Normativity : epistemic and practical

First Edition. Edition Mchugh

Visit to download the full and correct content document:
https://ebookmass.com/product/normativity-epistemic-and-practical-first-edition-editio
n-mchugh/
Normativity: Epistemic and Practical
Normativity:
Epistemic and
Practical

EDITED BY

Conor McHugh, Jonathan Way,

and Daniel Whiting

1
3
Great Clarendon Street, Oxford, OX2 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
# the several contributors 2018
The moral rights of the authors have been asserted
First Edition published in 2018
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: 2017954189
ISBN 978–0–19–875870–9
Printed and bound by
CPI Group (UK) Ltd, Croydon, CR0 4YY
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.
Contents

Acknowledgements vii
List of Contributors ix

Introduction 1
Conor McHugh, Jonathan Way, and Daniel Whiting
1. Putting Fallibilism to Work 12
Charity Anderson
2. Pragmatic Approaches to Belief 26
Jessica Brown
3. The Relevance of the Wrong Kind of Reasons 47
Ulrike Heuer
4. Directives for Knowledge and Belief 68
David Hunter
5. How Reasons Are Sensitive to Available Evidence 90
Benjamin Kiesewetter
6. Evidence and Its Limits 115
Clayton Littlejohn
7. The Explanatory Problem for Cognitivism about Practical Reason 137
Errol Lord
8. Pragmatic Encroachment: Its Problems Are Your Problems! 162
Matthew McGrath
9. Why Only Evidential Considerations Can Justify Belief 179
Kate Nolfi
10. Practical Interests and Reasons for Belief 200
Baron Reed
11. Two Theses about the Distinctness of Practical and Theoretical
Normativity 221
Andrew Reisner
vi CONTENTS

12. Reasoning with Reasons 241

Daniel Star
13. Epistemic Instrumentalism, Permissibility, and Reasons for Belief 260
Asbjørn Steglich-Petersen

Index 281
Acknowledgements

This volume emerges from a two-year research project, Normativity: Epistemic

and Practical, based at the University of Southampton. We are very grateful to
the Arts and Humanities Research Council for funding this project (grant
number AH/K008188/1). We are also grateful to our colleagues in Philosophy
at Southampton for supporting it in many different ways, including, but not
limited to, participating in the project’s various activities. Most of the contri-
butions to this volume are descendants or relatives of papers presented in the
seminars, workshops, and conferences accompanying the project. Thanks to all
those who took part in these events, not least the contributors themselves.
Finally, thanks of different sorts to Oxford University Press, especially Peter
Momtchiloff, for encouraging and overseeing the publication of this volume, and
to Sophie Keeling for preparing the index.
List of Contributors

C HARITY A NDERSON , Baylor University, Texas

J ESSICA B ROWN , University of St Andrews
U LRIKE H EUER , University College London
D AVID H UNTER , Ryerson University, Ontario

B ENJAMIN K IESEWETTER , Humboldt University of Berlin

C LAYTON L ITTLEJOHN , King’s College London
E RROL L ORD , University of Pennsylvania
M ATTHEW M C G RATH , University of Missouri and University of St Andrews
C ONOR M C H UGH , University of Southampton
K ATE N OLFI , University of Vermont
B ARON R EED , Northwestern University, Illinois
A NDREW R EISNER , Uppsala University
D ANIEL S TAR , Boston University, Massachusetts
A SBJØRN S TEGLICH -P ETERSEN , Aarhus University
J ONATHAN W AY , University of Southampton
D ANIEL W HITING , University of Southampton
 Introduction
Conor McHugh, Jonathan Way, and Daniel Whiting

1 Introduction
What should I do? What should I think? Traditionally, ethicists tackle the first
question, while epistemologists tackle the second. This division of labour corres-
ponds to a distinction theorists draw between practical and epistemic normativ-
ity, where normativity is a matter of what one should or may do or think, what
one has reason or justification to do or to think, what it is right or wrong to do or
to think, and so on, not simply of what one in fact does or thinks.1
The tendency is to investigate the issue of what to do independently of the
issue of what to think, that is, to do ethics independently of epistemology, and
vice versa. But there is a growing awareness that treating the two separately leads
to distortions, omissions, and misunderstandings. Debates over epistemic nor-
mativity have failed to take on board developments on the practical side; those
discussing practical normativity often make general claims whose implications in
the epistemic case remain unexplored.
Several considerations seem to support adopting a more synoptic approach to
matters normative. First, we use paradigmatic normative terms such as ‘ought’,
‘right’, and ‘justification’ when speaking both of action and of belief: for example,
‘You ought not to smoke’ and ‘You ought not to believe everything the tobacco
companies tell you’. There is no obvious ambiguity here. Second, epistemic
norms appear to govern actions of certain sorts, for example, asserting proposi-
tions or gathering evidence, while practical norms appear to govern thinking of

1
For ease of presentation, we will speak primarily of action and belief. But, of course, practical
attitudes like intention are also subject to norms, and epistemic norms govern other attitudes, such
as credences. Feelings are also subject to norms. The connection between those norms and the
norms that govern action and belief is an interesting and important one, but not the focus of this
volume (although see McGrath’s contribution in Chapter 8 for some discussion).
 CONOR M c HUGH , JONATHAN WAY , AND DANIEL WHITING

certain sorts, for example, practical reasoning. Finally, the question of what to do
often seems closely tied to that of what to think. For example, it would be surprising
if it were to turn out that the issue of whether to give a person chocolate were wholly
independent of the issue of whether to think that she is diabetic.
The aim of this volume is to examine the norms which concern us as agents
alongside the norms which concern us as inquirers. More specifically, it is to
explore substantive and explanatory connections between practical and epistemic
norms, to consider whether these norms are at some level unified, and to ask
what that might mean. The aim of this introduction is to provide an overview of
some—by no means all—of those issues and to indicate in which contributions
they appear.

2 Substantive Connections
In this section we will consider some of the ways in which epistemic and practical
normativity might be thought to be substantively connected. A place to begin is
the suggestion that there is pragmatic encroachment on knowledge, that whether
a person knows a proposition does not depend only on, say, whether her evidence
supports that proposition, but also on broadly practical considerations, such as
what is at stake in believing it.2 To illustrate:
LOW On Friday, Karen drives past the bank. She has a cheque to deposit
though there is no urgent need to do so. Karen wonders whether she could
come back tomorrow and recalls that the bank was open on Saturday two
weeks ago. She concludes that the bank will be open tomorrow, too, and so
decides to come back then. It turns out that Karen is right—the bank is open
tomorrow.
HIGH Like LOW except that there is an urgent need for Karen to deposit the
cheque. She has a mortgage payment to make by Sunday, when the bank is
closed, and her house will be repossessed if she fails to make it.3

Many are inclined to judge that in LOW Karen knows that the bank is open. In
contrast, many are inclined to judge that in HIGH she does not know this. In
both cases, paradigmatically epistemic factors, such as the level of evidential

2
See Fantl and McGrath 2002, 2009; Hawthorne 2004; Stanley 2005.
3
LOW and HIGH are variants on the ‘Bank’ cases of DeRose 1992. The original versions of those
cases involve explicit attributions or denials of knowledge, and were intended to support, not
pragmatic encroachment, but the view that the truth-conditions of such attributions depend on
the practical context in which they are made.
 INTRODUCTION 

support, the reliability of the belief-forming mechanisms, and so on, are the same.
The only difference is that Karen has more to lose in HIGH than in LOW.
Many of those who have the above inclinations are also inclined to say that in
LOW Karen is justified in believing that the bank is open, while in HIGH she is
not, at least, not fully or flat-out.4
Such cases, then, lend intuitive support to the idea that practical considerations
bear on whether a person knows or is justified in believing a proposition.
(McGrath defends the idea in Chapter 8, while Anderson and Reed oppose it
(Chapters 1 and 10, respectively); see also Brown (Chapter 2).)5 Lying behind
those intuitions might be a linking principle like the following:
If a person is justified in believing that p, she is justified in acting on the basis
that p.
Of course, this is just a first pass. The idea it seeks to capture is that, if a belief is
not an appropriate basis for action, if it is not appropriate to act on that belief ’s
content, it is not appropriate to have it.6 This is to say that there is a substantive
connection between the norms governing action and those governing belief. (For
discussion of this idea, independent of the debate over pragmatic encroachment,
see Littlejohn’s Chapter 6 and Star’s Chapter 12.)
If a principle like this holds, it explains the verdicts in HIGH and LOW. It is
appropriate for Karen in LOW to use the content of her belief that the bank is
open as a premise in her practical reasoning as to whether to come back
tomorrow. But it is not appropriate for Karen in HIGH to do the same. Hence,
in HIGH Karen is not justified in believing that the bank is open. Hence, she does
not know this.
The proponent of pragmatic encroachment maintains that in some way
practical considerations, specifically the costs of error, bear on whether to believe
a proposition. An interesting question (which Reed discusses in Chapter 10) is
how this relates to a more traditional form of pragmatism, according to which
the costs and benefits of believing can provide reasons for and against doing
so. Suppose, for example, that a representative of an unscrupulous oil company

4
See, for example, Fantl and McGrath 2009.
5
For further critical discussion, see Anderson 2015; Brown 2008; Levin 2008; Reed 2010.
6
The authors cited in n2 advance some version of this idea (see also Hawthorne and Stanley
2008). For critical discussion, see Brown 2008; Gerken 2011; McKinnon 2011; Neta 2009; Smithies
2012; Whiting Forthcoming.
A related, but distinct, principle is that, if a person is justified in believing that she is justified in
performing some act, then she is justified in performing that act. For discussion and defence of
principles in this ballpark, see Gibbons 2013; Kiesewetter Forthcoming; Littlejohn 2012; Way and
Whiting 2016.
 CONOR M c HUGH , JONATHAN WAY , AND DANIEL WHITING

offers you a financial reward for believing that global temperatures are falling.
The offer is not evidence for the proposition. So, one might say, it is not an
epistemic reason for believing it. But the offer nonetheless seems to many to
count in favour of belief; it is, they say, a practical or pragmatic reason for
believing.7
If there are practical reasons for belief in addition to epistemic reasons, this
raises the question of how, if at all, they weigh against each other. If the practical
considerations weigh in favour of belief but the evidential considerations weigh
against it, what should a person believe overall or all-things-considered? It is
difficult to know how to answer this question. That might suggest it is a bad one
to ask. Perhaps there is something that a person should believe from an epistemic
point of view and something that she should believe from a practical point of
view, but nothing that she should believe full stop. (For discussion, see Reisner’s
Chapter 11.)
The distinction between epistemic and practical reasons for belief seems to be
an instance of a more general distinction between what are sometimes called
reasons of the right kind and reasons of the wrong kind for attitudes.8 Suppose that
the oil company rep offers you a reward for intending to drive to work. To get the
reward, you do not have to drive to work—you only have to intend to do so. The
offer is, one might think, a reason for having the intention but it is a reason of an
unorthodox (wrong) sort. To see this, suppose that driving to work is quicker
than walking. This is a reason of an orthodox (right) kind to intend to drive—it
reveals or indicates something good about what you intend. But the rep’s offer
does not in the same way reveal or indicate anything good about what you
intend; instead, it indicates something good about intending it, about having
that attitude.
This raises the issue of how exactly to distinguish reasons of the wrong kind
from reasons of the right kind. In turn, it raises the issue of whether the way in
which we draw the distinction in the case of belief carries over to the case of
intention or other practical attitudes. If so, this might reveal some important
unity between the norms governing each attitude; if not (as Heuer argues in
Chapter 3), it might cast doubt on whether there is any such unity.
The suggestion that practical considerations bear in some way on the epistemic
status of a belief seems like a challenge to evidentialism. A standard way of
characterizing this is as the view that whether a person is justified in believing

7
For an overview of this issue, and extensive references, see Reisner Forthcoming.
8
For some influential discussions of this distinction, see Hieronymi 2005; Rabinowicz and
Rnnow-Rasmussen 2004; Schroeder 2012.
 INTRODUCTION 

a proposition is determined by her evidence.9 One way to understand this is to

think that whether a person is justified in performing some act or having some
attitude is in general determined by the reasons she possesses. The reasons she
possesses for believing a proposition are provided (only) by evidential consider-
ations, that is, considerations that indicate or make likely that that proposition is
true. On this account, the question of what to believe is settled by a person’s
evidence and her evidence alone. (In Chapters 4 and 6, Hunter and Littlejohn
criticize this view in different ways.)
Evidentialists deny that practical considerations of the above sort provide
reasons for believing.10 (If they are right to do so, the problem of how to reconcile
the apparent demands of such reasons with the demands of evidential consider-
ations does not arise.) Some deny more generally that reasons of the wrong kind
are reasons at all; they hold that they are reasons only in name. In doing so, they
often appeal to a principle along the following lines:
That p is a reason for a person to φ only if she can φ for the reason that p.11
As stated, this principle applies in both the epistemic and practical domains. If
true, it might seem to rule out reasons of the wrong kind. It seems that you
cannot believe that global temperatures are falling for the reason that by so
believing you will get a reward. Nor can you intend to drive for the reason that
by so intending you will get a reward.12 (For discussion, see Heuer’s Chapter 3.)
The above is a motivational constraint on reasons. Consider also this epistemic
constraint:
That p is a reason for a person to φ only if she is in a position to know that p.13
This is a substantive principle which also applies in both the epistemic and
practical domains. According to it, only facts within a person’s ken provide

9
See Conee and Feldman 2004.
10
More carefully, some deny this. Others make the more modest claim that only evidential
considerations bear on whether a belief possesses the kind of justification necessary for knowledge,
that is, for epistemic justification.
11
Williams 1981 is an influential proponent of this principle. For discussion and further
references, see Way Forthcoming-b; Way and Whiting Forthcoming.
12
For versions of this argument, see Kelly 2002; Shah 2006, 2008. For doubts about the claims
about what we can believe and intend that it relies on, see Frankish 2007; Pink 1991; Schroeder 2012;
Sharadin 2016; Way Forthcoming-b. For defence, see Archer Forthcoming; Shah and Silverstein
2013.
13
Proponents of epistemic constraints of this sort include Dancy 2000; Gibbons 2013; Lord
2015; Raz 2011; and Kiesewetter (Chapter 5, this volume).
 CONOR M c HUGH , JONATHAN WAY , AND DANIEL WHITING

reasons for her to act or to think.14 This is to say, whether a fact is a reason for a
person depends on her epistemic status with respect to that fact.
The motivational constraint might be taken to support the epistemic con-
straint. The idea, in short, is that if a person is in no position to know a reason,
she is no position to act on it.15
An epistemic constraint of the above sort bears on the debate between object-
ivists and perspectivists about what is sometimes called the deliberative ought, the
ought in play when a person asks herself, with the aim of making a decision,
‘What ought I to do?’16 Objectivists maintain that what a person ought in this
sense to do is determined by the facts, without restriction. Perspectivists, in contrast,
maintain that what a person ought in the deliberative sense to do is determined by
her perspective. What constitutes a person’s perspective? Different answers to this
question result in different versions of perspectivism. For present purposes, suppose
that a person’s perspective is constituted by her evidence, understood as including
only what she is in a position to know. In view of this, consider:
A patient has a treatable disease. If left untreated it will lead to death. All of the
doctor’s evidence indicates that drug A will cure her patient and drug B will kill
her. In fact, drug A will kill the patient and drug B will cure her.
(cf. Jackson 1991: 462–3)
According to objectivists, the doctor in this case ought to give drug B. According
to perspectivists, the doctor ought to give drug A. On the assumption that a
person’s reasons determine what she ought to do, the epistemic constraint on
reasons accords with the perspectivist’s verdicts (defended in Chapter 5 by
Kiesewetter). Perspectivists thus give epistemic factors a crucial role in the
substantive determination of central normative facts in the practical realm.

3 Explanatory Connections
We turn now from substantive connections to explanatory ones. These include
ways that epistemic norms might be thought to be explained in terms of the
practical, or vice versa, as well as unifying explanations of both.
Suppose that evidence for a proposition justifies or provides a reason for
believing it. What might explain this? A standard answer in epistemology is

14
One might, of course, propose alternative principles by appeal to a different epistemic status
than that of being in a position to know.
15
Lord 2015 argues at length that the epistemic constraint follows from the motivational one, at
least when these are restricted to decisive reasons.
16
For some further contributions to this debate, see Graham 2010; Kiesewetter 2011; Mason
2013; Zimmerman 2014.
 INTRODUCTION 

teleological. Having true beliefs (and avoiding false beliefs) is an end. If a person
has evidence for a proposition, this indicates that, in believing it, she will realize
or secure this end. Hence, such evidence is or provides a reason for so believing.17
In what way is believing what is true an end? One—not the only—way to
understand this is as an aim that people have in forming and revising beliefs.18
This proposal represents epistemic normativity as a species of instrumental
normativity, which is often taken to be a paradigmatic form of practical norma-
tivity. If a person aims to bake a cake, instrumental rationality in some way
requires her to take the means to satisfying this aim, say, by buying the ingredi-
ents. In a similar fashion, when a person aims to believe the truth, instrumental
rationality in some way requires her to take the means to satisfying this aim, say,
by following her evidence. (Steglich-Petersen defends such an account of epi-
stemic normativity in Chapter 13.)
Another proposal takes off from the thought that beliefs provide the basis for
action or the input to practical reasoning. For example, when a person believes
that the recipe requires eggs, she might on that basis fetch some. This proposal
can also be developed in a broadly teleological way. One might think that it is the
function or purpose of belief to serve this role. In turn, one might seek to explain
why evidence provides a reason for belief by appeal to that function or purpose.
Evidence for a belief indicates that it is an accurate guide for action, hence, that it
is fit for inclusion in practical deliberation. Hence, such evidence is or provides a
reason for belief.19 It is an interesting question whether this explanation com-
petes with or complements the preceding one. (Nolfi defends a proposal of this
sort in Chapter 9, while Brown challenges the idea that belief should be under-
stood in terms of its practical role in Chapter 2.)
These are examples of how one might try to ground epistemic normativity in
practical normativity or concerns. (For more examples, see Reed’s Chapter 10.)
In this way, we might find explanatory connections between the epistemic and
practical domains. Of course, such explanatory connections might run in the
other direction. Consider again instrumental rationality. Here is a rough formu-
lation of a norm of instrumental rationality:
You must intend to ψ if you intend to φ and believe that φing is a necessary
means to ψing.

17
Foley 1987 gives a classic development of such an approach; for a more recent defence, see
Ahlstrom-Vij and Dunn 2014. Influential critics include Berker 2013 and Kelly 2003.
18
For an influential discussion of this idea, see Velleman 2000. See also McHugh 2012-a, 2012-b;
Shah and Velleman 2005; Steglich-Petersen 2006.
19
For other ways of developing this idea, see Côté-Bouchard 2015; Whiting 2014.
 CONOR M c HUGH , JONATHAN WAY , AND DANIEL WHITING

Suppose that intending to do something involves believing that you will do it. In
that case, the above norm might seem to follow from a requirement of epistemic
rationality:
You must believe that you will ψ if you believe that you will φ and believe that
you will φ only if you will ψ.
So, given the ‘cognitivist’ assumption that intentions involve beliefs, one might
try to ground a certain sort of practical norm in a certain sort of epistemic
norm.20 (For criticism of this approach, see Lord’s Chapter 7.)
An alternative way to ground the practical in the epistemic is to offer an
account of some normative property, such as that of being a reason, which
appeals to an epistemic property. Consider the proposal that a reason for a
person to φ is evidence that she ought to φ.21 For example, that a person is in
pain is evidence that you ought to help her. Hence, on this view, it is a reason for
doing so. This account (which Star defends in Chapter 12) explains practical
reasons in terms of a central epistemic property, that of evidence.
It is arguably an attractive feature of this view that it offers a unified account
of reasons, one that applies to reasons for belief as well as reasons for action. Reasons
of all kinds are evidence; epistemic reasons are evidence that a person ought to
believe a proposition while practical reasons are evidence that a person ought to act.
Of course, it is not the only view of this unifying sort on the market. Consider
the view that reasons are facts that explain why a person ought to φ, where φing
might be acting or believing.22 Or consider the view that reasons are premises
of good reasoning, whether to a theoretical conclusion, a belief, or a practical
conclusion, an intention or action.23 The teleological proposal noted above
might also be understood in this way: reasons to φ—whether φing is believing
or acting—are considerations which help explain why φing promotes one’s
aims (desires, wants, etc.).24, 25
Despite their differences, these accounts of what it is to be a reason assume that
a unified explanation—that is, one that applies to both reasons for belief and
reasons for action—is possible. Such an account is attractive, insofar as it

20
For an early statement of this kind of view, see Harman 1976. For more recent defences, see
Setiya 2007; Velleman 1989; Wallace 2006. For criticism, see Bratman 1999: ch. 13; Brunero 2009.
21 22
See Kearns and Star 2009; Thomson 2008. For this view, see Broome 2013.
23
For versions of this view, see Setiya 2014; Way Forthcoming-a.
24
Schroeder 2007 expresses sympathy for this view. See also Finlay 2006.
25
Unlike some of the views discussed earlier in this section, these views are typically presented as
analyses of reasons—that is, accounts of what it is to be a reason—rather than simply as accounts of
what grounds reasons or of what explains reasons. For a recent discussion of the relationship
between analyses and grounds, see Rosen 2015.
 INTRODUCTION 

promises to explain the similarities between reasons for belief and action. None-
theless, one might, of course, doubt that any such account is possible. Perhaps
reasons for belief are just very different beasts from reasons for action (as Heuer
suggests in Chapter 3). Perhaps, in turn, oughts as they apply to belief are very
different beasts from oughts as they apply to action. More generally, perhaps the
norms governing belief are entirely independent of the norms governing action.
On these views, the domain of the normative is fragmented or, rather, there is not
one such domain but a number of them. Such views would also have their
attractions—for instance, they might explain why it is hard to make sense of
weighing practical reasons against epistemic reasons. (For discussion, see also
Reisner’s Chapter 11.)

4 Conclusion
The above provides an overview of some of the themes which the contributions to
this collection explore. We make no pretence that the overview, or for that matter
the collection itself, is exhaustive. The question of how practical norms and
epistemic norms relate raises many more issues than any one volume can cover.
The hope is that the collection as a whole demonstrates the importance and interest
of asking that question and the many lines of inquiry that lead from it.

References
Ahlstrom-Vij, K. and Dunn, J. 2014: A Defence of Epistemic Consequentialism. Philo-
sophical Quarterly 64: 541–51.
Anderson, C. 2015: On the Intimate Relationship of Knowledge and Action. Episteme 12:
343–53.
Archer, S. Forthcoming: Defending Exclusivity. Philosophy and Phenomenological
Research.
Berker, S. 2013: The Rejection of Epistemic Consequentialism. Philosophical Issues 23:
363–87.
Bratman, M. 1999: Faces of Intention. Cambridge: Cambridge University Press.
Broome, J. 2013: Rationality through Reasoning. Oxford: Wiley-Blackwell.
Brown, J. 2008: Subject-Sensitive Invariantism and the Knowledge Norm for Practical
Reasoning. Noûs 42: 167–89.
Brunero, J. 2009: Against Cognitivism about Practical Rationality. Philosophical Studies
146: 311–25.
Conee, E. and Feldman, R. 2004: Evidentialism. Oxford: Oxford University Press.
Côté-Bouchard, C. 2015: Epistemic Instrumentalism and the Too Few Reasons Objection.
International Journal of Philosophical Studies 23: 337–55.
Dancy, J. 2000: Practical Reality. Oxford: Oxford University Press.
 CONOR M c HUGH , JONATHAN WAY , AND DANIEL WHITING

DeRose, K. 1992: Contextualism and Knowledge Attributions. Philosophy and Phenom-

enological Research 52: 913–29.
Fantl, J. and McGrath, M. 2002: Evidence, Pragmatics, and Justification. Philosophical
Review 111: 67–94.
Fantl, J. and McGrath, M. 2009: Knowledge in an Uncertain World. Oxford: Oxford
University Press.
Finlay, S. 2006: The Reasons that Matter. Australasian Journal of Philosophy 84: 1–20.
Foley, R. 1987: The Theory of Epistemic Rationality. Cambridge, MA: Harvard University
Press.
Frankish, K. 2007: Deciding to Believe Again. Mind 116: 527–47.
Gerken, M. 2011: Warrant and Action. Synthese 178: 529–47.
Gibbons, J. 2013: The Norm of Belief. Oxford: Oxford University Press.
Graham, P. 2010: In Defence of Objectivism about Moral Obligation. Ethics 121: 88–115.
Harman, G. 1976: Practical Reasoning. Review of Metaphysics 29: 431–63.
Hawthorne, J. 2004: Knowledge and Lotteries. Oxford: Oxford University Press.
Hawthorne, J. and Stanley, J. 2008: Knowledge and Action. Journal of Philosophy 105:
571–90.
Hieronymi, P. 2005: The Wrong Kind of Reason. Journal of Philosophy 102: 437–57.
Jackson, F. 1991: Decision-Theoretic Consequentialism and the Nearest and Dearest
Objection. Ethics 101: 461–82.
Kearns, S. and Star, D. 2009: Reasons as Evidence. Oxford Studies in Metaethics: Volume 4,
ed. R. Shafer-Landau. Oxford: Oxford University Press.
Kelly, T. 2002: The Rationality of Belief and Some Other Propositional Attitudes. Philo-
sophical Studies 110: 163–96.
Kelly, T. 2003: Epistemic Rationality as Instrumental Rationality: A Critique. Philosophy
and Phenomenological Research 66: 612–40.
Kiesewetter, B. 2011: ‘Ought’ and the Perspective of the Agent. Journal of Ethics and Social
Philosophy 5: 1–24.
Kiesewetter, B. Forthcoming: The Normativity of Rationality. Oxford: Oxford University
Press.
Levin, J. 2008: Assertion, Practical Reason, and Pragmatic Theories of Knowledge.
Philosophy and Phenomenological Research 76: 359–84.
Littlejohn, C. 2012: Justification and the Truth-Connection. Cambridge: Cambridge
University Press.
Lord, E. 2015: Acting for the Right Reasons, Abilities, and Obligation. Oxford Studies in
Metaethics: Volume 10, ed. R. Shafer-Landau. Oxford: Oxford University Press.
McHugh, C. 2012-a: Belief and Aims. Philosophical Studies 160: 425–39.
McHugh, C. 2012-b: Control of Belief and Intention. Thought 1: 337–46.
McKinnon, R. 2011: Lotteries, Knowledge, and Practical Reasoning. Logos and Episteme 2:
225–31.
Mason, E. 2013: Objectivism and Prospectivism about Rightness. Journal of Ethics and
Social Philosophy 7: 1–21.
Neta, R. 2009: Treating Something as a Reason for Action. Noûs 43: 684–99.
Pink, T. 1991: Purposive Intending. Mind 99: 343–59.
 INTRODUCTION 

Rabinowicz, W. and Rnnow-Rasmussen, T. 2004: The Strike of the Demon: On Fitting

Pro-Attitudes and Value. Ethics 114: 391–423.
Raz, J. 2011: From Normativity to Responsibility. Oxford: Oxford University Press.
Reed, B. 2010: In Defense of Stable Invariantism. Noûs 44: 224–44.
Reisner, A. Forthcoming: Pragmatic Reasons for Belief. The Oxford Handbook of Reasons
and Normativity, ed. D. Star. Oxford: Oxford University Press.
Rosen, G. 2015: Real Definition. Analytic Philosophy 56: 189–209.
Schroeder, M. 2007: Slaves of the Passions. Oxford: Oxford University Press.
Schroeder, M. 2012: The Ubiquity of State-Given Reasons. Ethics 122: 457–88.
Setiya, K. 2007: Cognitivism about Instrumental Reason. Ethics 117: 649–73.
Setiya, K. 2014: What is a Reason to Act? Philosophical Studies 167: 221–35.
Shah, N. 2006: A New Argument for Evidentialism. Philosophical Quarterly 56: 481–98.
Shah, N. 2008: How Action Governs Intention. Philosophers’ Imprint 8: 1–19.
Shah, N. and Silverstein, M. 2013: Reasoning in Stages. Ethics 124: 101–13.
Shah, N. and Velleman, J. D. 2005: Doxastic Deliberation. Philosophical Review 114:
497–534.
Sharadin, N. 2016: Nothing but the Evidential Considerations? Australasian Journal of
Philosophy 94: 343–61.
Smithies, D. 2012: The Normative Role of Knowledge. Noûs 46: 265–88.
Stanley, J. 2005: Knowledge and Practical Interests. Oxford: Oxford University Press.
Steglich-Petersen, A. 2006: No Norm Needed: On the Aim of Belief. Philosophical
Quarterly 56: 499–516.
Thomson, J. J. 2008: Normativity. Peru, IL: Open Court.
Velleman, J. D. 1989: Practical Reflection. Princeton, NJ: Princeton University Press.
Velleman, J. D. 2000: The Possibility of Practical Reason. Oxford: Oxford University Press.
Wallace, R. J. 2006: Normativity and the Will. Oxford: Oxford University Press.
Way, J. Forthcoming-a: Reasons as Premises of Good Reasoning. Pacific Philosophical
Quarterly.
Way, J. Forthcoming-b: Two Arguments for Evidentialism. Philosophical Quarterly.
Way, J. and Whiting, D. 2016: If You Justifiably Believe that You Ought to Φ, You Ought
to Φ. Philosophical Studies 173: 1873–95.
Way, J. and Whiting, D. Forthcoming: Reasons and Guidance (or, Surprise Parties and Ice
Cream). Analytic Philosophy.
Whiting, D. 2014: Reasons for Belief, Reasons for Action, the Aim of Belief, and the Aim
of Action. Epistemic Norms, ed. C. Littlejohn and J. Turri. Oxford: Oxford University
Press.
Whiting, D. Forthcoming: Against Second-Order Reasons. Noûs.
Williams, B. 1981: Internal and External Reasons. Moral Luck. Cambridge: Cambridge
University Press.
Zimmerman, M. 2014: Ignorance and Moral Obligation. Oxford: Oxford University Press.
 1
Putting Fallibilism to Work
Charity Anderson

1 Introduction
A connection between knowledge and reasons for action is defended in recent
literature as follows: when one knows p one is in a good enough epistemic
position to treat p as a reason for action (hereafter, Sufficiency).1 Sufficiency—
or some nuanced version of it—is used to motivate pragmatic encroachment: the
view that pragmatic factors are relevant to whether a subject has knowledge.2
When combined with fallibilism—the widely accepted idea that knowledge is
compatible with an epistemic chance of error—Sufficiency results in a rather
counterintuitive picture.3 Namely, it results in the rejection of purism, the view
that pragmatic factors are irrelevant to knowledge.4 Fallibilism, purism, and
Sufficiency each have substantial prima facie intuitive support; and yet the three
seem to form an inconsistent triad.
To see why these three have been thought to form a trilemma, consider two
agents with the same rather strong epistemic position for a proposition. Suppose
the practical details are such that one agent ought to act, but the other ought not

1
A knowledge-reasons connection has also been defended in terms of a necessary principle:
one is appropriate to treat p as a reason for action only if one knows p. This chapter is neutral
with respect to this principle.
2
See Fantl and McGrath (2002, 2007, 2009), Hawthorne (2004), Stanley (2005), and Hawthorne
and Stanley (2008).
3
Fallibilism is a technical term and although there is no agreed-upon definition, there is a core
idea that is often expressed in terms of the compatibility of knowledge with a chance of error. In
Section 3, I offer a gloss on the notion of epistemic chance of error that provides a way to understand
fallibilism while also maintaining a knowledge-first framework.
4
Also known as intellectualism, this view is sometimes characterized as the idea that two subjects
with the same strength of epistemic position for p are in the same position to know p. It is not
entirely clear which factors count as ‘epistemic’ and which do not, but one point of agreement that
tends to guide discussion is that practical factors, such as the cost of being wrong about p, are not
epistemic.
 PUTTING FALLIBILISM TO WORK 

to act, for some action (often this is done by positing that one agent is in a ‘high-
stakes’ setting and the other in a ‘low-stakes’ setting). Assuming fallibilism, it’s
natural to think the agent in the ‘low-stakes’ setting knows. Given this assump-
tion, we can see why Sufficiency and purism cannot both be affirmed: if Suffi-
ciency is true, then the agent that ought not act does not know the relevant
proposition. Since the difference between the agent that knows and the agent that
doesn’t know is due to some non-epistemic factor (the agent’s practical situ-
ation), purism is violated. (If we instead start by holding purism fixed, we get the
result that Sufficiency is violated.)5
For better or worse (in my opinion, for better), fallibilism has earned the status
of an indispensable commitment of contemporary epistemology. From this
vantage point, the problem could be stated as a dilemma for fallibilists: fallibilists
must choose between purism and Sufficiency.6 One goal of this chapter is to show
that there is a fallibilist option that can avoid the trilemma.
My primary goal, though, is to argue against pragmatic encroachment by
arguing against Sufficiency. The structure of this chapter is as follows: in
Section 2, I challenge the account of reasons that underlies one prominent way
of arguing for Sufficiency and then propose an alternative account of the rela-
tionship between knowledge and reasons for action. Section 3 examines the
trilemma between purism, fallibilism, and Sufficiency and delineates a position
that can maintain all three. Thus, there is a way out of the trilemma. I conclude
with a final consideration in support of my preferred resolution of the trilemma:
rejection of Sufficiency.

2 Against Safe Reasons and KJ

2.1 Unpacking the principles
The knowledge-action links have been formulated in a variety of nuanced ways.
I will focus here on the following conception of Sufficiency, defended at length by
Jeremy Fantl and Matthew McGrath (2009):
(KJ) If you know that p, then p is warranted enough to justify you in ϕ-ing,
for any ϕ.7

5
See Fantl and McGrath (2009: 84–6) for a detailed explanation of the trilemma.
6
Fantl and McGrath (2009: last chapter) discuss infallibilism as a tenable option, though they do
not go as far as to recommend that we reject fallibilism; on the contrary, they seem to strongly favor
maintaining fallibilism.
7
Alternative wordings of KJ replace “p is warranted enough to justify you in ϕ-ing, for any ϕ”
with other phrases, such as “one is in a good enough epistemic position to rely on p in practical
 CHARITY ANDERSON

The central argument Fantl and McGrath offer to motivate KJ involves a

framework that connects knowledge and rational action via reasons. Their case
for KJ is based on the following two principles about reasons:
(KR) If you know that p, then p is warranted enough to be a reason you have
to ϕ, for any ϕ. (2009: 69)
(Safe Reasons) If p is a reason you have to ϕ, then p is warranted enough to
justify you in ϕ-ing, for any ϕ. (2009: 77)8
In what follows I will argue that Safe Reasons is false. If my argument succeeds,
one motivation for KJ is undermined. The same line of reasoning I offer against
Safe Reasons can be advanced against KJ directly, showing KJ to be false.
A few preliminary points are in order. First, there are multiple ways a prop-
osition could fail to justify you in ϕ-ing; epistemic propriety is just one dimension
along which we can evaluate reasons for action. These principles speak only to
the epistemic propriety of p for justified action. To fix on how epistemic propriety
comes apart from other kinds of propriety, consider a setting where free dough-
nuts are being given out just outside your office, but you have no idea that this
is the case. You fail to be in the right kind of epistemic position towards the
proposition there are free doughnuts outside your office that is required for you to
be justified in treating this proposition as your reason for stepping outside.
Alternatively, a proposition could fail to justify you in ϕ-ing because, despite
having strong warrant for the proposition, the proposition fails to be connected
in the right way to ϕ-ing. For example, even if I have top-grade warrant for the
proposition I exist, this proposition does not justify me in walking my dog
because it fails to be connected in the right way to the action. Nevertheless, the
proposition I exist is warranted enough to justify me in walking my dog. In this
case, something other than epistemic propriety prevents I exist from justifying
me in walking my dog.

reasoning” (Brown 2012: 155) and “it is appropriate to treat p as a reason for acting” (Hawthorne
and Stanley 2008: 578). The differences in wording will not be significant for my purposes: the
difficulties I raise for KJ can be extended to these versions of the principle. In Section 3, I discuss a
way of understanding Hawthorne and Stanley’s principle that does not equate it with KJ.
8
The argument depends on a further connecting principle:
(Connecting Principle): If p is warranted enough to be a reason you have for ϕ-ing,
then p is warranted enough to justify you in ϕ-ing, for any ϕ.
I will focus my discussion on Safe Reasons rather than the Connecting Principle for two reasons:
first, it is the principle that plays a central role in Fantl and McGrath’s discussion; and second, the
locution of a ‘reason one has’ maps onto ordinary language more easily than the ‘warranted enough
to be a reason one has.’ But note that the argument I level against Safe Reasons is also a
counterexample to the Connecting Principle.
 PUTTING FALLIBILISM TO WORK 

The discussion in this chapter concerns when weakness in one’s epistemic

position makes it inappropriate to treat a proposition as a reason for action. In
Fantl and McGrath’s preferred terminology, these are situations where weakness
in epistemic position ‘stands in the way’ of a proposition justifying action.9 To
identify when weakness in epistemic position stands in the way of a proposition
justifying action, I introduce the following test:
Strengthening test: what S ought to do is the same as what S ought to do
conditional on p.
When what one ought to do differs from what one ought to do conditional on
some proposition, one fails the strengthening test for that proposition. When one
fails the test, epistemic weakness stands in the way of appropriately treating that
proposition as a reason to act.10 The strengthening test is naturally applied in a
decision theory framework, where what one ought to do is the top ranked action
on an ordering given by one’s epistemic position and utilities.11 What one ought
to do conditional on some proposition is the action that ranks top conditional on
that proposition.12
The strengthening test plays an important role in the principles above. Passing
the strengthening test is a necessary condition on a proposition being a reason
one has to ϕ (by Safe Reasons). In the same way, KJ makes passing the strength-
ening test a necessary condition on knowledge. I will argue against both of these
principles. My discussion leaves KR untouched as a plausible connection between
knowledge and reasons for action. The picture I am inclined towards is one
whereby when one knows a proposition the proposition is a reason one has to act,
but where being a reason one has does not entail that one is appropriate to treat

9
See Fantl and McGrath (2009: 67–8).
10
Passing the strengthening test is not a sufficient condition for one’s epistemic position to be
warranted enough to justify action because one might pass the test even though one’s epistemic
position is very weak. Plausibly there is some minimal threshold required, in addition to passing the
strengthening test, in order for p to be warranted enough to justify action.
11
The notion ‘epistemic position’ is flexible with respect to various interpretations; as I use it
here, it is subject to a few constraints: first, it excludes non-epistemic features of one’s situation;
second, it is externalist—one’s epistemic position does not supervene on one’s internal mental state.
Third, one’s epistemic position is not to be construed as including one’s total knowledge, as on such
a picture one will automatically pass the strengthening test for each proposition known. See
Section 3 for further discussion and for a relaxing of this last constraint.
12
Some advocates of the knowledge-action principles may object to a test that relies on decision
theory. For my purposes here, the helpfulness of the test relies on only basic and relatively harmless
features of decision theory. The strengthening test bears strong similarity to the slogan test that Fantl
and McGrath offer (2009: 68), which is as follows: “if merely strengthening your epistemic position
can make a difference as to whether p justifies you in ϕ-ing, then weaknesses in your epistemic
position stand in the way of its so justifying you.”
 CHARITY ANDERSON

the proposition as a reason to act—thus, being a reason one has to act does not
require passing the strengthening test.13 Our epistemic position for reasons we
have is not always strong enough to pass the strengthening test. (This should be
unsurprising, given fallibilism.)14
Finally, we need to make it clear that, according to Safe Reasons, reasons you
have are able to justify any available action. As Fantl and McGrath (2009: 77)
state: if p is a “reason you have, then the chance that [p] is false can’t stand in
the way of [p] justifying you in doing anything.”15 Safe Reasons claims that
when p is a reason one has to ϕ, p is warranted enough to justify you in any
available action.16 Hence, the strengthening test is an apt test, as it tests all
available actions simultaneously. If one ought to x and y, but conditional on
p one ought to x and ~y, then one fails the strengthening test. One cannot pass
the strengthening test for a particular available action while failing for other
available actions: failing for one action is failing for all.
By requiring that one pass the strengthening test, Safe Reasons places a
stringent requirement on being a reason one has. KJ likewise places a stringent
test on knowledge. Whether p is a reason one has to ϕ and whether one knows p,
on this picture, will depend on what actions are available in any given scenario. In
situations where there are multiple available actions, for p to be warranted
enough to justify S in ϕ-ing, it must be that for all actions available to S, what
S ought to do does not differ from what S ought to do conditional on p. P is not a
reason S has to ϕ if p is not warranted enough to justify S in ψ-ing. This may strike
some as a particularly demanding requirement on reasons one has. In Section 2.2,
we will see how this feature plays out in counterintuitive ways.

13
There is an expansive literature discussing ‘reasons one has.’ The discussion in this chapter is
limited to the epistemic position required for p to be a reason one has—it is neutral with respect to
other requirements, and in particular it is neutral with respect to the conditions under which p is an
objective reason to ϕ. Throughout the chapter I will speak as though knowledge is sufficient to make
p a reason one has, but of course knowing p is not sufficient for p to meet the objective requirements
on being a reason one has, if there are such requirements. See Schroeder (2008) for a helpful
discussion of objective and subjective reasons.
14
Although I find the knowledge-reasons connection represented by KR attractive, the argument
offered in this chapter does not commit one to KR. Rather, it leaves open the option to reject KR. See
Brown (2011) for discussion of this route.
15
An explicit formulation of Safe Reasons, as Fantl and McGrath intend it to be understood, is
thereby as follows: for all S, p, ϕ, if p is a reason S has to ϕ, then for all ψ, p is warranted enough to
justify S in ψ-ing.
16
Although Fantl and McGrath do not qualify ‘any action’ as any available action, as I do here,
this seems to be their intended meaning. In (2009: 67) they use specifications such as ‘in a certain
situation’ or ‘merely strengthening your epistemic position’ holding fixed other factors (such as the
agent’s stakes). In holding fixed a subject’s stakes, I take it that we hold fixed what actions are
available to the agent (that the subject is not facing a high-stakes bet, etc.). The addition of ‘available’
merely makes this evident. See also Fantl and McGrath (2009: 224–9) for relevant discussion.
 PUTTING FALLIBILISM TO WORK 

2.2 Against safe reasons

Consider the following case:
Dinner
Alli tells her husband Tim that she is going out for the evening and won’t be
home until late. On the basis of her testimony, Tim’s epistemic position for the
proposition Alli will be home late (hereafter HOMELATE) is very strong. Tim
decides to make pizza for dinner. Alli hates pizza, so he only has pizza for
dinner when she is out. He considers inviting his brother over to play chess,
but decides not to since Alli recently had a huge disagreement with his brother
and she made it clear to Tim that she does not want to see him for a while.17

Tim’s epistemic position for HOMELATE is very strong but not top level. Given
his epistemic position, he is justified in making pizza but he shouldn’t invite his
brother over. Although Alli almost never comes home early when she goes out for
the evening, if she did she would be very upset to run into Tim’s brother. We can
imagine Tim saying to his brother on the phone: “I’m making pizza for dinner
tonight—Alli isn’t going to be home until late. I’d invite you over to play chess,
but if she came home early that’d be a disaster; I wouldn’t want you guys to run
into each other.” Tim ought not invite his brother over, even though strength-
ening his epistemic position for HOMELATE makes it such that Tim ought to
invite his brother over—that is, conditional on HOMELATE, Tim ought to invite
his brother over. Tim fails the strengthening test.
It is natural to think that HOMELATE is a reason Tim has to do certain
actions, and in particular, it is natural to think HOMELATE is a reason Tim has
to make pizza. But HOMELATE is not warranted enough to justify Tim in any
available action. Sometimes one’s epistemic position is strong enough to justify
one action while simultaneously not strong enough to justify another action.
Consider the following statements:
(1) HOMELATE is a reason Tim has to make pizza.
(2) HOMELATE is a reason Tim has to invite his brother over.
(3) Tim ought to make pizza and ought not invite his brother over.
(4) HOMELATE is warranted enough to justify Tim in making pizza.
(5) HOMELATE is not warranted enough to justify Tim in inviting his
brother over.
(6) Conditional on HOMELATE, Tim ought to invite his brother over.

17
This case first appears in Anderson (2015).
 CHARITY ANDERSON

Each of (1)–(6) are plausible things to say about this case. But (1), (5), and (6) are
jointly incompatible with Safe Reasons. Two points deserve emphasis here. First,
(1) is prima facie very plausible: Tim’s epistemic position for the proposition is
very strong and making pizza is the rational thing for him to do. In this way,
denial of (1) is a cost. But according to Safe Reasons, (1) is false. Second, the
reason the proponent of Safe Reasons must deny (1) is surprising: the advocate of
Safe Reasons must deny (1) because (5) and (6) are true. But intuitively, whether
HOMELATE is a reason Tim has to make pizza should not depend on whether
HOMELATE is warranted enough to justify Tim in inviting his brother over. Safe
Reasons forces this counterintuitive result.
It might be tempting to think that the problem could be avoided by introdu-
cing probabilities. The envisioned strategy claims that the proposition available to
function as Tim’s reason to make pizza is Probably Alli is coming home late
(hereafter, HOMELATE*). The idea is that if HOMELATE* is a reason Tim has
for action, rather than HOMELATE, then HOMELATE* can explain why Tim is
justified in doing one action but not the other.
The first thing to note about this strategy is that if Alli comes home late and
asks Tim why he made pizza for dinner, he won’t give as his reason that she
probably was coming home late. His reason for making pizza is that she was
coming home late. Propositions about probabilities do not always provide plaus-
ible alternatives when it comes to the reasons that actually motivate us.
Furthermore, in at least some cases, the shift to HOMELATE* will merely push
the problem back a step. Consider that according to Safe Reasons, if HOME-
LATE* is a reason Tim has to ϕ, then HOMELATE* is warranted enough to
justify Tim in ϕ-ing, for any ϕ. But assume that Tim has a very strong but less
than top-level epistemic position for HOMELATE*. We can construct the details
of the case in such a way that Tim is rational to make pizza, but is not rational to
invite his brother over, even though conditional on HOMELATE* he is rational
to invite his brother over. When HOMELATE* fails the strengthening test, by
Safe Reasons, HOMELATE* is not a reason Tim has to do any action. There is
nothing special about propositions about probabilities that make them resistant
to counterexamples of the kind in Dinner.18

18
The strategy can be repeated at higher levels. The advocate of Safe Reasons might suggest that
Tim’s reason is only probably HOMELATE*. But there is no reason in principle why the counter-
example cannot be constructed around this alternative proposition. For each iteration there will be
cases where, given fallible knowledge of the relevant proposition and the right details of the case, the
replacement proposition will justify one action even though weakness in epistemic position for
the proposition stands in the way of another action. Of course, the more iterations that are added, the
less plausible it will be that we actually treat the suggested proposition as our reason for acting.
 PUTTING FALLIBILISM TO WORK 

More importantly, the availability of HOMELATE* or other propositions that

Tim could have treated as his reason to make pizza for dinner does not change the
fact that it is plausible to think that HOMELATE is a reason Tim has to make
pizza for dinner. (This applies equally to alternatives such as Alli said she was
coming home late.) If as a matter of fact Tim treats HOMELATE as his reason to
make pizza, the framework under consideration delivers the result that he does
something inappropriate. This strikes me as the wrong result.19
The reasoning in Dinner can be advanced directly against KJ by positing that
Tim knows HOMELATE. Given certain assumptions about Alli’s reliability as a
testifier, it is highly plausible that Tim comes to know HOMELATE on the basis
of her testimony. If Tim knows HOMELATE, and HOMELATE is not warranted
enough to justify Tim in making pizza for dinner and inviting his brother over,
then KJ is false. We can construct a similar line of reasoning for HOMELATE*.
It is plausible that Tim knows HOMELATE, and even more plausible that he
knows HOMELATE*. In this way, Dinner is a counterexample to both Safe
Reasons and to KJ.20
We should reject Safe Reasons and KJ. Passing the strengthening test for all
available actions is too strong a requirement for a proposition to be a reason one
has and too strong a requirement for knowledge.
In light of these difficulties, one might be inclined to formulate Safe Reasons
and KJ such that neither require passing the strengthening test for all available
actions. That is, when ϕ and ψ are available actions, one might hold that one
knows p (or, that p is a reason one has to ϕ) if one passes the strengthening test
with respect to ϕ-ing even if one does not pass the strengthening test with respect
to ψ-ing. Weakening KJ in this way is not a very promising option, since knowing
p with respect to one action when one does not know p with respect to a different
action will involve a kind of compartmentalization of knowledge that advocates
of pragmatic encroachment are not likely to embrace. Weakening Safe Reasons is

19
In more recent work, Fantl and McGrath (manuscript) state that the action need only be
available to the subject in order for failing the strengthening test for that action to deprive the subject
of p as a reason S has (and also to deprive the subject of knowledge). The result is that Tim fails to
have HOMELATE as a reason to make pizza for dinner when inviting his brother over is an available
action even in a setting where Tim does not consider inviting his brother over. In my opinion, this
adds further implausibility to the overall picture.
20
It’s worth considering how we might respond if some action that prevents us from passing the
strengthening test were always available—for example, if there were a genie offering us a high-stakes
bet for every proposition we were prepared to treat as a reason. I’d wager that we would hold fast to
the idea that these propositions are reasons we have to act, for all sorts of actions, and conclude that
our inability to treat them as reasons to accept the genie’s bets is irrelevant to whether we can
permissibly treat the propositions as reasons to act in other ways.
 CHARITY ANDERSON

likewise not an appealing option for the KJ theorist because Safe Reasons will no
longer be able to motivate KJ.
At this point it should be clear that KJ relies on a thought that is unnatural
given fallibilism. Given fallibilism, there is no reason to expect that when we
know p, we are justified in acting on p for any available action. In fact, fallibilism
predicts exactly the opposite. That is, fallibilism predicts that sometimes when we
know p, the difference between our actual epistemic position for p and a stronger
position with respect to p will make a difference to whether p is warranted
enough to justify some available action. In this way, KJ (and the framework of
reasons underlying KJ) involves a picture that does not fit well with fallibilism.

3 The Trilemma
I turn now to examine the trilemma between fallibilism, purism, and Sufficiency.
A central term in Sufficiency involves an ambiguity which, when elucidated,
provides a way out of the trilemma. Yet even in light of a position that can maintain
all three commitments, I contend that we have reason to reject Sufficiency.
Here are the principles involved in the trilemma, as they are commonly
glossed:
Purism: pragmatic factors do not affect whether one knows.
Fallibilism: knowledge is compatible with an epistemic chance of error.
Sufficiency: when S knows p, S’s epistemic position for p is such that p is warranted
enough to justify S in ϕ-ing, for any ϕ.

The notion of strength of epistemic position is notoriously vague. Without offering

a complete account, I will draw a distinction between two disparate ways of
thinking about epistemic position. The first construes one’s epistemic position in
such a way that it aligns naturally with the notion of epistemic chance used to
articulate fallibilism. The second departs from this alignment. In what follows,
I will paint in broad brushstrokes. I do not intend to give an account of fallibilism,
but rather to offer a general picture that many if not all fallibilists agree upon,
despite significant disagreement regarding exactly how to articulate fallibilism.
An important feature of the notion of epistemic chance as it is used in
statements of fallibilism is that it concerns something other than the facts of
success in a particular case—it represents something more like an approximation
of one’s modal success rate across similar worlds.21 Even when we get it right in

21
The modal success rate here involves the assumption that the relevant sets of worlds are finite.
Things are more complicated when it comes to infinite worlds. For the most part, these complexities
 PUTTING FALLIBILISM TO WORK 

the actual world, there are similar worlds where we form a false belief (in either
the same proposition or a similar proposition).22, 23 Fallibilists agree that although
our modal success rate is less than perfect, nevertheless, we can possess know-
ledge.24 For example, suppose you come to know that Paul is coming to the party
because your friend tells you. Many cases of testimony are such that there is
some similar world where the testifier is misinformed or misrepresenting infor-
mation but where you form a belief on the basis of the testimony nevertheless.25
Such worlds negatively affect an agent’s modal profile with respect to particular
cases of believing on testimony.26 When considering an unrestricted set of similar
worlds, fallibilists can agree about one thing: an imperfect modal success rate is
compatible with knowledge.
A further choice point concerns whether knowledge is compatible with an
imperfect epistemic position relative to some restricted subset of the similar
worlds. It’s fairly standard to think that not all similar worlds are relevant to
whether you know, and thus to restrict the worlds relevant to knowledge to some
more narrow subset. There are similar worlds, after all, where extremely unlikely
events occur—the bank is closed because terrorists attack, or an earthquake
destroys the building. Nevertheless, it’s common to think that in the actual
world (where these events do not occur), we can know the bank is open. With
respect to this restriction, there is further division between fallibilists concerning
whether knowledge requires a perfect modal success rate within this restricted
set. (This choice point lines up nicely with a distinction sometimes drawn
between strong and weak safety.) Some fallibilists allow that one can know
even if one believes p falsely within the aforementioned set, others require
perfection within the restricted set. It’s important to note that this is an ‘in-
house’ disagreement among fallibilists of various stripes.

are not relevant to the picture I present, though it’s worth noting that notions like ‘epistemic chance’
are problematic in the infinite setting and are not interchangeable with ‘epistemic possibility.’
22
Where by ‘similar world’ we hold fixed that you believe using the same or a similar method.
23
A standard way to account for fallible knowledge of necessary truths is to include in the set of
relevant worlds those where one forms beliefs about similar propositions.
24
How strong your success rate must be to be compatible with knowledge is a contentious issue.
Since I do not intend here to give an account of fallibilism, I have no need to articulate a threshold.
25
A fully general account of fallibilism will inevitably want to provide a principled reason for
drawing the line of similar/relevant worlds. Various options are available here—some may be
inclined to use something like the nearby worlds compatible with one’s mental states. I leave open
what constraints are needed.
26
It’s worth noting that because the modal success rate is an externalist notion, one’s epistemic
position is less than perfect even if one is not aware of one’s own fallibility—one is not always in a
position to know one’s modal profile.
 CHARITY ANDERSON

While fallibilists disagree on this point, what they have in common is that one’s
unrestricted modal success rate is imperfect. I’ll call this notion of epistemic
position (derived from consideration of imperfection in the unrestricted set)
one’s EI epistemic position.27 It is this notion that lies behind the gloss on
fallibilism above.
An alternative conception of epistemic position can be construed by looking
only at the restricted set of similar worlds. The view that one’s modal success rate
within the restricted set must be perfect in order for one to know results in a
corresponding picture of epistemic position whereby when one knows, there is
no chance of believing falsely. On this picture, all the worlds where you mess up
are outside the relevant range and are thereby irrelevant to this construal of
epistemic position. I will refer to this gloss on epistemic position (which requires
perfection within the relevant set of similar worlds) as one’s EP epistemic
position.
There is an important choice point regarding whether the former or latter
notion of epistemic position is relevant to justified action, and thus to Sufficiency.
Some but not all versions of Sufficiency involve thinking of one’s epistemic
position using the notion of one’s EI chance of error. Given the various construals
of epistemic position available, fallibilists face a choice point with respect to
which gloss on one’s strength of epistemic position is relevant to what one ought
to do: one’s EI or one’s EP. There is room for disagreement among fallibilists
concerning the relevancy of EI to justified action. Fallibilism is a commitment
about the compatibility of knowledge with an epistemic chance of error—a
chance that I have suggested can be understood with respect to one’s EI epistemic
position. It is a further (though somewhat natural) commitment for the fallibilist
to affirm that the same notion of epistemic position according to which one has a
chance of error is the notion of epistemic position relevant to determining what
one ought to do.28 It is open to the fallibilist to hold that one’s EI epistemic
position is irrelevant to rational action and instead to think of one’s EP epistemic
position as the relevant notion.
We are now in a position to see the role that the notion of one’s EI plays in the
trilemma. Here is the trilemma again, where the notion of one’s EI is made explicit:

27
One could also speak of the epistemic probability of p relative to one’s EI position, if one holds
that epistemic position is the kind of thing that satisfies the probability axioms. Since there is some
doubt as to whether and how epistemic position can be construed in terms of probability, I leave it to
the reader to make the application.
28
One might try to build into the notion of ‘fallibilism’ the idea that what one ought to do
depends on one’s EI epistemic position. But although many fallibilists do, in fact, think one’s EI
position is relevant to justified action, this constraint is not built into the definition of fallibilism.
There is space for a fallibilist position that rejects this picture.
 PUTTING FALLIBILISM TO WORK 

Purism: pragmatic factors do not make a difference to whether one knows p.

Fallibilism*: for some p, S knows p and there is an EI chance for S that ~p.
Sufficiency*: when S knows p, S’s EI epistemic position for p is such that p is
warranted enough to justify S in ϕ-ing, for any ϕ.
This version of Sufficiency gives rise to the trilemma. But fallibilists who hold that
one’s EP epistemic position determines what one ought to do are likely to
construe sufficiency instead as follows:
Sufficiency**: when S knows p, S’s EP epistemic position is such that p is
warranted enough to justify S in ϕ-ing, for any ϕ.
In virtue of maintaining that one’s EP determines what one ought to do, on this
view one will always pass the strengthening test when one knows p (since what
one ought to do will always be the same as what one ought to do conditional on
p). In this way, Sufficiency** is compatible with fallibilist* purism. Fallibilists who
hold that one’s EP determines what one ought to do, and thereby hold Suffi-
ciency** instead of Sufficiency*, avoid the trilemma.29, 30
We’ll call views that affirm fallibilism* and Sufficiency** chance-irrelevant falli-
bilism. According to chance-irrelevant fallibilism, when S knows p, and we consider
what S is rational to do, there is no need to consider S’s EI position for p. Since what
one ought to do is not given by one’s EI position, it is irrelevant on this view whether
one passes or fails the strengthening test with respect to one’s EI epistemic position.
Alternatively, there is chance-relevant fallibilism. Chance-relevant fallibilists
hold that one’s EI epistemic position is relevant to rational action. Thus, on this
view, sometimes the difference between an EI position of 0.97 for p and an EI
position of 0.99 will make a difference to what one ought to do. (I assume here an
idealization of epistemic position such that it can be represented using numbers
between 0 and 1. Those who find this objectionable may substitute ‘strong
epistemic position’ and ‘stronger epistemic position’ for the relevant values.)

29
On one reading of Hawthorne and Stanley (2008) their position is an example of a view that
uses a subject’s EP to decide what one ought to do. Their position denies purism, despite the fact that
Sufficiency** does not force the denial of purism. So denial of purism is a superfluous feature of the
view. See also Dutant (manuscript) for relevant discussion.
30
There are other options I haven’t considered here. One could try to construe epistemic position
using a scale such that one’s epistemic position is represented as perfect only when one knows that
one knows. When wedded to decision theory, this view will favor a version of Sufficiency along the
following lines: when S knows that she knows p, S is warranted enough to act as if p. One who holds
this principle will also avoid the trilemma. Further iterations of knowledge to set the top threshold of
the scale of the epistemic notion relevant to determining what one ought to do are also an option.
Exploration of these various positions would take us too far afield.
 CHARITY ANDERSON

There is a further divide between chance-relevant fallibilists. According to

chance-relevant fallibilists who affirm Sufficiency*, when the difference between
an EI position of 0.97 and an EI position of 0.99 makes a difference to what
S ought to do, S does not know p. Hence, not only is the difference between a strong
EI position and a top-level epistemic position relevant to justified action, it is also
relevant to knowledge. This results in pragmatic encroachment. An alternative
chance-relevant position is stable fallibilism. Stable fallibilists hold that one’s EI
position is relevant to justified action, and one can know p even when weakness in
one’s EI stands in the way of acting (that is, even when one’s EI position is such that
one fails the strengthening test). Knowledge is stable on this view—whether one
has knowledge does not depend on one’s practical decision setting.
In closing, I advance a consideration in favor of stable fallibilism that has gone
unappreciated. The consideration is this: only stable fallibilism allows us to put
fallibilism to work. A key fallibilist insight is that knowing p is compatible with a
small EI chance that ~p. Stable fallibilism is the only view among the three that
allows this insight to be widely relied upon. Chance-irrelevant fallibilism,
although it is committed to the compatibility of knowledge and a small EI chance
that ~p, does not allow these chances to make a difference to rational action
because EI chances are always irrelevant to what one ought to do. Chance-
relevant fallibilism tied to Sufficiency* also does not allow the fallibilist insight
to be put to work because when the EI chance that ~p is such that one fails the
strengthening test, one fails to know p. Consequently, one’s fallible knowledge
cannot be relied on for anything in these settings. Both of these views hold that
you know p only when it is rational for you to act just as you ought to act
conditional on p (that is, conditional on there being no EI chance of error). In this
way, both views make the key fallibilist insight—the compatibility of knowledge
and a chance of error—idle.
In conclusion, I’ve argued that KJ and Safe Reasons ought to be rejected and
that affirmation of Sufficiency involves fallibilists in a kind of double-mindedness.
I’ve delineated two notions of epistemic position relevant to unpacking the
principles in the trilemma, and shown how Sufficiency can be understood in a
way consistent with purism and fallibilism. Finally, I’ve argued that stable
fallibilism has an unappreciated advantage over the competing two views: it is
the only position among the three that allows us to put a key fallibilist insight
to work.31

31
I am particularly grateful to Julien Dutant and John Hawthorne for providing detailed
comments on drafts of this chapter. Thanks also to Trent Dougherty, John Greco, Jeffrey Russell,
Matthew McGrath, and the editors of this volume for helpful comments and discussion.
 PUTTING FALLIBILISM TO WORK 

References
Anderson, C. (2015). On the intimate relationship of knowledge and action. Episteme,
12(3): 343–53.
Brown, J. (2011). Fallibilism and the knowledge norm. In Assertion: New Philosophical
Essays. Edited by J. Brown and H. Cappelen. Oxford: Oxford University Press.
Brown, J. (2012). Assertion and practical reasoning: Common or divergent epistemic
standards? Philosophy and Phenomenological Research, 84(1): 123–57.
Dutant, J. Normative sceptical paradoxes. Unpublished manuscript.
Fantl, J. and McGrath, M. (2002). Evidence, pragmatics, and justification. Philosophical
Review, 111(1): 67–94.
Fantl, J. and McGrath, M. (2007). On pragmatic encroachment in epistemology. Philoso-
phy and Phenomenological Research, 75(3): 558–89.
Fantl, J. and McGrath, M. (2009). Knowledge in an Uncertain World. Oxford: Oxford
University Press.
Fantl, J. and McGrath, M. (manuscript). On two ultimately unsuccessful objections to
pragmatic encroachment. Presented at Oxford University’s Workshop on Contextual-
ism, Pragmatic Encroachment, and Religious Epistemology.
Hawthorne, J. (2004). Knowledge and Lotteries. Oxford: Oxford University Press.
Hawthorne, J. and Stanley, J. (2008). Knowledge and action. Journal of Philosophy, CV
(10): 571–90.
Schroeder, M. (2008). Having reasons. Philosophical Studies, 139: 57–71.
Stanley, J. (2005). Knowledge and Practical Interests. Oxford: Oxford University Press.
 2
Pragmatic Approaches to Belief
Jessica Brown

1 Introduction
A number of authors have recently argued for a pragmatic approach to belief on
which full belief is understood in terms of its role in practical reasoning (e.g.
Weatherson 2005 and 2014; Ganson 2008; Fantl and McGrath 2009; and Ross
and Schroeder 2014). This approach has been motivated on the grounds that it
deals with the threshold problem for belief and simultaneously explains why we
need a notion of outright belief in addition to degrees of belief, or credence.
However, I argue that it does not provide a satisfactory solution to these problems
because of cases in which a subject believes a proposition which is not relevant for
her practical reasoning. I consider both pragmatic approaches which endorse an
extremely tight relationship between belief and practical reasoning and those
which endorse a weaker relationship.

2 Pragmatic Encroachment about Belief

Pragmatic approaches have been motivated as offering a solution to the well-
known problems in unifying the frameworks of outright belief and degrees of
belief. The most obvious way to unify these two frameworks is to regard belief as
equivalent to some degree of belief. However, requiring credence 1 for outright
belief threatens to have the result that we have very few beliefs at all. For at least
on standard assumptions connecting credence and decision theory, if one places
credence 1 in a proposition, then one is prepared to bet on it at any stakes. But,
there are very few propositions we are prepared to bet on at any stakes. As a
result, many have endorsed a ‘Lockean’ conception of belief on which one
believes that p if and only if one’s credence in p is greater than some fixed
threshold, t, less than 1. Presumably, the relevant threshold must be considerably
 PRAGMATIC APPROACHES TO BELIEF 

more than 0.5. But where should it be? At 0.8, 0.81, 0.82, 0.83? It seems hard to
provide an answer to this question which is not arbitrary (this is the ‘threshold
problem for belief ’). Furthermore, we might wonder what is the point of adding
the Lockean notion of outright belief to a framework which contains degrees of
belief? We can call the latter problem ‘the value problem for belief ’.
These problems have been used to motivate a pragmatic account of belief
according to which belief should be understood in terms of its role in practical
reasoning. According to one way to develop this approach—pragmatic encroach-
ment about belief, or PEB—to believe that p is for your credence in p to be high
enough for you to rely on p in your practical reasoning, or to take p as a premise
in reasoning concerning what to do (e.g. Weatherson 2005 and 2014, Ganson
2008, and Fantl and McGrath 2009: chapter 5). PEB has the consequence that
whether one believes that p depends on the stakes. For, whether one is prepared
to rely on a proposition in one’s practical reasoning depends in part on what is at
stake. Proponents of PEB argue that it can solve the threshold problem since it
holds that what distinguishes the credence in p required for belief in p is the fact
that it is the credence required for one to rely on p in one’s practical reasoning.
Further, they argue that PEB explains the value of belief in terms of its role in
practical reasoning (e.g. Weatherson 2005; Fantl and McGrath 2009).
To see whether PEB can fulfil its promise of solving the threshold and
value problems, we need to clarify the link it proposes between belief and action.
To do so, let’s start with the following claim about the relation between belief
and action:
The belief-action link. You believe p if and only if your credence that p is high
enough for p to be your motivating reason for φ-ing for all φ.
To understand the belief-action link we need to know how we should understand
the expression ‘all φ’. What kinds of action does φ range over? In particular, does
it include only those live options for action present in one’s actual situation, or
also merely hypothetical actions which are not currently available? For instance,
suppose that I have high credence that there is no cereal in the house as I seem to
remember finishing it off last night. Which actions are relevant to whether this
credence constitutes belief? Do they include only the actual actions available to
me (e.g. buying some cereal at the supermarket) or merely possible actions (e.g.
taking a bet which pays out £10 if there is no cereal in the house and on which
I lose my house otherwise)? In addition, do the relevant actions include only
those to which the proposition is relevant (e.g. buying some more cereal at the
supermarket) or any action I face whether or not the proposition is relevant to it?
For instance, if I also face the decision whether to take a new job, is this latter
 JESSICA BROWN

Table 2.1. Table of options

Only live options Live and hypothetical options

Content-restriction Live and content-restriction. Live and hypothetical; content-

restriction.
No content-restriction Live; no content-restriction. Live and hypothetical; no
content-restriction.

action relevant to determining whether my credence that there is no cereal in the

house constitutes belief?
These choices generate a table of options for those defending PEB (Table 2.1).
Weatherson (2005) and Fantl and McGrath (2009) defend versions of PEB which
occupy the top left-hand box of the table.1 However, Weatherson (2014) defends
a version of PEB occupying the top right-hand box of the table (21–2). To my
knowledge, no one has endorsed either the bottom left-hand or bottom right-
hand boxes. In what follows, I will consider all four options, arguing that none of
them enables PEB to adequately answer the value and threshold problems. I start
in Section 3 by considering the combination of live options for action with a
content-restriction.

3 Live Options for Action with a Content-Restriction

At least initially, it seems that, in the belief-action link, φ must be restricted to live
options for action. For, if it ranges over all possible actions whatsoever it would
have the result that we hardly believe anything at all. For instance, suppose that
I have a high credence in the proposition that there is no cereal in the house, or c.
Nonetheless, I wouldn’t be prepared to use c as a premise in deciding whether to
take the hypothetical bet which pays out a penny if there is no cereal in the house
and yields eternal torture otherwise. So, initially, let us assume that φ ranges over
only live options for action in one’s current practical reasoning situation. Further,
it may seem natural to develop PEB with a content-restriction.2 On this view,
whether my high credence in c counts as a belief depends on whether I would rely
on c in those practical reasoning decisions to which c is relevant.
While the combination of live options for action with a content-restriction
seems intuitively plausible, we may worry about its consequences for propositions

1
See Weatherson 2005: 422, and Fantl and McGrath 2009: 137–8.
2
This understanding of PEB is analogous to one understanding of pragmatic encroachment
about knowledge advocated in Hawthorne and Stanley (2008). That view about knowledge faces
analogous difficulties with practically irrelevant propositions (Brown 2012).
 PRAGMATIC APPROACHES TO BELIEF 

which are not relevant to any practical reasoning decision one is facing. Consider,
for example, the proposition that Mars has two moons. Even though I have some
credence in this proposition, it is irrelevant to any practical decisions I now face.
I will argue that, depending on how the belief-action link is interpreted, it either
fails to determine a threshold of credence for belief in practically irrelevant
propositions, or does determine a threshold of credence for belief but one which
is trivially met.
On a ‘presuppositional’ reading, the belief-action link determines a threshold
of credence in p for one to have a belief that p only if there are live options for
action to which p is relevant. If there are no such actions, it fails to determine a
threshold of credence at all. On this view, the belief-action link fails to determine
a threshold of credence for me to believe that Mars has two moons.
On an alternative ‘conditional’ reading, the belief-action link amounts to the
claim that any credence, n, which S has in p at t, constitutes a belief if and only if
the following conditional holds: for any φ (if φ is a live option for action to which
p is relevant, then n is high enough for S to rely on p in deciding whether to φ).
Given that any conditional with a false antecedent is true, if p is practically
irrelevant then this conditional is trivially met regardless of one’s credence in
p. Thus, for instance, even if I have an extremely low credence in the proposition
that Mars has two moons, say 0.2, I count as fully believing this proposition.
Furthermore, if I have a 0.2 credence in the proposition that Mars has two moons
and a 0.2 credence in its negation, I count as fully believing both that Mars has
two moons and as fully believing that it does not have two moons. But, surely
I don’t have contradictory beliefs so easily! Thus, on either the presuppositional
or conditional reading, understanding the belief-action link as restricted to live
options for action to which the proposition is relevant fails to provide a satisfac-
tory answer to the threshold problem for practically irrelevant propositions.
In addition, it fails to satisfactorily address the value problem. For, where a belief
is practically irrelevant, it is hard to explain the importance of that belief in terms
of its connection to live options for action.
In reply, the defender of PEB may claim that, for any proposition p in which
one has some credence, there is always some potential action it is relevant to,
namely asserting p. Certainly, it’s usually physically possible for someone to
assert any proposition in which they have some credence. However, speakers
usually have no desire to assert propositions regardless of their relevance, e.g. to
the ongoing conversation. So, there is no live practical decision for them of
whether to assert a proposition irrespective of its relevance to the ongoing
conversation. A similar problem affects the suggestion that for any proposition,
p, in which one has some credence, one can always act on it by drawing further
 JESSICA BROWN

consequences from it in thought. Agents do not usually have a desire to draw

further consequences from their credences no matter whether those conse-
quences are relevant to their situation. Thus, appealing to the mere fact that it
is standardly physically possible for a subject who has a credence in p to assert
that p, or infer further credences from that credence, does not overcome the
problem concerning practically irrelevant propositions.
It’s important to notice that the difficulties PEB faces with practically irrelevant
propositions do not just affect a few propositions. Most adults acquire credences
in a vast range of propositions many of which are, for much of the time, irrelevant
to any practical reasoning situation they face. These propositions may concern
geographical and historical matters, propositions about cultural products such as
films, novels, pop songs, or plays, or propositions concerning scientific matters.
Moreover, these credences are acquired in mundane, familiar ways, such as via
memory, testimony, and perception. To take just a few examples, I have some
credence in the following propositions: that Lima is the capital of Peru; that
the French Revolution started in 1789; that Chopin was born in Poland; that
strontium was once mined in Scotland, and that Cole Porter wrote Kiss Me Kate.
None of these propositions is relevant to any current practical reasoning situ-
ations I now face. Thus, understood with a content-restriction, the belief-action
link does not merely face difficulties in a few, borderline cases. Rather, it doesn’t
offer an adequate account of the threshold of belief for a vast range of credences
formed in familiar ways. Relatedly, it doesn’t offer an adequate account of the
value of belief in these propositions.
One potential way to try and deal with this problem would be to add a
condition requiring a certain minimal level of confidence for one to count as
believing that p. For instance, Weatherson (2005) adds some further conditions
to his original account of belief which have the net effect that when a proposition
p is practically irrelevant, one believes that p only if one’s credence in p is above
0.5 (424–5). But, specifying that 0.5 is the threshold of credence for outright belief
in practically irrelevant propositions seems ad hoc. Moreover, as Weatherson
(2014) acknowledges, merely requiring credence 0.5 for one to have outright
belief in a practically irrelevant proposition seems far too weak a requirement.
Surely, I don’t believe every proposition which is both practically irrelevant to my
current practical reasoning situation and in which I have a more than 0.5
credence (Ross and Schroeder 2014).
I conclude that the belief-action link understood so that φ is restricted to live
options for action with a content-restriction is unable to offer a satisfactory
solution to either the threshold or value problem for practically irrelevant
propositions.
 PRAGMATIC APPROACHES TO BELIEF 

4 Live Options for Action; No Content-Restriction

Let us now examine the belief-action link with φ ranging over live options for
action with no content-restriction. To see how the belief-action link should be
understood without a content-restriction suppose that I’m giving dinner to my
brother’s children. I face a high-stakes decision of whether to allow little Tommy
to have some of the Indian takeaway given his potentially fatal peanut allergy.
Although I have a high credence of say 0.9 in the proposition that the takeaway
doesn’t contain peanuts, let’s suppose that this is not high enough for me to rely
on that proposition in making this decision. Now suppose that I also have 0.9
credence in another proposition, say that Cole Porter wrote Kiss Me Kate. Even
though this proposition is entirely irrelevant to the decision whether to allow
little Tommy to have some of the Indian takeaway, we can still ask the question
whether the credence is high enough for me to rely on this proposition in making
the takeaway decision. Of course, there would be something deeply bizarre about
relying on it given that it is indeed irrelevant to the decision about whether
Tommy can have the takeaway. But, setting that aside, we can still ask whether
my credence in the Cole Porter proposition is high enough for me to rely on it in
making the decision about the takeaway. Given that a 0.9 credence in a relevant
proposition was not sufficient, it seems that a 0.9 credence in the Cole Porter
proposition is not sufficient either. Thus, endorsing the belief-action link without
a content-restriction can yield a non-trivial threshold for belief in practically
irrelevant propositions.
Even though understanding the belief-action link without a content-restriction
yields a non-trivial threshold for belief in practically irrelevant propositions, I will
argue that it faces the objection that it is too sceptical about belief. To show this,
I start by arguing that it has the result that the relevant threshold for one to
believe any proposition at a time is determined by the highest-stakes practical
reasoning decision one faces at that time. I then argue that we are frequently in
high-stakes situations.
Suppose that one simultaneously faces a high-stakes decision and a low-stakes
decision. For instance, the high-stakes decision could be the decision whether or
not to give little Tommy some of the Indian takeaway given his potentially fatal
peanut allergy. The low-stakes decision might be the decision whether to allow
little Sally to have a big glass of milk. The latter decision is low stakes given that,
at worst, all that is at risk is whether there is enough milk to have a cup of tea later
in the evening (the milkman will deliver new supplies tomorrow morning). Given
that the milk decision is a low-stakes decision, it seems that the credence required
for one to rely on a proposition in that decision is relatively undemanding.
 JESSICA BROWN

Certainly, having a credence 0.9 would suffice. So, if we allow the milk decision to
fix the threshold for belief for any proposition, then a 0.9 credence in the
takeaway proposition would count as an outright belief. But, plausibly, given
that little Tommy has a potentially fatal peanut allergy, I would not rely on that
proposition in deciding whether to give him the takeaway. Thus, if we allow the
single threshold for belief to be determined by anything other than the highest-
stakes situation, the tight connection between belief and action presupposed by
PEB would be broken. Thus, the belief-action link without a content-restriction
is best understood as having the effect that, for any subject, S, and any time,
there is a single threshold of credence for S then to believe any proposition, where
that threshold is determined by the highest-stakes practical reasoning situation
S then faces.
Notice that many of us face high-stakes decisions in our lives. For instance, we
may face decisions about care options for serious diseases, such as cancer and
heart disease; decisions about whether or not to enter or end a marriage, to have
children, to take a certain job, to move house, to invest one’s life savings in certain
stocks, and so on. Since these are high-stakes decisions, making them often takes
a considerable period of time. So, for example, someone might be in the situation
of deciding whether or not to enter into or end a marriage over many months, or
even in some cases years, before finally coming to a resolution. Of course
someone making this decision would not be consciously considering it at every
second of every day over several years. But lack of conscious reflection does not
show one is not in a decision situation. Rather, one can be in a practical reasoning
situation for a period even if one is not consciously considering it at every second in
that period. So, given the everyday nature of many of these high-stakes decisions,
and the fact that we are in some of them for long periods of time, many of us are
frequently in high-stakes situations.
Now the credence required for relying on a proposition in a quotidian high-
stakes situation may be very demanding. For instance, consider a decision that
many of us have faced, namely whether to change job. Changing job is a complex
decision which may involve moving house or even country of residence, moving
one’s children into new schools in the new location, finding a new job for one’s
spouse as well as oneself, and changing one’s pension and private health insur-
ance entitlements. In making such a complex decision with such high-stakes
consequences, the credence required for relying on relevant propositions in
making the decision is high. For instance, suppose that a friend living in the
area assures one that places in good local schools are available, and an employee
at the new institution says they have a good health insurance scheme. Even if
one trusts both the friend and the employee and so places high credence in the
 PRAGMATIC APPROACHES TO BELIEF 

relevant propositions, it seems that one would likely not simply be disposed to
employ these propositions in one’s reasoning about whether to take the job
but instead seek independent confirmation. Alternatively, suppose one finds
oneself with a high credence that in a big city such as the city one is considering
moving to, there would be many opportunities for one’s spouse to find work.
Again, it seems that one would not simply be disposed to take this for granted
in reasoning to take the job, but would instead seek evidence, say by looking
on a few internet sites for job vacancies in the relevant area. So, in such everyday
high-stakes situations, one would not simply rely on a proposition on the
grounds of apparent memory or non-expert testimony, even if one has high
credence in it.
When we combine the facts that we frequently face high-stakes practical
decisions, and that they push the threshold for belief high, we can see that
endorsing the belief-action link with no content-restriction runs the risk of
crediting us with very few beliefs at all. Of course, that’s not to say that it has
the result that we have no beliefs at all. First, a subject may happen to face no
high-stakes situation at a time. Second, even supposing that one is in a high-
stakes scenario, say a high-stakes divorce decision scenario, that decision might
motivate one to acquire the credence required for one to have belief in relevant
propositions, say propositions concerning the custody of children following a
separation. Furthermore, even if one doesn’t have the credence in a proposition,
p, required to have belief in it, one could have the credence required to have belief
in a related proposition, say that it is likely that p. Still, endorsing the belief-action
link without a content-restriction does have the consequence that we don’t
believe many of the propositions we ordinarily take ourselves to believe. In
particular, we would ordinarily take ourselves to believe many propositions in
which we have high credence supported by apparent memory or non-expert
testimony. But, it seems that in the kinds of high-stakes practical decisions we
routinely face, high credences supported in this way do not constitute belief. That
seems a sceptical enough result.
So far, I’ve argued that there is no acceptable version of the belief-action link
understood so that φ can range only over live options for action. In Section 5,
I consider whether φ should range over live and hypothetical options.

5 Live and Hypothetical Options for Action

Supposing that we allow φ to range over both live and hypothetical options
for action, we could do so with or without a content-restriction. It seems that we
can quickly reject the combination of live and hypothetical actions with no
 JESSICA BROWN

content-restriction. As we saw in Section 4, treating the relevant actions as

restricted to live options with no content-restriction yielded a view which was
too sceptical about belief. It’s hard to see how adding in merely hypothetical
actions can ameliorate this result. To the extent that those hypothetical actions
involve lower stakes than the actual live options we face, then the single threshold
for belief is determined by the actual live options we face. So, adding in hypo-
thetical actions doesn’t help lower the standard. Alternatively, the hypothetical
actions might involve even higher stakes than those we actually face. This would
merely serve to push the single threshold for belief even higher and so exacerbate
the problem of scepticism about belief.
It seems, then, that if one takes the relevant actions to include both live and
hypothetical options, one should also embrace a content-restriction. Weatherson
(2014) pursues this strategy in developing his view to deal with practically
irrelevant propositions. On Weatherson’s basic approach, one believes that p if
and only if one has preferences which make sense, by one’s own lights, in a world
in which p is true. This account is trivially met by any practically irrelevant
proposition. For, where some proposition is practically irrelevant, neither it
nor its negation make any difference to one’s preferences for action. Thus,
Weatherson’s basic account has the result that, for any practically irrelevant
proposition, one both believes it and believes its negation. In response, Weatherson
suggests adding an extra condition to his account, namely that ‘to believe that
p, there must be some decision problem such that some table the agent would
be disposed to use to solve it encodes that p. If there is no such problem, the agent
does not believe that p’ (21). Given that Weatherson’s additional condition is
supposed to help with practically irrelevant propositions, it must be understood
as involving merely possible decision problems, rather than live ones the subject
is facing. For, if p is a practically irrelevant proposition for me, there is no live
decision problem I’m facing which turns on whether p.
To see whether Weatherson’s additional condition helps, let’s apply it to an
example of a belief in a practically irrelevant proposition. Plausibly, I believe that
Mars has two moons (or m) as I read this in an encyclopaedia. Now consider a
hypothetical decision problem to which this proposition is relevant, say a bet
which pays one penny if m, and loses one penny otherwise. The decision problem
can be represented as in Table 2.2. It is plausible that in solving this decision
problem I would ignore the possibility that not-m, and use the truncated table
formed by deleting the not-m column (Table 2.3). In Weatherson’s terminology,
Table 2.3 ‘encodes that p’ in the sense that it ignores the possibility that not-p.
So Weatherson’s additional condition amounts to the idea that one believes that
p only if there is some possible decision problem such that there is some table
Another random document with
no related content on Scribd:
Hyacinth
Hyde
hydraulic
hydrogen
hydrographic
hydrographical
Hydrological
hygiene
hygienic
Hyndman
Hyperborean
hypnotized
hypocrisy
hypocritical
hypotheses
hypothesis
hysterical
Hæmamœbidæ
hésitez
hêng
Hübner
I
Ian
iao
Ibayat
Ibero
Ibi
Ibrahimieh
ice
ICELAND
icequake
Ichang
IDAHO
Ide
idea
ideal
ideals
ideas
Identic
identical
identification
identified
identify
identifying
identity
idiocy
idiots
idle
idleness
idol
idolatrous
idolatry
idols
if
Ignacio
Ignatius
ignited
ignorance
ignorant
ignore
ignored
ignoring
II
iii
Il
Iles
Ilios
ill
illegal
illegality
illegalité
illegally
illiberal
illicit
illimitable
Illinois
illiteracy
illiterate
illiterates
illness
Illocos
illogical
ills
illuminant
illuminate
illuminating
illumination
illumined
illusion
illusory
illustrate
Illustrated
Illustrating
illustration
illustrations
illustrative
illustrious
Illustrirte
Ilocanos
Ilocos
Iloilo
Ilorin
im
Image
imaginable
imaginary
imagination
imagine
imagined
Imarovatana
Imataka
imbued
Imerina
Imgur
imitate
imitated
imitating
imitation
immaterial
immeasurable
immeasurably
Immediate
Immediately
immemorial
immense
immensity
immigrant
immigrants
immigrated
immigration
imminence
imminent
immoral
immorality
immortal
Immortalité
immovable
immune
immunities
immunity
immunized
impact
impair
impaired
impairing
impairment
impairs
impart
imparted
impartial
impartiality
impartially
impassable
Impassioned
impatience
impatient
impatiently
impeached
impeachment
impede
impeded
impediment
impediments
impending
imperative
imperatively
imperfect
imperfection
imperfections
imperfectly
imperfectness
IMPERIAL
imperialism
Imperialist
Imperialistic
imperialists
imperil
imperilled
imperilling
imperils
imperious
imperishable
impermeable
impersonation
impertinent
impervious
impetus
impinge
impinges
impiously
implacable
implants
implements
implicate
implicated
implication
implicitly
implied
implies
implore
implored
imploring
imply
implying
impolicy
impolitic
import
importance
Important
importantly
importation
importations
imported
importer
importers
importing
imports
importunate
importunately
importuned
impose
imposed
imposes
imposing
imposition
impossibility
impossible
impost
imposts
impotence
impotency
impoverish
impoverished
impoverishment
impracticability
impracticable
impregnable
impregnated
impress
impressed
impressing
impression
impressionable
impressions
impressive
impressively
impressiveness
imprison
imprisoned
imprisonment
improbable
improper
improperly
impropriety
improve
improved
Improvement
Improvements
improvidence
improving
imprudence
impudence
impudent
impulse
impulses
impulsive
impunity
imputation
imputed
Imus
in
inability
inaccessible
inaccuracy
inaccurate
inaction
inactive
inactivity
inadequacy
inadequate
inadequately
inadmissible
inadvisability
inadvisable
inalienable
inapplicable
inarticulate
inasmuch
Inaugural
inaugurate
inaugurated
inaugurating
inauguration
Inayat
incalculable
incalculably
incandescent
incantations
incapable
incapacitated
incapacity
incarcerated
incarceration
Incarnation
incarnations
Incas
incendiarism
incense
incensed
incentive
inception
incessant
incessantly
inch
inches
incidence
INCIDENT
Incidental
incidentally
incidents
incipient
incisions
incisive
incite
incited
inciting
inclination
inclinations
inclined
inclose
inclosed
incloses
inclosure
inclosures
include
Included
includes
Including
Inclusion
inclusions
inclusive
incognito
incoherent
INCOME
incomers
incomes
incoming
incomparable
incompatibility
incompatible
incompetent
incomplete
incomprehensible
inconceivable
inconclusive
incongruous
inconsiderable
inconsistency
inconsistent
incontestable
incontinently
incontrovertible
inconvenience
inconveniences
inconvenient
inconveniently
incorporate
incorporated
incorporates
incorporating
incorporation
incorrect
incorrectly
increase
increased
Increases
increasing
increasingly
incredible
incredibly
incredulity
increment
incriminated
incubation
inculcated
incumbency
incumbent
incumbered
incur
incurable
incurably
incurred
incurs
incursion
incursions
Indang
indebted
indebtedness
indecision
indecisive
indeed
indefensible
indefinite
indefinitely
indelible
indemnified
indemnify
Indemnities
indemnity
indenture
Independence
Independent
independently
Independents
indescribable
indeterminate
index
indexes
indexing
INDIA
Indian
INDIANA
Indianapolis
Indians
indicate
indicated
indicates
indicating
indication
indications
indicted
indictment
indictments
Indies
indifference
indifferent
indifferently
indigenous
indigestible
indigestion
indignant
indignantly
indignation
indignity
indigo
Indios
indiquant
indirect
indirectly
indiscretion
indiscriminate
indispensable
indispensably
indisputable
indisputably
indissoluble
individual
individuality
individually
individuals
indomitable
INDONESIAN
Indonesians
indoor
indorse
indorsed
indorsement
indubitable
induce
induced
inducement
induces
inducing
inductance
inducted
induction
indulge
indulged
indulgence
indulgences
indulgent
indulges
indulging
indunas
Indus
industrial
industrialism
Industries
industrious
industry
ineffaceably
ineffective
Ineffectual
inefficiency
inefficient
inelastic
inequality
inequitable
ineradicable
inertia
inestimable
inevitable
inevitably
inexhaustible
inexpediency
inexpedient
inexpensive
inexplicable
inexplicit
inexpressible
inextricably
infallible
infallibly
infamous
infamy
infancy
infant
Infanta
Infanticide
INFANTRY
infants
infect
infected
infecting
infection
infectious
infer
inference
Inferior
inferiority
infernally
inferred
inferring
infested
infidelium
infinite
infinitely
infirm
infirmaries
inflamed
inflammable
inflammation
inflammatory
inflexible
inflict
inflicted
inflicting
Influence
influenced
influences
influential
influenza
influx
inform
informal
informally
informant
informants
information
informed
informing
informs
infraction
infractions
infrequency
infrequently
infringe
infringement
infringing
infused
ING
ingenious
ingeniously
ingenuity
Ingersoll
ingested
Ingolf
ingots
inhabit
inhabitant
inhabitants
inhabited
inhabiting
inhabits
inhaled
inharmonious
inherent
inherit
inheritance
inheritances
inherited
inheritor
inherits
inhibition
inhuman
inhumanity
iniquities
iniquitous
iniquity
initial
initiate
initiated
initiation
initiative
Initiator
injected
injecting
injection
injudicious
injunction
injunctions
injure
injured
injuries
injuring
injurious
injuriously
injury
injustice
Inland
inlet
inlets
Inman
inmates
inn
Inner
Innes
innings
innocence
innocent
innocently
innocuous
innovation
innovations
innumerable
inoculated
Inoculation
inoculations
inoffensive
inoperative
inopportune
Inouye
inquire
inquired
inquirer
inquires
inquiries
inquiring
inquiry
inquisitive
inroad
inroads
insane
insanity
insatiable
insatiate
inscribe
inscribed
inscription
Inscriptions
insect
insecticide
insects
insecure
insecurity
inseparable
insert
inserted
insertion
inshore
inside
insight
insignia
insignificance
insignificant
insincere
insincerity
insinuation
insinuations
insist
insisted
insistence
insistent
insisting
insists
insolence
insolent
insolvency
insolvent
insolvents
insomnia
insomuch
inspect
inspected
inspecting
INSPECTION
inspections
Inspector
Inspectorate
inspectors
inspiration
inspire
inspired
inspiring
inst
installation
installations
installed
installment
installments
instalment
instalments
instance
instances
instant
instantaneous
instantaneously
instanter
instantly
Instead
instigated
instigates
instigating
instigation
instigator
instigators
instill
instinct
instinctive
instinctively
instincts
Institut
Institute
instituted
institutes
institution
institutional
institutions
instruct
instructed