Professional Documents
Culture Documents
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
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
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
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.
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.
6 THE EPISTEMOLOGY OF GROUPS
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
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
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:
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
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.
10 THE EPISTEMOLOGY OF GROUPS
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.
12 THE EPISTEMOLOGY OF GROUPS
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.
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
14 THE EPISTEMOLOGY OF GROUPS
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.
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
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
18 THE EPISTEMOLOGY OF GROUPS
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
The Epistemology of Groups. Jennifer Lackey, Oxford University Press (2021). © Jennifer Lackey.
DOI: 10.1093/oso/9780199656608.003.0002
20 THE EPISTEMOLOGY OF GROUPS
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.
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:
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
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
24 THE EPISTEMOLOGY OF GROUPS
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:
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
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.
26 THE EPISTEMOLOGY OF GROUPS
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:
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)
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.
28 THE EPISTEMOLOGY OF GROUPS
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:
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
30 THE EPISTEMOLOGY OF GROUPS
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.
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
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:
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.
32 THE EPISTEMOLOGY OF GROUPS
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.
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
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:
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).
38 THE EPISTEMOLOGY OF GROUPS
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.
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:
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.
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
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.
44 THE EPISTEMOLOGY OF GROUPS
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:
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.
48 THE EPISTEMOLOGY OF GROUPS
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.
50 THE EPISTEMOLOGY OF GROUPS
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
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
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?
The Epistemology of Groups. Jennifer Lackey, Oxford University Press (2021). © Jennifer Lackey.
DOI: 10.1093/OSO/9780199656608.003.0003
56 THE EPISTEMOLOGY OF GROUPS
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 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
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,
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)
3 See Levi (1962), Fallis (2006), Riggs (2008), and Mathiesen (2011).
WHAT IS JUSTIFIED GROUP BELIEF? 59
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:
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.
60 THE EPISTEMOLOGY OF GROUPS
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
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.
WHAT IS JUSTIFIED GROUP BELIEF? 61
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:
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.
62 THE EPISTEMOLOGY OF GROUPS
11 For the sake of ease of expression, I will often drop the “openly.”
12 See, for instance, Tuomela (2004).
64 THE EPISTEMOLOGY OF GROUPS
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:
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).
WHAT IS JUSTIFIED GROUP BELIEF? 65
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
66 THE EPISTEMOLOGY OF GROUPS
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
WHAT IS JUSTIFIED GROUP BELIEF? 67
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.
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.
70 THE EPISTEMOLOGY OF GROUPS
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
WHAT IS JUSTIFIED GROUP BELIEF? 71
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.
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.
72 THE EPISTEMOLOGY OF GROUPS
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.
WHAT IS JUSTIFIED GROUP BELIEF? 73
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:
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:
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.
76 THE EPISTEMOLOGY OF GROUPS
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)
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.
78 THE EPISTEMOLOGY OF GROUPS
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).
WHAT IS JUSTIFIED GROUP BELIEF? 79
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)
-(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.)
80 THE EPISTEMOLOGY OF GROUPS
(AvBvCvDvE)
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.
82 THE EPISTEMOLOGY OF GROUPS
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.
WHAT IS JUSTIFIED GROUP BELIEF? 83
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.
84 THE EPISTEMOLOGY OF GROUPS
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.
WHAT IS JUSTIFIED GROUP BELIEF? 85
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
WHAT IS JUSTIFIED GROUP BELIEF? 87
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
88 THE EPISTEMOLOGY OF GROUPS
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.
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:
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.
31 See List (2005). 32 For an extended discussion of this, see List (2005).
92 THE EPISTEMOLOGY OF GROUPS
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
94 THE EPISTEMOLOGY OF GROUPS
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.
WHAT IS JUSTIFIED GROUP BELIEF? 95
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.”
98 THE EPISTEMOLOGY OF GROUPS
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
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).
100 THE EPISTEMOLOGY OF GROUPS
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.
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,
102 THE EPISTEMOLOGY OF GROUPS
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.
WHAT IS JUSTIFIED GROUP BELIEF? 103
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.
WHAT IS JUSTIFIED GROUP BELIEF? 105
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
106 THE EPISTEMOLOGY OF GROUPS
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?
WHAT IS JUSTIFIED GROUP BELIEF? 107
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.
108 THE EPISTEMOLOGY OF GROUPS
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
WHAT IS JUSTIFIED GROUP BELIEF? 109
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,
110 THE EPISTEMOLOGY OF GROUPS
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.
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
The Epistemology of Groups. Jennifer Lackey, Oxford University Press (2021). © Jennifer Lackey.
DOI: 10.1093/OSO/9780199656608.003.0004
112 THE EPISTEMOLOGY OF GROUPS
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:
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)
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.
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).
116 THE EPISTEMOLOGY OF GROUPS
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).
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).
118 THE EPISTEMOLOGY OF GROUPS
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
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
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
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.:
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.
124 THE EPISTEMOLOGY OF GROUPS
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
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
126 THE EPISTEMOLOGY OF GROUPS
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
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.
128 THE EPISTEMOLOGY OF GROUPS
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
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
132 THE EPISTEMOLOGY OF GROUPS
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
134 THE EPISTEMOLOGY OF GROUPS
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
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:
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.
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.
140 THE EPISTEMOLOGY OF GROUPS
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
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.
144 THE EPISTEMOLOGY OF GROUPS
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.
146 THE EPISTEMOLOGY OF GROUPS
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
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.
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
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.
150 THE EPISTEMOLOGY OF GROUPS
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
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
154 THE EPISTEMOLOGY OF GROUPS
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:
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
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).
160 THE EPISTEMOLOGY OF GROUPS
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
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.
162 THE EPISTEMOLOGY OF GROUPS
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
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
166 THE EPISTEMOLOGY OF GROUPS
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
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.
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).
168 THE EPISTEMOLOGY OF GROUPS
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:
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.
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
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.
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:
15 Carson (2010) is interested in both lying and deception, but is clear that he regards the
latter as not necessary for the former.
172 THE EPISTEMOLOGY OF GROUPS
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.)
174 THE EPISTEMOLOGY OF GROUPS
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
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.
176 THE EPISTEMOLOGY OF GROUPS
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:
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
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:
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
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.
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:
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
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
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* 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:
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.
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.
182 THE EPISTEMOLOGY OF GROUPS
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.
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?”:
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.
184 THE EPISTEMOLOGY OF GROUPS
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
186 THE EPISTEMOLOGY OF GROUPS
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.
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
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
188 THE EPISTEMOLOGY OF GROUPS
5.9 Conclusion
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
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
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
Schmid, Daniel Sirtes, and Marcel Weber (eds.), Collective Epistemology.
Frankfurt: Ontos Verlag, 75-117.
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
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.