a teacher to include pre
senting material that is best supported by the available evidence, which
clearly includes the truth of evolutionary theory. As a result, while present
ing her biology lesson today, Stella asserts to her students, “Modern day
Homo sapiens evolved from Homo erectus? though she herself does not
believe this proposition.21
Despite the fact that Stella’s statement that Homo sapiens evolved from
Homo erectus satisfies all three non-deception accounts of lying above,
Stella does not seem to be lying to her students. Why not? My answer to this
question is that it is precisely because Stella does not intend to be deceptive
to her students.22

21 An article in The New York Times (February 12, 2007, “Believing Scripture but Playing by
Science’s Rules”) about Dr. Marcus R. Ross, a creationist who also holds a geosciences Ph.D. in
paleontology, makes clear that the situation described in creationist teacher is by no
means merely a thought experiment. As the author of the article writes, “For him, Dr. Ross
said, the methods and theories of paleontology are one ‘paradigm’ for studying the past, and
Scripture is another. In the paleontological paradigm, he said, the dates in his dissertation are
entirely appropriate. The fact that as a young earth creationist he has a different view just
means, he said, ‘that I am separating the different paradigms.’” (I am grateful to Cristina Lafbnt
for bringing this article to my attention.)
22 For other cases of selfless assertion that pose a problem for LIE-F, LIE-C, and LIE-S, see
Lackey (2007).
GROUP LIES 175
To see this, the first point to notice is that Stellas statement that Homo
sapiens evolved from Homo erectus clearly satisfies the conditions put forth
in LIE-F, LIE-C, and LIE-S. She offers this statement to her students, where
she herself believes that it is false. Moreover, since she clearly does not
regard the context of her classroom as an ironic, humorous, or theatrical
one, she does so while believing that the following norm of conversation is
in effect: Do not make statements that you believe to be false. The reason that
she violates this norm is that she believes it is overridden or defeated by the
duty to state what the scientific evidence best supports when teaching her
biology lesson.23 Stella also intends to warrant the truth of the proposition
that Homo sapiens evolved from Homo erectus since she is promising her
students that what she says is true, just as she does when she states what she
herself believes.24 And, finally, there is nothing about her statement or the
context that prevents her statement from qualifying as an assertion.
The second point to notice is that Stella does not in any way aim to be
deceptive to her students in stating that Homo sapiens evolved from Homo
erectus. For though she does not herself believe this, she regards her own
personal beliefs regarding religion—particularly those that are grounded in
her relationship with God—as irrelevant to the information she conveys
during her biology lesson. Reporting to her students what her religious
beliefs are about the origin of humans would, for Stella, be comparable to
sharing with them what her favorite aspect of evolutionary theory is. Both
are irrelevant to her biology lesson. Given this, when Stella states to her stu
dents a proposition that she believes is false, her aim is not to bring about a
false belief in her students or to conceal her own beliefs on the matter. In
fact, we can imagine that she would willingly share her own views about
evolutionary theory with her students, were they to ask her. Instead, Stella’s
aim is to convey to her students the theories that are best supported by the
current scientific evidence, which include evolutionary theory but not
creationism.
23 In Fallis (2009, pp. 51-3), he discusses at length how this conversational norm can be
overridden or defeated. This case can also be understood as Stella choosing to violate Grice’s
first norm of quality—Do not make statements that you believe to be false—in order to obey
his second norm of quality—Do not say that for which you lack adequate evidence. For more
on this, see Grice (1989) and note 25 below.
24 It is important to keep in mind that Carson believes that one can intend to warrant the
truth of a proposition even when one is lying, and thus one can promise one’s hearer that what
one says is true, even when one knows that it is false. It is comparable to making a promise that
one knows one cannot keep.
The final point to make is that Stella is not lying to her students. Beyond
the intuitiveness of this conclusion, it can be further supported by consider
ing a slightly modified version of creationist teacher. Suppose that
everything about the case remains the same, except that Stella states to her
students that Homo sapiens evolved from Homo erectus, not because she
regards her religious beliefs on the matter as irrelevant to her biology les
son, but because she will get fired from her teaching job if she reveals such
beliefs to her students. In such a case, the aim of Stella reporting what she
herself does not believe is to conceal her own religiously grounded beliefs
on the topic, and thus she intends to be deceptive to her students.
Corresponding to this, Stella’s statement also seems to be a lie.25
We have seen, then, that LIE-F, LIE-C, and LIE-S all count as lies asser
tions that are clearly not, thereby showing that such non-deception accounts
of lying fail to provide sufficient conditions for lying. Should we then simply
add an intention-to-be-deceptive requirement to these accounts? No, since
the alternative requirements found in these views (i.e., conditions (3) in
LIE-F and LIE-C) also fail to provide necessary conditions for lying.26 In
particular, such accounts fail to count as lies assertions that clearly are. To
see this, consider first the case below:
deceptive anthropologist: Shawn is an anthropologist who visits

a highly isolated tribal community living in the Amazon rainforest. He does
not have any beliefs at all about the norms governing the conversations in
their interactions. Nevertheless, he wishes to gain their trust quickly, and so
25 Interestingly, Fallis considers a version of my creationist teacher (which was used

for a different purpose in the paper that he cites), but does not seem to recognize the full force
of the case. He writes:
Norms of conversation can clash’ with each other as well as with other interests that
we have .... For example when a teacher who believes in creationism has to give a
lesson on evolution, Grice’s first maxim of quality comes into conflict with Grice’s
second maxim of quality. If the teacher violates the norms against saying what she
believes to be false solely in order to obey the norm against saying that for which
she lacks adequate evidence, some... might want to say that she is not lying. In
order to accommodate that intuition, my definition might be modified to include
an exemption for such cases. (Fallis 2009, p. 52, note 74)
It is unclear what sort of “exemption” could be added to Fallis’s account of lying to respond
to this counterexample that would not simply be ad hoc. Moreover, by regarding this case as
requiring only such a slight modification, it seems to have blinded Fallis to the far deeper point
that there is a necessary connection between lying and deception. The positive view that I
defend respects this point, provides a unified account of lying, and does not need to resort to
such ad hoc moves.
26 I here set aside LIE-S since it does not provide a third condition that is intended to be a
substitute for the intention-to-deceive requirement.
GROUP LIES 177
he states to them, “My grandmother was an anthropologist who lived with

members of your tribe decades ago, and so I feel as though I already know
you.” Not only does Shawn believe that this is false, he also states this with
the intention to deceive the tribe members into believing that he has a per
sonal connection with their ancestors.
Shawn does not satisfy condition (3) of LIE-F since he fails to believe that
the following conversational norm is in effect: Do not make statements that
you believe to be false. But surely this lack of belief does not prevent him
from lying. For not only does Shawn state what he believes to be false, he
does so with the explicit aim to deceive the tribe members into believing
this falsehood. That Shawns statement is a lie is further supported by noting
that a committee investigating a supposed ethics violation involving his
research methods would hardly think the matter resolved when he says,
“Look, I didn’t lie to the tribe members since I had no idea what conversa
tional norms were operative in their community.”
Consider, now, the following case:
sabotaging friend: Fran wants to sabotage the relationship between

Sam and Betty, but does not want to be held responsible for their break-up.
So, she tells Sam, “Betty is cheating on you, but don’t take my word for it.”
Fran not only believes that it is false that Betty is being unfaithful, she also
offers her assertion with the deliberate intention to deceive Sam.
Fran does not satisfy condition (3) of LIE-C since she does not intend to
warrant the truth of the proposition that Betty is cheating on Sam.
Specifically, because she does not want to shoulder the responsibility of
Betty and Sam breaking up, she makes it explicit that she is not promis
ing or guaranteeing that what she says is true. Despite this, Fran is clearly
lying to Sam, a verdict supported by the similarity the situation bears to
the paradigmatic case of Iago lying to Othello about Desdemona’s
fidelity.27
27 Fallis (2009) presents a counterexample to Carson’s view where a witness to a murder

follows up his statement that “Tony was with me at the time of the murder” by saying, “Of
course, you know I am really bad with dates and times” (p. 49). Carson responds to this case as
follows: “If the proviso ‘you know that I am bad with dates’ is intended to weaken, but not
remove, the assurance of truth, then my... definition counts this statement as a case of lying.
On the other hand, if the proviso is intended to completely remove or nullify any assurance of
truth then the statement is not a lie” (Carson 2010, pp. 38-9). Even if this response works with
respect to Fallis’s case, it does not seem at all plausible with respect to sabotaging friend.
Given this, I propose the following account of lying:
LIE-L: A lies to B if and only if (1) A states thatp to B, (2) A believes thatp
is false, and (3) A intends to be deceptive to B in stating thatp.
LIE-L avoids all of the problems afflicting rival views. It delivers the correct
result in the three kinds of counterexamples to LIE-T: since the aim of the
speaker is to be deceptive in bald-faced lies, knowledge-lies, and coercion
lies, my view counts all three as lies. It also provides the right verdict in
creationist teacher and its modified version: Stella does not lie to
her students in the original scenario because her aim is to report what cur
rent scientific evidence supports regarding evolutionary theory, but she
does lie in the modified version since her intention is to conceal her own
religiously grounded beliefs about creationism in order to avoid termin
ation. Still further, it provides the correct verdict in the counterexamples to
conditions (3) of LIE-F and LIE-C: since there is the intention to be decep
tive in Shawn’s statement to the tribal community in deceptive
anthropologist and in Fran’s statement to Sam in sabotaging
friend, LIE-L correctly regards both as lies. Finally, my account distin
guishes lying from irony, joking, and acting since the intention to be decep
tive is present in the former statements, but not in the latter. While there
may be reason, then, to sever the connection between lying and the
intention to deceive, lying nonetheless remains fundamentally tied to the
intention to be deceptive.
5.5 Summativism and Sufficiency
Given the account of individual lying found in LIE-L, let’s now turn to
understanding group lies. To frame our discussion, it will be helpful to
return to the paradigmatic group lie presented in Chapter 1:
tobacco company: Philip Morris, one of the largest tobacco compan

ies in the world, is aware of the massive amounts of scientific evidence
revealing not only the addictiveness of smoking, but also the links it has
with lung cancer and heart disease. While the members of the board of
directors of the company believe this conclusion, they all jointly agree that,
For there is no doubt that Fran is lying to Sam, yet she also clearly intends to nullify any assur
ance of truth by explicitly disavowing responsibility for the truth of the statement.
GROUP LIES 179
because of what is at stake financially, the official position of Philip Morris is

that smoking is neither highly addictive nor detrimental to one’s health,
which is then published in all of their advertising materials.
Arguably, no group is more infamous for its lies than Philip Morris, and
thus the scenario described in tobacco company, which is fairly close
to non-fiction, is precisely the sort of case that any account of group lies
must have the resources to accommodate.
As with other group phenomena, perhaps group lying can be understood
in terms of what we might call simple summativism. Such a view can be
expressed as follows:
SS: A group, G, lies to B in stating thatp if and only if all of the members
of G lie to B in stating thatp.
The account of individual lying found in LIE-L can then be used to under
stand what it is for the members of G to lie to B. In this way, a group lie is
simply constructed out of the individual lies of its members.
There are, however, two immediate problems with SS that require minor
modification. First, requiring that all of the members of G lie to B in order
for the group to lie to B is too stringent. Surely, Philip Morris doesn’t fail to
lie to the public simply because a few of its employees are on vacation or
home ill and thus do not satisfy the conditions found in LIE. So perhaps SS
should be revised so that only some of the members of G need to lie to B in
order for the group to lie to B.
But this is still not enough to render SS plausible. In particular, what if
the individual members of the group who lie about whether p to B are all
utterly irrelevant to the domain in question? For instance, suppose that
while all of the members of the board of directors of Philip Morris do not lie
to B about the harmful effects of smoking, the custodians do. Is this enough
for Philip Morris to lie to B about this matter? Clearly not. What this shows
is that the individual lies in SS need to be made, not just by any members,
but by the right ones. Most groups have members with vastly different roles,
only some of whom have the authority or power to determine certain out
comes for the group as a whole. As we saw in earlier chapters, those who
have the relevant decision-making authority are often called operative
members.23 Thus, while the custodians in this case might at least in part
determine whether Philip Morris lies about the cleanliness of the company’s
28 See, for instance, Tuomela (2004).

facilities, they are irrelevant to whether the group lies about the harmful
effects of smoking; this is because they are operative members regarding the
former, but not the latter, question. SS should, then, be modified in the fol
lowing way:
SS*: A group, G, lies to B in stating that p if and only if most of the

operative members of G lie to B in stating that p.
SS* has clear virtues. It delivers the correct verdict that Philip Morris lies to
the public in tobacco company with ease: given that all of the members
of the board of directors individually lie in stating to the public that smok
ing does not cause lung cancer, the company lies about this as a group, too.
Moreover, it explains group lying while utilizing only resources from the
account of individual lying, which is not only simple but also avoids posit
ing phenomena, such as group belief, that require further explication.
Despite these virtues, SS* faces two objections that, to my mind, show
decisively that we need to look in an altogether different place to under
stand group lies. The first is that all of the operative members of G might
state that p to B, believe that p is false, and intend to be deceptive to B with
respect to whether p in stating thatp, but G still might not lie regarding this
question. Consider the following:
personal lies: Philip Morris has three operative members regarding

the question whether smoking causes lung cancer, Mx-M . All three mem
bers lie to B in stating that there is not a causal connection between the two,
but they do so entirely in the context of their personal relationships with
B. B is M^s wife, and Mx lies to her because he does not want her to worry
about their sons smoking habit when there is nothing that she can do to
prevent it. M2 is B’s best friend, and M2 lies to her so as to not cause marital
problems by contradicting M/s testimony. And M3 is B’s son, and M3 lies to
her so as to avoid his mother’s nagging to quit smoking.
In personal lies, though all three operative members of Philip Morris

satisfy conditions (1)—(3) of LIE and thereby lie to B regarding the question
whether smoking causes lung cancer, the tobacco company itself does not
lie to B. This is because each member lies to B, not in his or her role as an
employee of Philip Morris, but entirely because of the personal relationship
shared with B. Mx lies to B as her husband, M2 lies to B qua her best friend,
and M3 lies to B in his role as her son. Otherwise put, M -M would have
GROUP LIES 181
behaved exactly the same way toward B, even if none of them worked at the
tobacco company.29 This shows that the context in which an individual
member lies can affect whether the group itself lies. As we saw in Chapter 4,
similar considerations apply with respect to assertion or testimony more
broadly. If the CEO of Philip Morris testifies that smoking does not pose
health risks, whether it is the company’s statement depends on where and to
whom it was offered. It is one thing to state it at a board meeting to co
workers and quite another to say this to one’s spouse while on vacation.
personal lies thus reveals the following result about group lies: a
group can fail to lie to someone with respect to the question whether p,
despite the fact that every single one of its operative members lies to this
person regarding this question. SS*, then, does not specify sufficient condi
tions for a group lie.
5.6 Summativism and Necessity
A second objection to the version of simple summativism found in SS* is

that none of the operative members of G might state thatp to B, believe that
p is false, and intend to be deceptive to B with respect to whether p in stat
ing that p, but G might still lie to B regarding this question. Consider the
following:
manipulated spokesperson: Philip Morris hires spokesperson S—

who is not a member of the group and is known to be naive—to be the voice
of the company’s views. S’s job is to attend the meetings of the board of
directors at Philip Morris and, on the basis of the information therein pre
sented, to draw the relevant conclusions and convey them to the public on
the company’s behalf. At a recent meeting, the board of directors strategic
ally presented only the very small body of scientific evidence that indicates
that there is not a connection between smoking and lung cancer. This was
done both with the knowledge that there is overwhelming evidence show
ing that this is in fact false and with the deliberate intention that S would
29 Suppose that one were to argue that this is not enough to prevent the statements to B
from being group lies because M -M3 might be implicitly relying on a kind of authority from
being known to be experts on tobacco. In this case, even though they are not speaking in their
official capacity, what they say may be taken to represent the group’s view anyway. This compli
cation can be avoided by simply adding to the case that B is entirely unaware that M -M work
at Philip Morris.
then draw the conclusion that there is not such a connection and state this
to the public. Nevertheless, no operative member of Philip Morris actually
ever states, to one another in the boardroom or to S, that there is not a con
nection between smoking and lung cancer. Moreover, as the company
expected, S herself knows very little about the scientific evidence in ques
tion and so actually believes that there is not such a connection. At a recent
press conference, S, in her role as the official spokesperson for Philip
Morris, stated that there is not a connection between smoking and
lung cancer.
Let p be the proposition that there is not a connection between smoking

and lung cancer. In manipulated spokesperson, while all of the
operative members of Philip Morris believe that p is false, none of them
actually states that p, either to S or to the public. So, while they all satisfy
condition (2) of LIE, none of them satisfies (1) and (3) of LIE. Moreover,
while S states that p to the public, she believes that p is true and does not
intend to be deceptive to the public with respect to whether p in stating that
p. Instead, S’s intention is to convey the view of Philip Morris in stating that
p. S, then, satisfies (1) of LIE, but not (2) or (3). Thus, neither the operative
members of the company nor the spokesperson satisfies the three condi
tions found in LIE. Nevertheless, Philip Morris lies to the public about
the connection between smoking and lung cancer in manipulated
SPOKESPERSON.
To see this, notice that Philip Morris knowingly brings it about that S—
who is the company’s official spokesperson—says something false to the
public on its behalf, and does so with the explicit intention to be deceptive.
In this context, the mere fact that the words do not literally come out of the
mouths of the operative members of Philip Morris is irrelevant to whether
the group lies to the public. This is because, as we saw in Chapter 4, the
assertion, though offered by the spokesperson, is nevertheless the group's.
This, in turn, is due to the special relationship that exists between a spokes
person and the party she represents: the assertion she offers qua spokes
person is that of the represented party, not her own. Thus, she can
unknowingly lie on a group’s behalf while believing herself in the truth of
that which she is reporting.
This point can be made clearer by considering the distinction between
lying and misleading. Jennifer Saul provides the following case to illustrate
the difference between these phenomena: suppose that a politician, Tony,
GROUP LIES 183
believes that there are no weapons of mass destruction in Iraq but wants to
convince his audience otherwise. Now compare the two utterances that he
might offer in response to the question, “Are there weapons of mass destruc
tion in Iraq?”:
(a) There are weapons of mass destruction in Iraq.

(b) Saddam Hussein is a very dangerous man.
If we suppose that in both cases Tony’s intention is to bring it about that his
audience believes that Iraq has weapons of mass destruction, then while he
has lied in (a), he has merely misled in (b). A central difference between the
two is that Tony’s statement is true in only (b), despite the fact that he delib
erately conveyed something false in both (Saul 2012, p. 4).
Given this, one might ask why it isn’t the case that S’s statement is merely
an instance of misleading, not lying to, the public. By way of response to
this question, notice that the operative members of Philip Morris do merely
mislead S. By virtue of cherry-picking only the handful of studies that sug
gest that there is not such a connection, they offer true statements to S with
the deliberate intention to convey something false. But their misleading S
results in the company itself lying to the public. This is because all parties
involved recognize that S’s statements are on behalf of Philip Morris—that
her reports are the company’s assertion. In this way, so long as S is function
ing properly in her capacity as the spokesperson for Philip Morris, the com
pany itself asserted that there is not a connection between smoking and
lung cancer by virtue of S stating this. Thus, while the operative members’
manipulation of S avoids their lying to her, it does not get Philip Morris off
the hook of lying to the public.
This can be further supported by considering how easy and convenient it
would be for groups, such as corporations, to avoid lying if what Philip
Morris did in manipulated spokesperson failed to count as a lie.
Groups could hire naive spokespeople, mislead them with cherry-picked
information, and have false statements thereby conveyed, all while avoiding
any responsibility for lying. Surely this is an unwelcome result.
manipulated spokesperson thus reveals another result about
group lies: a group can lie to someone with respect to the question whether
p, despite the fact that not a single one of its operative members lies regard
ing this question. SS*, therefore, does not specify necessary conditions for a
group lie, either.
5.7 The Joint Acceptance Account of Group Lies
As we have seen, a common move that is made when summative accounts

of group phenomena fail is to embrace joint acceptance theories. Recall
that, according to these theories, collective phenomena, such as group belief
and group justification, must be understood in terms of the joint acceptance
of the operative members of the group. On such a view, it is neither neces
sary nor sufficient for a group state that its individual members instantiate
that state. Such an approach, then, might lend itself very nicely to account
ing for the cases that are problematic for SS*. Specifically, perhaps group lies
can be understood in terms of joint acceptance or agreement in the fol
lowing way:
JAA: A group, G, lies to B in stating that p if and only if most of the

operative members of G jointly agree to lie (in the sense found in LIE) to B
in stating thatp.30
The JAA delivers the correct result in personal lies with ease: even
though M1-M3 each lies to B in stating that there is not a causal connection
between smoking and lung cancer, they do not jointly agree to lie to B, and
thus Philip Morris doesn’t lie to B either. This view also seems to provide the
right verdict in manipulated spokesperson, at least on a certain
reading of the case. If, for instance, the board of directors jointly agreed to
strategically present only the very small body of scientific evidence that
indicates that there is not a connection between smoking and lung cancer,
and this was done with the intention of being deceptive toward the spokes
person, then it might be argued that this amounts to the operative members
agreeing to lie to the public regarding this question. Thus, Philip Morris
would end up lying to the public, too.
But it is not difficult to see that the JAA fails as a general account of group
lies. On the one hand, the operative members of a group jointly agreeing to
lie isn’t sufficient for a group’s lying. Suppose, for instance, that all of the
operative members of Philip Morris conceal from one another their per
sonal beliefs regarding the safety of smoking and yet, because of peer
30 Given that Fricker (2012) argues for a broadly joint acceptance account of group testi
mony—where a group’s testifying requires a joint commitment to trustworthiness—it is not
implausible to think she might espouse a joint acceptance account of group lying—where a
groups lying requires a joint commitment to untrustworthiness.
GROUP LIES 185
pressure, they all nonetheless agree to lie to the public that smoking is safe.
If it turns out that all of the operative members in fact believe that smoking
is safe, and there is a collective commitment to the proposition that smok
ing is safe, then the group hasn’t lied to the public in saying that it is. This is
evidenced by the fact that if the public learned all of the details of the case,
they might regard Philip Morris as ignorant and misinformed for believing
that smoking is safe, and they might even regard it as deceitful—given the
members’ agreement to lie—but they wouldn’t say that the company lied.
This is because in every sense, Philip Morris believes that smoking is safe
when it states that it is, despite the joint agreement to lie about this. In this
way, just as agreeing to be happy does not in fact make it the case that one is
happy, agreeing to lie does not in fact make it the case that one lies.
On the other hand, the operative members of a group jointly agreeing to
lie isn’t necessary for a group’s lying. For instance, suppose that everything
in manipulated spokesperson is exactly the same, except that the
board of directors strategically presented only the very small body of scien
tific evidence that indicates that there is not a connection between smoking
and lung cancer without jointly agreeing to do so. Philip Morris seems to be
engaged in a lie no less than in the original scenario. Yet, according to the
JAA, these are instances of two radically different phenomena, one involv
ing a group lie while the other doesn’t.
Moreover, we can imagine a slight variation to the paradigmatic group
lie—tobacco company—presented earlier in this chapter: suppose that
the members of the board of directors of Philip Morris decide to never
jointly agree to lie to the public precisely to avoid being responsible for
lying. Thus, the company is aware of the massive amounts of scientific evi
dence revealing the causal connection between smoking and lung cancer
and yet they deny such a connection to the public with the intention to be
deceptive. According to the JAA, Philip Morris doesn’t lie in such a case
because of the lack of joint acceptance to do so. But, surely, this is the
wrong result.
It is worth noting that these objections are not ones that can be avoided
through simple modification. Joint acceptance or agreement necessarily
requires intentional activity on the part of the members of the group at
issue. Indeed, as we have seen in earlier chapters, it is this very feature that
enables such accounts of group phenomena to avoid the problems facing
summative views. Regardless of what the individual members believe, for
example, the group’s doxastic states depend crucially on what the group
chooses to do. Similarly, even if each member of a group strategically reports
what she herself does not believe with the explicit intention to be deceptive,
the group itself does not lie unless there is joint agreement to do so.
However, while the intentional component of joint activity is what allows
for the response to summativism, it is also what leaves the view open to
decisive counterexample. For simply jointly agreeing to lie does not make it
the case that a group lies and merely refusing to jointly agree to lie does not
necessarily prevent a group from lying. Hence, it is the very heart of a joint
acceptance account of lying that is the problem.
5.8 Group Lies
We have seen that group lies cannot be understood in summative terms

since a group’s lying can diverge significantly from the lies of its individual
members. In particular, a group can fail to lie even though every operative
member does, and a group can lie even though no operative member does.
We have also seen that the standard non-summative move of embracing a
joint acceptance account doesn’t help here since a group can fail to lie even
though every operative member jointly agrees to, and a group can lie even
when there is no joint acceptance to do so. So where does that leave us?
I propose that we understand group lies in the following way:
G-LIE: A group, G, lies to B if and only if (1) G states that p to B, (2) G

believes thatp is false, and (3) G intends to be deceptive to B with respect to
whetherp in stating thatp.
The key lesson that we have learned earlier is that group lies need to be, in
an important sense, made by the group rather than by the individual mem
bers. G-LIE captures this by having the group be the agent at the center of
the view. Moreover, condition (1) can be understood in terms of the account
of group assertion offered in Chapter 4, and (2) can be fleshed out in terms
of the view of group belief developed in Chapter 1. As for (3), unlike with
group belief and group assertion, much work has been done on group
intentions,31 so this condition can be filled out with one’s preferred view.32
31 See, for instance, Tuomela (2006), Chant and Ernst (2007), Ludwig (2007), Gilbert
(2009), Pacherie (2013), and Kopec and Miller (2018).
32 Tliis is not entirely true, as some views of group intention will be unable to accommodate
most of the instances of lying discussed in this paper. For instance, accounts of intention that
require common knowledge, such as Bratman’s (1993) highly influential view of shared inten
tion, would be at odds with the view of group lies developed here.
GROUP LIES 187
Group lies, then, should be understood in terms of groups offering either

coordinated or authority-based assertions. What is important is that just as
groups can assert that p without any individual member asserting that p,
groups can lie that p without any individual member lying that p. This
divergence between what groups do and what their members do can be
explained precisely through spokespersons possessing authority to assert on
behalf of the group. This is why group assertion is the core feature of group
lying: while others might have the authority to speak or assert on our behalf,
it is far more puzzling how others might have the authority to believe or
intend on our behalf. Thus, the account of group assertion offered in Chapter 4
is what enables an understanding of the distinctiveness of group lies.
Moreover, the account of authority-based group assertion developed in
the previous chapter provides the framework for understanding not only
group lies, but also other phenomena in the neighborhood. For instance,
while the statement in manipulated spokesperson is clearly a group
lie, the assertion in autonomous spokesperson from Chapter 4 might
be better characterized as group bullshit. This is because while Philip Morris
clearly believes that it is false that smoking does not cause lung cancer in
manipulated spokesperson, thereby satisfying (2) of G-LIE, the
company does not have any relevant beliefs about disease X in autono
mous spokesperson. In this sense, it might fit better with Harry
Frankfurt’s description of bullshit, which, it may be recalled, he describes as
follows:
It is impossible for someone to lie unless he thinks he knows the truth.

Producing bullshit requires no such conviction. A person who lies is
thereby responding to the truth, and he is to that extent respectful of it.
When an honest man speaks, he says only what he believes to be true; and
for the liar, it is correspondingly indispensable that he considers his state
ments to be false. For the bullshitter, however, all these bets are off: he is
neither on the side of the true nor on the side of the false. His eye is not on
the facts at all, as the eyes of the honest man and the liar are, except insofar
as they may be pertinent to his interest in getting away with what he says.
He does not care whether the things he says describe reality correctly. He
just picks them out, or makes them up, to suit his purpose.
(Frankfurt 2005, pp. 55-6)
Because Philip Morris doesn’t have any relevant beliefs about disease X in
autonomous spokesperson but states that it is safe merely to suit its
economic purposes, the assertion seems to be a case of group bullshit. My

authority-based account of group assertion provides the resources for cap
turing this: because the spokesperson has the authority to assert on behalf
of Philip Morris when the safety of smoking is concerned, the assertion
about disease X is Philip Morris’s, and thus the company is the one bullshit
ting, not the spokesperson.
Finally, it is worth pointing out that the model of group assertion and
group lies presented in this book might be viewed as providing the frame
work for an account of group action more broadly. That is, group action in
general might be understood as either coordinated or authority-based, with
the latter involving, not always a spokesperson, but another agent who has
the authority to act on behalf of the group. In this way, the view developed
here can shape our grasp of group agency and, therewith, group responsi
bility in ways that go far beyond lying and bullshitting.
5.9 Conclusion
In this chapter, I have provided an account of group lies. On my view, a

group’s lying cannot be understood merely in terms of features that take
place at the level of its members, such as their offering individual lies or
jointly agreeing to lie. Instead, it is the group itself that lies, in virtue of the
group’s stating that p, the group’s believing that p is false, and the group’s
intending to be deceptive with respect to whether p in stating thatp. A cen
tral virtue of my account is that it appreciates the unique relationship that
exists between most groups and their spokespersons, as well as the subtle
and complex interactions made possible by that relationship, such as the
possibility that what a group says may come apart from what its individual
members say. In these ways, my view provides the basis not only for
understanding how groups are responsible for their linguistic behavior, but
also for determining when it is appropriate to trace this responsibility to
the individual members of the group and the spokespersons who repre
sent them.
References
Adler, Jonathan E. 2002. Belief's Own Ethics. Cambridge, MA: The MIT Press.
Alston, William P. 1988. “The Deontological Conception of Epistemic Justification.”
Philosophical Perspectives 2: 257-99.
Antony, Louise. 2011. “Against Langton’s Illocutionary Treatment of Pornography.”
Jurisprudence 2: 387-401.
Audi, Robert. 1997. “The Place of Testimony in the Fabric of Knowledge and
Justification.” American Philosophical Quarterly 34:405-22.
Audi, Robert. 1998. Epistemology: A Contemporary Introduction to the Theory of
Knowledge. London: Routledge.
Audi, Robert. 2006. “Testimony, Credulity, and Veracity,” in Jennifer Lackey and
Ernest Sosa (eds.), The Epistemology of Testimony. Oxford: Oxford University
Press, 25-49.
Augustine. 1952 [395]. “Lying” in Roy. J. Deferrari (ed.), Treatises on Various
Subjects, Volume XVI. New York: Fathers of the Church, 53-120.
Austin, J. L. 1962. How to Do Things with Words. Cambridge, MA: Harvard
University Press.
Bergmann, Michael. 1997. “Internalism, Externalism and the No-Defeater
Condition.” Synthese 110: 399-417.
Bird, Alexander. 2010. “Social Knowing: The Social Sense of‘Scientific Knowledge?’
Philosophical Perspectives 24: 23-56.
Bird, Alexander. 2014. "When Is there a Group that Knows? Distributed Cognition,
Scientific Knowledge, and the Social Epistemic Subject,” in Jennifer Lackey (ed.),
Essays in Collective Epistemology. Oxford: Oxford University Press, 42-63.
Bonjour, Laurence. 1980. “Externalist Theories of Epistemic Justification.” Midwest
Studies in Philosophy 5: 53-73.
Bonjour, Laurence. 1985. The Structure of Empirical Knowledge. Cambridge, MA:
Harvard University Press.
Bonjour, Laurence and Ernest Sosa. 2003. Epistemic Justification: Internalism vs.
Externalism, Foundations vs. Virtues. Oxford: Blackwell Publishing.
Bratman, Michael. 1993. “Shared Intention.” Ethics 104: 97-113.
Briggs, Rachael, Fabrizio Cariani, Kenny Easwaran, and Branden Fitelson. 2014.
“Individual Coherence and Group Coherence,” in Jennifer Lackey (ed.), Essays in
Collective Epistemology. Oxford: Oxford University Press, 215-39.
Broncano-Berrocal, Fernando. 2013. “Lies and Deception: A Failed Reconciliation.”
Logos and Episteme 4: 227-30.
Burge, Tyler. 1993. “Content Preservation.” The Philosophical Review 102:457-88.
Burge, Tyler. 1997. “Interlocution, Perception, and Memory.” Philosophical Studies
86: 21-47.
Cariani, Fabrizio. 2011. “Judgment Aggregation.” Philosophy Compass 6: 22-32.
190 REFERENCES
Cariani, Fabrizio. 2013. “Aggregating with Reasons.” Synthese 190: 3123-47.

Carson, Thomas L. 2010. Lying and Deception: Theory and Practice. Oxford: Oxford
University Press.
Carter, J. Adam. 2015. “Group Knowledge and Epistemic Defeat.” Ergo 2: 711-35.
Chant, Sara Rachel and Zachary Ernst. 2007. “Group Intentions as Equilibria.”
Philosophical Studies 133: 95-109.
Chisholm, Roderick M. 1989. Theory of Knowledge, 3rd edn. Englewood Cliffs, NJ:
Prentice-Hall.
Chisholm, Roderick M. and Thomas D. Feehan. 1977. “The Intent to Deceive.” The
Journal of Philosophy 74:143-59.
Christensen, David. 2004. Putting Logic in Its Place. Oxford: Oxford University Press.
Coady, C. A. J. 1992. Testimony: A Philosophical Study. Oxford: Clarendon Press.
Cohen, L. Jonathan. 1989. “Belief and Acceptance.” Mind 98: 367-89.
Cohen, L. Jonathan. 1992. An Essay on Belief and Acceptance. Oxford:
Clarendon Press.
Cohen, Stewart. 2004. “Knowledge, Assertion, and Practical Reasoning,” in Ernest
Sosa and Enrique Villanueva (eds.), Philosophical Issues 14: 482-91.
Corlett, J. Angelo. 1996. Analyzing Social Knowledge. Lanham, MD: Rowman &
Littlefield.
Corlett, J. Angelo. 2007. “Analyzing Social Knowledge.” Social Epistemology 21:231-47.
Craig, Edward. 1990. Knowledge and the State of Nature: An Essay in Conceptual
Synthesis. Oxford: Clarendon Press.
Davidson, Donald. 2001. Essays on Actions and Events. Oxford: Oxford
University Press.
DeRose, Keith. 2002. “Assertion, Knowledge, and Context.” The Philosophical Review
111: 167-203.
Dietrich, Franz. 2005. “Judgment Aggregation: (Im)possibility Theorems.” Journal of
Economic Theory 126: 286-98.
Douven, Igor. 2006. “Assertion, Knowledge, and Rational Credibility.” The
Philosophical Review 115: 449-85.
Elgin, Catherine Z. 2002. “Take It from Me: The Epistemological Status of
Testimony.” Philosophy and Phenomenological Research 65:291-308.
Fallis, Don. 2006. “Epistemic Value Theory and Social Epistemology.” Episteme
2:177-88.
Fallis, Don. 2009. What Is Lying?” The Journal of Philosophy 106: 29-56.
Fallis, Don. 2014. “Are Bald-Faced Lies Deceptive After All?” Ratio 28:81-96.
Fantl, Jeremy and Matthew McGrath. 2002. “Evidence, Pragmatics, and Justification.”
The Philosophical Review 111: 67-94.
Fantl, Jeremy and Matthew McGrath. 2009. Knowledge in an Uncertain World.
Oxford: Oxford University Press.
Faulkner, Paul. 2006. “On Dreaming and Being Lied To.” Episteme 3:149-59.
Foley, Richard. 1987. The Theory of Epistemic Rationality. Cambridge, MA: Harvard
University Press.
Frankfurt, Harry G. 2005. On Bullshit. Princeton, NJ: Princeton University Press.
Fricker, Elizabeth. 1987. “The Epistemology of Testimony,” Proceedings of the
Aristotelian Society, 61 (supp): 57-83.
REFERENCES 191
Fricker, Elizabeth. 1994. “Against Gullibility,” in Bimal Krishna Matilal and Arindam
Chakrabarti (eds.), Knowing from Words. Dordrecht: Kluwer Academic
Publishers, 125-61.
Fricker, Elizabeth. 1995. “Telling and Trusting: Reductionism and Anti-
Reductionism in the Epistemology of Testimony.” Mind 104: 393-411.
Fricker, Elizabeth. 2006. “Second-hand Knowledge.” Philosophy and
Phenomenological Research 73: 592-618.
Fricker, Miranda. 2007. Epistemic Injustice: Power and the Ethics of Knowing. Oxford:
Oxford University Press.
Fricker, Miranda. 2010. “Can There Be Institutional Virtues?” in Tamar Szabo
Gender and John Hawthorne (eds.), Oxford Studies in Epistemology. Oxford:
Oxford University Press, 235-52.
Fricker, Miranda. 2012. “Group Testimony: The Making of a Collective Good
Informant.” Philosophy and Phenomenological Research 84: 249-76.
Fumerton, Richard. 2004. “Epistemic Probability.” Philosophical Issues 14:149-64.
Gettier, Edmund. 1963. “Is Justified True Belief Knowledge?” Analysis 23:121-3.
Gilbert, Margaret. 1987. “Modelling Collective Belief.” Synthese 73: 185-204.
Gilbert, Margaret. 1989. On Social Facts. London and New York: Routledge.
Gilbert, Margaret. 1993. “Agreements, Coercion, and Obligation.” Ethics 103:
679-706.
Gilbert, Margaret. 1994. “Remarks on Collective Belief,” in Frederick F. Schmitt
(ed.), Socializing Epistemology: The Social Dimensions of Knowledge. Lanham,
MD: Rowman & Littlefield, 235-55.
Gilbert, Margaret. 2002. “Belief and Acceptance as Features of Groups.”
Protosociology 16: 35-69.
Gilbert, Margaret. 2004. “Collective Epistemology.” Episteme 1: 95-107.
Gilbert, Margaret. 2009. “Shared Intention and Personal Intentions.” Philosophical
Studies 144: 167-87.
Gilbert, Margaret. 2013. Joint Commitment: How We Make the Social World. Oxford:
Oxford University Press.
Gilbert, Margaret and Daniel Pilchman. 2014. “Belief, Acceptance, and What
Happens in Groups: Some Methodological Considerations,” in Jennifer Lackey
(ed.), Essays in Collective Epistemology. Oxford: Oxford University Press, 189-212.
Goldberg, Sanford C. 2015. Assertion: On the Philosophical Significance ofAssertoric
Speech. Oxford: Oxford University Press.
Goldberg, Sanford C. 2017. “Should Have Known.” Synthese 194: 2863-94.
Goldman, Alvin. I. 1979. “What Is Justified Belief?” in George Pappas (ed.),
Justification and Knowledge. Dordrecht: Reidel, 89-104.
Goldman, Alvin. I. 1986. Epistemology and Cognition. Cambridge, MA: Harvard
University Press.
Goldman, Alvin. I. 2004. “Group Knowledge Versus Group Rationality: Two
Approaches to Social Epistemology.” Episteme 1:11-22.
Goldman, Alvin. I. 2014. “Social Process Reliabilism: Solving Justification Problems
in Collective Epistemology,” in Jennifer Lackey (ed.), Essays in Collective
Epistemology. Oxford: Oxford University Press, 11-41.
Graham, Peter J. 1997. “What Is Testimony?” The Philosophical Quarterly 47:227-32.
192 REFERENCES
Grice, Paul. 1989. Studies in the Way of Words. Cambridge, MA: Harvard
University Press.
Hagemann, Thomas A. and Joseph Grinstein. 1997. “The Mythology of Aggregate
Corporate Knowledge: A Deconstruction.” George Washington Law Review
65: 210-47.
Hakli, Raul. 2007. “On the Possibility of Group Knowledge without Group Belief.”
Social Epistemology 21:249-66.
Hakli, Raul. 2011. “On Dialectical Justification of Group Beliefs,” in Hans Bernhard
Schmid, Daniel Sirtes, and Marcel Weber (eds.), Collective Epistemology.
Frankfurt: Ontos Verlag, 119-53.
Hardwig, John. 1985. “Epistemic Dependence.” The Journal of Philosophy 82: 335-49.
Hardwig, John. 1991. “The Role of Trust in Knowledge.” The Journal of Philosophy
88: 693-708,
Harman, Gilbert. 1973. Thought. Princeton NJ: Princeton University Press.
Hawthorne, John. 2004. Knowledge and Lotteries. Oxford: Oxford University Press.
Hawthorne, John and Jason Stanley. 2008. “Knowledge and Action.” The Journal of
Philosophy 105: 571-90.
Huebner, Bryce. 2014. Macrocognition: A Theory of Distributed Minds and Collective
Intentionality. New York: Oxford University Press.
Hughes, Justin. 1984. “Group Speech Acts.” Linguistics and Philosophy 7: 379-95.
Hutchins, Edwin. 1995. Cognition in the Wild. Cambridge, MA: The MIT Press.
Isenberg, Arnold. 1964. “Deontology and the Ethics of Lying.” Philosophy and
Phenomenological Research 24:465-80.
Kallestrup, Jesper. 2016. "Group Virtue Epistemology.” Synthese doi: 10.1007/
si 1229-016-1225-7.
Kenyon, Tim. 2003. “Convention, Pragmatics, and Saying ‘Uncle.” American
Philosophical Quarterly 40:241-8.
Klausen, Soren Harnow. 2015. “Group Knowledge: A Real-Wo rid Approach.”
Synthese 192: 813-39.
Klein, Peter. 1985. “The Virtues of Inconsistency.” TheMonist68:105-35.
Kolodny, Niko. 2007. “How Does Coherence Matter?” Proceedings of the Aristotelian
Society 107:229-63.
Kopec, Matthew and Seumas Miller. 2018. “Shared Intention Is Not Joint
Commitment.” Journal of Ethics and Social Philosophy 13:179-89.
Krishna, Daya. 1961. “‘Lying’ and the Compleat Robot.” British Journal for the
Philosophy of Science 12:146-9.
Kukla, Rebecca. 2012. “‘Author TBD’: Collaboration in Contemporary Biomedical
Research. Philosophy of Science 79: 845-58.
Kyburg, Henry E. 1961. Probability and the Logic of Rational Belief. Middleton, CT:
Wesleyan University Press.
Kyburg, Henry E. 2001. “Probability as a Guide in Life.” The Monist 84:135-52.
Lackey, Jennifer. 2006. “The Nature of Testimony.” Pacific Philosophical Quarterly
87: 177-97.
Lackey, Jennifer. 2007. “Norms of Assertion.” Nous 41: 594-626.
Lackey, Jennifer. 2008. Learning from Words: Testimony as a Source of Knowledge.
Oxford: Oxford University Press.
REFERENCES 193
Lackey, Jennifer. 2010. “Acting on Knowledge.” Philosophical Perspectives 24: 361-82.

Lackey, Jennifer. 2013. “Lies and Deception: An Unhappy Divorce.” Analysis
73\ 236-48.
Lackey, Jennifer. 2014a. “A Deflationary Account of Group Testimony,” in Jennifer
Lackey (ed.), Essays in Collective Epistemology. Oxford: Oxford University
Press, 64-94.
Lackey, Jennifer. 2014b. “Socially Extended Knowledge.” Philosophical Issues 24:282-98.
Lackey, Jennifer. 2016. “What Is Justified Group Belief?” The Philosophical Review
125: 341-96.
Lackey, Jennifer. 2018a. “Group Assertion.” Erkenntnis 83:21-42.
Lackey, Jennifer. 2018b. “Group Lies,” in Eliot Michaelson and Andreas Stokke
(eds.), Lying: Language, Knowledge, Ethics, and Politics. Oxford: Oxford University
Press, 262-84.
Lahroodi, Reza. 2007. “Collective Epistemic Virtues.” Social Epistemology 21:281-97.
Langton, Rae. 2009. Sexual Solipsism: Philosophical Essays on Pornography and
Objectification. Oxford: Oxford University Press.
Levi, Isaac. 1962. “On the Seriousness of Mistakes.” Philosophy of Science 29:47-65.
List, Christian. 2005. “Group Knowledge and Group Rationality: A Judgment
Aggregation Perspective.” Episteme 2: 25-38.
List, Christian. 2014a. “Three Kinds of Collective Attitudes.” Erkenntnis 79(9): 1601-22.
List, Christian. 2014b. “When to Defer to Supermajority Testimony—and When
Not,” in Jennifer Lackey (ed.), Essays in Collective Epistemology. Oxford: Oxford
University Press, 240-9.
List, Christian and Phillip Pettit. 2002. “Aggregating Sets of Judgments: An
Impossibility Result.” Economics and Philosophy 18: 89-110.
List, Christian and Phillip Pettit. 2004. “Aggregating Sets of Judgments: Two
Impossibility Results Compared.” Synthese 140: 207-35.
List, Christian and Phillip Pettit. 2011. Group Agency: The Possibility, Design, and
Status of Corporate Agents. Oxford: Oxford University Press.
Ludwig, Kirk. 2007. “Collective Intentional Behavior from the Standpoint of
Semantics.” Nous 41: 355-93.
Ludwig, Kirk. 2014. “Proxy Agency in Collective Action.” Nous 48: 75-105.
Mahon, James Edwin. 2008. “The Definition of Lying and Deception,” in Edward N. Zalta
(ed.), The Stanford Encyclopedia of Philosophy (fall 2008 edn). Available at: https://
plato.stanford.edu/entries/lying-definition/ accessed April 27,2020.
Makinson, David. 1965. “The Paradox of the Preface,” Analysis 25:205-7.
Makinson, David. 2012. “Logical Questions Behind the Lottery and Preface
Paradoxes: Lossy Rules for Uncertain Inference.” Synthese 186: 511-29.
Mathiesen, Kay. 2011. “Can Groups Be Epistemic Agents?,” in Hans Bernhard
McDowell, John. 1994. "Knowledge by Hearsay,” in Bimal Krishna Matilal and
Arindam Chakrabarti (eds.), Knowingfrom Words. Dordrecht: Kluwer Academic
Publishers, 195-224.
McKinnon, Rachel. 2013. “The Supportive Reasons Norm of Assertion.” American
Philosophical Quarterly 50: 121-35.
194 REFERENCES
McMahon, Christopher. 2003. “Two Modes of Collective Belief.” Protosociology

18/19: 347-62.
Meijers, A. W. M. 2002. “Collective Agents and Cognitive Attitudes.” Protosociology
16: 70-85.
Mokyr, Joel. 2002. The Gifts ofAthena, Princeton, NJ: Princeton University Press.
Nozick, Robert. 1981. Philosophical Explanations, Cambridge, MA: The
Belknap Press.
Owens, David. 2000. Reason without Freedom: The Problem of Epistemic Normativity.
London: Routledge.
Owens, David. 2006. “Testimony and Assertion.” Philosophical Studies 130:105-29.
Pacherie, Elisabeth. 2013. “Intentional Joint Agency: Shared Intention Lite.” Synthese
190: 1817-39,
Pauly, Marc and Martin van Hees. 2006 “Logical Constraints on Judgement
Aggregation.” Journal of Philosophical Logic 35: 569-85.
Pettit, Philip. 2003. “Groups with Minds of Their Own,” in Frederick Schmitt (ed.),
Socializing Metaphysics. New York: Rowman & Littlefield, 167-93.
Plantinga, Alvin. 1993. Warrant and Proper Function. Oxford: Oxford
University Press.
Platts, Mark. 1979. Ways of Meaning. London: Routledge and Kegan Paul.
Pollock, John. 1986. Contemporary Theories of Knowledge. Totowa, NJ: Rowman &
Littlefield.
Quinton, Anthony. 1975/1976. “Social Objects.” Proceedings of the Aristotelian
Society 75:1-27.
Ragozino, Anthony. 1995. “Replacing the Collective Knowledge Doctrine with a
Better Theory for Establishing Corporate Mens Rea: The Duty Stratification
Approach.” Southwestern University Law Review 24:423-72.
Reed, Baron. 2006. “Epistemic Circularity Squared? Skepticism about Common
Sense.” Philosophy and Phenomenological Research 73:186-97.
Reynolds, Steven L. 2002. “Testimony, Knowledge, and Epistemic Goals.”
Philosophical Studies 110:139-61.
Riggs, Wayne D. 2008. “Epistemic Risk and Relativism.” Acta Analytica 23:1-8.
Ritchie, Katherine. 2013. “What Are Groups?” Philosophical Studies 166:257-72.
Ross, Angus. 1986. “Why Do We Believe What We Are Told?” Ratio 28: 69-88.
Rupert, Robert D. 2005. “Minding One’s Cognitive Systems: When Does a Group of
Minds Constitute a Single Cognitive Unit?” Episteme 1:177-88.
Rupert, Robert D. 2011. “Empirical Arguments for Group Minds: A Critical
Appraisal.” Philosophy Compass 6: 630-9.
Saul, Jennifer Mather. 2012. Lying Misleading, and What Is Said. Oxford: Oxford
University Press.
Schmitt, Frederick F. 1994. “The Justification of Group Beliefs,” in
Frederick F. Schmitt (ed.), Socializing Epistemology: The Social Dimensions of
Knowledge. Lanham, MD: Rowman & Littlefield, 257-87.
Schmitt, Frederick F. 2006. “Testimonial Justification and Transindividual Reasons,”
in Jennifer Lackey and Ernest Sosa (eds.), The Epistemology of Testimony. Oxford:
Oxford University Press, 193-224.
Searle, John. 1995. The Construction of Social Reality. New York: Free Press.
REFERENCES 195
Silva Jr., Paul. 2019. “Justified Group Belief Is Evidentially Responsible Group Belief.’*
Episteme 16: 262-81.
Sorensen, Roy. 2007. “Bald-faced Lies! Lying without the Intent to Deceive.” Pacific
Philosophical Quarterly 88: 251-64.
Sorensen, Roy. 2010. “Knowledge-Lies.” Analysis 70: 608-15.
Staffel, Julia. 2011. “Reply to Roy Sorensen, ‘Knowledge-Lies’.” Analysis 71: 300-2.
Stalnaker, Robert. 1984. Inquiry. Cambridge, MA: The MIT Press.
Stanley, Jason. 2005. Knowledge and Practical Interests. Oxford: Oxford University Press.
Tollefsen, Deborah. 2007. “Group Testimony.” Social Epistemology 21: 299-311.
Tollefsen, Deborah. 2009. “ Wikipedia and the Epistemology of Testimony.” Episteme
6: 8-24.
Tollefsen, Deborah. 2014. “Review of Bryce Huebner’s Macrocognition: A Theory of
Distributed Minds and Collective Intentionality! Notre Dame Philosophical
Reviews. Available at: https://ndpr.nd.edu/news/macrocognition-a-theory-of-
distributed-minds-and-collective-intentionality/, accessed April 26, 2020.
Tuomela, Raimo. 1992. “Group Beliefs.” Synthese 91: 285-318.
Tuomela, Raimo. 1993. “Corporate Intention and Corporate Action.” Analyse und
Kritik 15:11-21.
Tuomela, Raimo. 1995. The Importance of Us. Stanford, CA: Stanford University Press.
Tuomela, Raimo. 2004. “Group Knowledge Analyzed.” Episteme 1:109-27.
Tuomela, Raimo. 2006. “Joint Intention, We-Mode and I-Mode.” Midwest Studies in
Philosophy 30: 35-58.
Tuomela, Raimo. 2011. “An Account of Group Knowledge,” in Hans Bernhard
Unger, Peter. 1975. Ignorance: A Casefor Skepticism. Oxford: Oxford University Press,
van Fraassen, Bas C. 1980. The Scientific Image. Oxford: Clarendon Press.
Velasquez, Manuel. 2003. “Debunking Corporate Moral Responsibility.” Business
Ethics Quarterly 13: 531-62.
Welbourne, Michael. 1979. “The Transmission of Knowledge.” The Philosophical
Quarterly 29:1-9.
Welbourne, Michael. 1981. “The Community of Knowledge.” The Philosophical
Quarterly 31: 302-14.
Welbourne, Michael. 1986. The Community of Knowledge. Aberdeen: Aberdeen
University Press.
Welbourne, Michael. 1994. “Testimony, Knowledge and Belief,” in Bimal Krishna
Matilal and Arindam Chakrabarti (eds.), Knowing from Words. Dordrecht:
Kluwer Academic Publishers, 297-313.
Wigmore, John H. 1904. “The History of the Hearsay Rule.” Harvard Law Review
17: 437-58.
Williams, Bernard. 2002. Truth and Truthfulness: An Essay in Genealogy. Princeton,
NJ: Princeton University Press.
Williams, Michael. 1999. Groundless Belief: An Essay on the Possibility of
Epistemology, 2nd edn. Princeton, NJ: Princeton University Press.
Williamson, Timothy. 1996. “Knowing and Asserting.” The Philosophical Review
105: 489-523.
196 REFERENCES
Williamson, Timothy. 2000, Knowledge and Its Limits. Oxford: Oxford

University Press.
Williamson, Timothy 2005. “Contextualism, Subject-Sensitive Invariantism and
Knowledge of Knowledge.*’ The Philosophical Quarterly 55: 213-35.
Wray, K. Brad. 2001. “Collective Belief and Acceptance.” Synthese 129: 319-33.
Wray, K. Brad. 2003. “What Really Divides Gilbert and the Rejectionists.”
Protosociology 18/19: 363-76.
Index
For the benefit of digital users, indexed terms that span two pages (e.g., 52-53) may, on
occasion, appear on only one of those pages.
ACLU of Illinois 19 Condorcet jury theorem 91

Adler, Jonathan 159 n. 25 Corlett, J. Angelo 23-24, 72 n. 19
aggregation 13,25 n. 13,28-30,37-44,49,55, Craig, Edward 155
71-76,78-80,91, 96-98,129,135-137
Alston, William 34 n. 24 Davidson, Donald 122 n. 19
Antony Louise 142 n. 6 defeater 37 n. 28, 57,62, 81-82, 87-89,
Audi, Robert 82 n. 28,150 n. 19,163 n. 32 106 n. 51,108,123-127
Augustine 166,170 n. 11 DeRose, Keith 119,159 n. 25
Dietrich, Franz 28 n. 17,72 n. 18
Baril, Anne 152 n. 20 discursive dilemma 29 see also
Bergmann, Michael 82 n. 28 doctrinal paradox
Bird, Alexander 7, 8 n. 11,14,19 n. 1, divergence arguments 56-57, 59,62,
22 n. 8, 111-115, 120-121,126 n. 24 67-68, 71
Bonjour, Laurence 82 n. 28-29 different evidence cases 56-57,59,
Bratman, Michael 147 n. 13,186 n. 32 62-63,68-71
Briggs, Rachel 50 n. 33,99 n. 45 different epistemic risk settings cases 56,
Broncano-Berrocal, Fernando 170 n. 11 58-59,62, 70-71
Burge, Tyler 82 n. 28,163 n. 32 doctrinal paradox 29,97 n. 39 see also
discursive dilemma
Cariani, Fabrizio 50 n. 33,72 n. 18, 78 n. 24, Douven, Igor 159 n. 26
99 n. 45
Carson, Thomas 30 n. 20, 166 n. 4,168-172, Easwaran, Kenny 50 n. 33, 99 n. 45
175 n. 24, 177 n. 27 Elgin, Catherine Z, 142 n. 6
Carter, J. Adam 37 n. 28 Ernst, Zachary 186 n. 31
Chant, Sara Rachel 186 n, 31
Chisholm, Roderick M. 82 n. 28,166-167, Fallis, Don 30 n. 20, 58 n. 3,166 n. 4-5,
173 n. 19 169-170,172-173, 175-177
Christensen, David 81 n. 26 Fantl, Jeremy 94 n. 35,117,119
Coady, C.A.J. 142 n. 6 Faulkner, Paul 163 n. 32
Cohen, L. Jonathan 25 n. 12,144 n. 9 Feehan, Thomas D. 166-167,173 n. 19
Cohen, Stewart 107 n. 52,159 n. 25 Fitelson, Branden 50 n. 33,99 n. 45
Coherence 50, 99-100 see also incoherence Foley, Richard 81 n. 26
collective action 3,17,141 Frankfurt, Harry G. 32-33,35,170 n. 12,187
proxy agent 14,17, 94,116 n. 8,134, Fricker, Elizabeth 82 n. 28,142 n. 6,159 n. 25
140-141,146 n. 11 Fricker, Miranda 22 n. 8,142 n. 5-6,
collective knowledge 7,14,72 n. 19, 154-155,184 n. 30
111-112,128-137 Fumerton, Richard 101 n. 48
collective responsibility 2-3 see also group
responsibility Gettier, Edmund 56 n. 2, 74-75
Collins, Rep. Chris 1-2 Gilbert, Margaret 6-8,19 n. 1, 22-27,
Comesana, Juan 107 n. 52 35 n. 26,186 n.31
198 INDEX
Goldberg, Sanford C. 128 n. 27,160 deliberative and non-deliberative 9-11,

Goldman, Alvin I. 26 n. 14,71-77, 82-83, 25 n. 13
92, 96,101 n. 48 established and non-established
Graham, Peter J. 142 n. 6 7-8,11,112
Grice, Paul 175 n. 23,176 n. 25 operative members 26-27,29-30, 32-33,
Grinstein, Joseph 129 48-50, 54,64-65,67, 97-101,103-106,
group assertion 3,14,17-18,138-140, 111,133-134,145,179-186
146 n. 12,147,149-151, 154,158-159, organic 112-113
161-163, 186-187 subject to normative evaluation 9,12
authority-based 14,138,140,148-150, versus mere collections 6,8-11
161-163,186-188 with “minds of their own” 3,20, 53,115
coordinated 14,139,149-150,156,163, group responsibility 123,188
186,188 see also collective responsibility
joint acceptance account 154
group belief 4, 6,10, 12-14,17-54,113,124 Hagemann, Thomas A. 129
base fragility 45-48, 50, 54, 78 n. 24 Hakli, Raul 25 n. 12, 34,52 n. 35,61,144 n. 9
direction of fit 13,44,50,54,147 Halliburton 19
eliminativism 4, 52 n. 35 Hardwig, John 163 n. 32
Group Agent Account 13,20, 48-52, 54 Harman, Gilbert 104 n. 50
joint acceptance account 12,17, 20, Hawthorne, John 82 n. 28,94 n. 35,101 n. 48,
24-29, 32-38,40-41,43-45,49,124 117,119,159 n. 25
judgment fragility 41-43,49 Horn, Michael 1
premise-based aggregation Huckabee Sanders, Sarah 19
account 28-29,37, 39-44 Huebner, Bryce 19 n. 1,126-127
summative view 20-21, 23, 30 n. 18,40, Hughes, Justin 158 n. 24
48, 50, 52 n. 35 Hutchins, Edwin 7,111-112
group bullshit 30, 33-37, 39-41,43,
51-52,187 incoherence 12,46,49 n. 32,50 see also
Group Bullshit Desideratum 34,37 coherence
group lies 4-5,36-41,43,48,51-54, Isenberg, Arnold 166 n. 3
165—166,178-188 see also lies
Group Lie Desideratum 31,34,37 justified group belief 3,6,13-14,18, 55-56,
j oint acceptance account 183-186 71 n. 16,73,79-80, 91-98,100,102,
summativism 178—179,181,183,185-186 104-106,107-109, 111, 128,133-134,
group phenomena 3, 6,105,179,183,185 137,161
deflationary views 3-4,13-17,55,71-72, Collective Evidence Problem 84-88,
75, 88, 92, 95, 98,107,109,138-39, 101,105
159-61,163 Condorcet-inspired account of justified
inflationary views 3,12-18,55-56,59-60, group belief 91-95
63-64, 67, 71, 92, 95, 98,100, 107,109, Defeater Problem 71,77,81-84,
111-112, 115,138-139, 158-159, 98,100-101
161-163 deflationary summativism 71-72,75, 88,
groups 3-20,23 n. 9,25 n. 13,26-27, 29, 32, 92, 95, 98,107-109
34,48, 53-57,64,69-70, 94-96, Group Epistemic Agent Account 13-14,56,
99-100,104,108,112-113,115-116, 95,97-98,100-101,103-107,109-110
120 n. 17,124,126-128,134-135, Group Justification Paradox 13, 71,
138-139,141-143,146 n. 11,148-149, 77-81, 83 n. 30, 84,98,100-101
151,154-155,157,159,162-164,166, Group Normative Obligations Problem
179,183, 186,188 88-89,101,106
INDEX 199
Illegitimate Manipulation of Evidence Owens, David 163 n. 32

Problem 13,67, 70, 95, 100-101
joint acceptance account 6,13, 55,59-68, Pacherie, Elisabeth 186 n. 31
71,95, 98,100,108-109,113, 124 Pauly, Marc 28 n. 17,72 n. 18
Pettit, Philip 3 n. 7, 8, 19 n. 1,20 n. 7, 28,29,
Kallestrup, Jesper 6 n. 9,69 n. 15, 37,40, 53 n. 37, 72,115-116
126 n. 24 Pilchman, Daniel 35 n. 26
Kenyon, Tim 168 n. 9 Plantinga, Alvin 82 n. 28,163 n. 32
Klausen, Soren Harnow 7,112 n. 2, 127 Platts, Mark 44
Klein, Peter 81 n. 26 Pollock, John 82 n. 28
Kolodny, Niko 50 n. 33, 99 n. 45 Preface Paradox 49 n. 32, 80-81, 99
Kopec, Matthew 186 n. 31
Krishna, Daya 173 n. 18 Quinton, Anthony 4,12
Kukla, Rebecca 127 n, 25
Kyburg, Henry E. 81 n. 26,101 n. 48 Ragozino, Anthony 129
Reed, Baron 82 n. 28-29
Lackey, Jennifer 15 n. 16, 30 n. 19-20, Reller, Fredric 31
82 n. 28-29, 94 n. 36-37,118 n. 12, Reynolds, Steven L, 159 n. 25,163 n. 32
138-139,142 n. 6,150 n. 16,156, Riggs, Wayne D. 58 n. 3, 99 n. 45
158-159,161-163,165 n. 2, Ritchie, Katherine 19 n. 1
174 n. 20 Ross, Angus 163 n. 32,174 n. 21
Lafont, Cristina 174 n. 21 Rupert, Robert D. 19 n. 1
Lahroodi, Reza 8 n. 11
Langton, Rae 142 n. 6 Saul, Jennifer Mather 182-183
Lauffer, Nathan 42 n. 29 Schmitt, Frederick E 6, 22-24, 57, 59 n. 5,
Leonard, Nick 103 n. 49 60-61, 69-70
Levi, Isaac 58 n. 3,104 Searle, John 140
lies 15,19 n. 1, 30, 32-33, 53-54,155, selfless assertion 167,173-174
165-188 see also group lies Silva Jr., Paul 97 n. 39,106 n. 51
bald-faced 166-168,170-173,178 social knowledge 14,111-115,119,121-123,
coercion 166, 169-171,173,178 125-128,137
knowledge 166,168-172,178 Sorensen, Roy 30 n. 20,166-170,173 n. 19
List, Christian 8,19 n. 1, 28 n. 17, 39, 41-43, Sosa, Ernest 82 n. 28
60-63,72, 74,91,99 n. 44,116 spokesperson 2,14-15,17-18,94,120,
Ludwig, Kirk 94 n. 35, 140-146,186 n. 31 139-164,181-188
bad 153
Mahon, James Edwin 166 n. 3 incompetent 153
Makinson, David 81 n. 26 negligent 153
Mathiesen, Kay 57-59, 70 rogue 151-153
McDowell, John 82 n. 28,163 n. 32 Staffel, Julia 173 n. 19
McGrath, Matthew 94 n. 34,117,119 Stalnaker, Robert 25 n. 12,144 n. 9
McMahon, Christopher 34 Stanley, Jason 94 n. 35,117,119
Meijers, A.W.M. 34,52 n. 35
Miller, Brian 107 n. 52 Thomson, Mark 69 n. 15
Miller, Seumas 107 n. 52,186 n. 31 Tollefsen, Deborah 22 n. 8,24 n. 10,121 n. 18,
Mokyr, Joel 72 n. 19 127,155-156,158 n. 24
Trump Administration 19
National Semiconductor 2-3,15-17 Tuomela, Raimo 24 n. 10,26-28, 64 n. 13,
Nozick, Robert 82 n. 28 72 n. 19,179 n. 28
200 INDEX
Unger, Peter 159 n. 25 Welbourne, Michael 163 n. 32

Wigmore, John H. 68
van Fraassen, Bas C. 25 n. 12,144 n. 9 Williams, Bernard 166 n. 3
van Hees, Martin 28 n. 17,72 n. 18 Williams, Michael 82 n. 28
Velasquez, Manuel 2 n. 6,122 n. 21 Williamson, Timothy 94 n. 35,117,163 n. 32
Volkswagen of America 1-3,15-17 Wray, K. Brad 25 n. 12,34,52 n. 35,144 n. 9

Jennifer Lackey - The Epistemology of Groups-Oxford University Press (2021)

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Jennifer Lackey - The Epistemology of Groups-Oxford University Press (2021)

Uploaded by

Copyright:

Available Formats

OXFORD

the epistemology of group

3.5 Collective Knowledge 128

My greatest debt of all is to my husband, Baron Reed—my most brilliant

ethics.” National Semiconductor prevailed: no individual employee was ever

7 See Pettit (2003).

0.1 On the Very Existence of Group Beliefs

One fundamental type of objection I have frequently heard to a project on

0.2 The Nature of Groups

In contrast, those who are interested in more permissive accounts of

10 See Bird (2010).

Collections of individuals come in many forms. Some change identity

11 See also Lahroodi (2007) and Bird (2010).

in membership,13 yet there may be significant asymmetries between this

incapable of engaging in collective reasoning and was therefore non-

14 Of course, being deliberative can also go out of existence.

0.3 Chapter Overviews

Chapter 2. The first, developed and defended in most detail by Alexander

0.4 The Bigger Picture

16 Lackey (2018b) is the only exception.

explicitly designed to produce fraudulent test results in 11 million diesel

1 A discussion of what distinguishes a group from a mere collection of individuals lies

1.1 Summative and Non-Summative

Let’s begin with the traditional summative account, according to which a

LSA: A group G believes that p if and only if some of the members of G

On this version of the summative view, it may be appropriate to attribute

philosophy department: The philosophy department at a leading

admission to its graduate program. After hours of discussion, all of the

but because collectively they regard these positions as increasing their

philosophy departmenT2: The same philosophy department that is

Despite the unanimity of individual belief in such a case, it is argued that it

These problems have motivated the now widely accepted non-summative

JAA: A group G believes that p if and only if the members of G jointly

It should be understood that: (1) Joint acceptance of a proposition p by a

Thus, according to Gilberts non-summative view, so long as a group jointly

medical association: The Board of Directors of the American

As it stands, the JAA cannot countenance group belief in medical asso­

JAA2: G believes that p in the social and normative circumstances C if and

Tuomela’s account of group belief is quite similar to Gilbert’s, though

factory: The employees of a factory are deciding whether to forgo a

16 See also Tuomela (1993 and 1995).

Serious danger? Effective measure? Bearable loss? Pay sacrifice?

1.2 Group Lies and Group Bullshit

To begin, I take the following to be a paradigmatic group lie:

tobacco company: Philip Morris, one of the largest tobacco companies

Since it is not my purpose in this chapter to provide an account of lying, I

belief—that is, when group belief is understood non-summatively on a judgment aggregation

Group Lie Desideratum: An adequate account of group belief should have

The scenario in oil company is a classic instance of what we may call

It is impossible for someone to lie unless he thinks he knows the truth.

Whereas the group in tobacco company believes that smoking is highly

operative members of BP—namely, the executive management team—jointly

Group Bullshit Desideratum: An adequate account of group belief should

26 See Wray (2003) and McMahon (2003), respectively.

JUDGMENT AGGREGATION TOBACCO COMPANY (or JA-TOBACCO

No lung cancer? No emphysema? No heart disease? Safe to health?

Following Pettit, one way of determining the group’s belief injA-TOBACCO

publish in their advertising materials that smoking is safe. Moreover, assume

1.3 Judgment Fragility

history department: The History Department at a leading university

In history department, the members of the History Department spent

JUDGMENT AGGREGATION HISTORY DEPARTMENT (or JA-HISTORY

I want to take a step back to discuss a feature that the phenomenon of

As it stands, the JAA cannot countenance group belief in medical asso

English department: The English Department at a leading univer

According to inflationary non-summativism, then, a group justifiedly believ

ignoring evidence: Philip Morris is one of the largest tobacco com

Given the serious problems facing an inflationary approach to understand

On Goldmans view, the justificational statuses of members’ doxastic atti

The contradiction found in the Group Justification Paradox is then gener