Shifty Speech and Independent Thought: Epistemic Normativity in Context [1 ed.] 0192895281, 9780192895288

Citation preview

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

Shifty Speech and Independent Thought

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

Shifty Speech and Independent Thought Epistemic Normativity in Context MONA SIMION

1

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

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 © Mona Simion 2021 The moral rights of the author have been asserted First Edition published in 2021 Impression: 1 All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, without the prior permission in writing of Oxford University Press, or as expressly permitted by law, by licence or under terms agreed with the appropriate reprographics rights organization. Enquiries concerning reproduction outside the scope of the above should be sent to the Rights Department, Oxford University Press, at the address above You must not circulate this work in any other form and you must impose this same condition on any acquirer Published in the United States of America by Oxford University Press 198 Madison Avenue, New York, NY 10016, United States of America British Library Cataloguing in Publication Data Data available Library of Congress Control Number: 2020949831 ISBN 978–0–19–289528–8 DOI: 10.1093/oso/9780192895288.001.0001 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.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

For my mother

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

Preface I have been thinking about the relation between practical stakes and epistemically proper thought and speech for many years now, and I have benefited from an unusual amount of support from my peers and mentors for this research. I would like to especially thank Jessica Brown, Herman Cappelen, Adam Carter, Sandy Goldberg, Peter J. Graham, Patrick Greenough, Chris Kelp, and Tim Williamson for the extensive feedback on this material and all the support. I would also like to thank Bob Beddor, Sven Bernecker, Corine Besson, Michael Blome-Tillman, Emma Borg, Cameron Boult, Michael Brady, Elke Brendel, Fernando Broncano-Berrocal, Annalisa Coliva, Robert Cowan, Marian David, Katherine Dormandy, Julien Dutant, Miguel Egler, Claire Field, Giada Fratantonio, Lizzie Fricker, Miranda Fricker, Manuel Garcia-Carpintero, Mikkel Gerken, Olav Gjelsvic, Emma Gordon, John Greco, Alex Gregory, Thomas Grundmann, Josh Habgood-Coote, Michael Hannon, Anandi Hattiangadi, Katherine Hawley, Frank Hofmann, Joachim Horvath, Christoph Jaeger, Jonathan JenkinsIchikawa, Matt Jope, Jesper Kallestrup, Klemens Kappel, Dirk Kindermann, Igal Kvart, Christos Kyriacou, Jennifer Lackey, Maria Lasonen-Aarnio, Janet Levin, Max Lewis, Clayton Littlejohn, Fed Luzzi, Jack Lyons, Susanne Mantel, Deborah Marber, Aidan McGlynn, Matt McGrath, Conor McHugh, Robin McKenna, Anne Meylan, Lisa Miracchi, Luca Moretti, Nikolaj Nottleman, Jonas Olson, Erik Olsson, Orestis Palermos, Carlotta Pavese, Andy Peet, Tommaso Piazza, Christian Piller, Duncan Pritchard, Andrew Reisner, Sven Rosenkrantz, Lewis Ross, Sarah Sawyer, Susanne Schellenberg, Johanna Schnurr, Nick Shackel, Susanna Siegel, Martin Smith, Justin Snedegar, Ernie Sosa, Rachel Sterken, Ema Sullivan-Bisset, Kurt Sylvan, Alessandra Tanesini, Jonathan Way, John Webber, Alan Weir, Asa Wikforss, Daniel Whiting, and Robbie Williams for the very helpful discussions. I am sure that I have missed people who deserve to be acknowledged by name and hope

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

viii



they will accept my apologies. Many thanks also to two anonymous referees for Oxford University Press who gave me excellent and very detailed comments on the book. I would also like to thank the organizers and audiences of the Philosophy Senior Seminar 2020, University of Glasgow; the University of St Andrews Philosophy Society 2020; the Joint Session of the Aristotelian Society and the Mind Association 2019, Durham University; the 2017 Meeting of the European Normativity Network, University of Stockholm; the Assertion: Foundational Issues Workshop, University of Oslo, 2017; the British Society for Ethical Theory conference 2017, University of Bristol; The Joint Session of the Aristotelian Society and the Mind Association 2017, University of Edinburgh; the 2nd LeuvenSouthampton Epistemic Normativity Workshop, KU Leuven; the 3rd PLM Masterclass with Timothy Williamson, Institute of Philosophy, University of London, 2016; the 2nd Cologne-Leuven Epistemology Workshop, KU Leuven, Belgium, 2016; the Leuven Epistemic Norms Conference, KU Leuven, 2015; the Postgraduate Sessions of the Joint Session of the Aristotelian Society and the Mind Association 2015, University of Warwick; the 2015 Bled Philosophical Conferences, Slovenia; the 4th Belux Colloquium in Epistemology, University of Luxembourg, 2015; the 3rd Belux Colloquium in Epistemology, KU Leuven, 2015; the Graz Young Epistemology and Philosophy of Logic Workshop, University of Graz 2014; the Yonsei Philosophy Conference, Seoul, 2014; the 1st Belux Colloquium in Epistemology, KU Leuven, 2014. A few chapters of the book draw on previously published work: Chapters 1 and 5 draw on my 2019 Mind & Language article ‘Assertion: The Context Shiftiness Dilemma’; Chapter 2 relies on my 2018 Synthese article ‘A Puzzle for Epistemic WAMs’; Chapter 6 draws on parts of ‘Saying and Believing: The Norm Commonality Assumption’ (Philosophical Studies 2018). Finally, Chapter 10 draws on my 2018 Ethical Theory and Moral Practice article ‘The Explanation Proffering Norm of Moral Assertion’. I want to thank the editors of these journals as well as the anonymous referees who provided me with excellent comments on these articles. Finally, many thanks to Peter Momtchiloff at Oxford University Press for all his excellent advice, support, and efficiency.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

Contents Introduction

xi

I . T H E TH O U G H T / S P E E C H S H I F T I N E S S D I L E MM A 1. The Context Shiftiness Dilemma Generalized 1.1 Classical Invariantism and the Knowledge Norm of Assertion 1.2 Contextualist Cases and the Shiftiness Dilemma 1.3 The Shiftiness Dilemma Generalized

2. Epistemic WAMs 2.1 Assertion Sensitivism and Value/Norm Type Association 2.2 Sensitive Epistemic Goals 2.3 Sensitive Discursive Justification

3. Pragmatic WAMs 3.1 3.2 3.3 3.4

Pragmatic Compatibilism Shiftiness without Error Possibilities Shiftiness without Stakes Shiftiness and Relevance

4. KK Compatibilism 4.1 Knowledge of Knowledge 4.2 The Normative Status of the Blameworthiness Claim 4.3 Knowledge of Knowledge of Knowledge

3 4 8 13

18 19 24 27

32 33 34 36 38

40 40 42 48

II. THOU GHT IN VARIANT ISM AN D SPEECH FUNCTIONALISM 5. Against the Shiftiness Dilemma 5.1 Against Conditional Incompatibility 5.2 Epistemic Norms and Norms with Epistemic Content 5.3 No Shiftiness Dilemma

57 58 61 65

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

x



6. Assertion Functionalism and Context 6.1 6.2 6.3 6.4 6.5

Etiology A Functionalist Account of Assertion Knowledge vs. Truth Overriding Functions Contextualist Rejoinders

7. Knowledge Norms for Constative Speech Acts 7.1 Bach and Harnish’s Taxonomy 7.2 Normative Transfer 7.3 Two More Arguments for Knowledge Norms for Constatives

68 68 73 77 79 83

87 88 91 95

I I I . C O N S T A T I V E S I N CO N T E X T 8. The Epistemic Normativity of Conjecture 8.1 8.2 8.3 8.4 8.5

103

Conjecture Orthodoxy Against Conjecture Orthodoxy The Knowledge Norm of Conjecture The Knowledge Norm vs. the Credence Norm Practical Shiftiness and Scientific Revolutions

104 107 113 118 126

9. No Special Treatment for the Epistemic Normativity of Telling

131

9.1 Telling as Directive or Commissive 9.2 Pelling’s Synchronic Different Stakes

10. A Special Case: Moral Assertion

132 135

141

10.1 Testimony and Moral Understanding 10.2 The Function and Norm of Moral Assertion 10.3 Response to Critics

142 145 149

Conclusion

153

References Index

157 163

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

Introduction This is an essay in epistemology and the philosophy of language. It concerns epistemology in that it is a manifesto for epistemic independence: the independence of good thinking from practical considerations. It defends the independence of thought from the most prominent threat that has surfaced in the last twenty years of epistemological theorizing: the phenomenon of shiftiness of proper assertoric speech with practical context. It concerns philosophy of language in that it defends a functionalist account of the normativity of assertion in conjunction with an integrated view of the normativity of constative speech acts. Roughly, the foe of independence argues as follows: proper assertoric speech and proper thought go hand in hand. Intuitively, proper assertoric speech is shifty with practical stakes. Therefore, proper thought is shifty with practical stakes. This book argues that the foe’s argument does not go through, and that shiftiness of assertoric speech is exactly what we should expect in a picture where epistemically proper thought is practical-context-invariant. To do that, it develops a view of the normativity of constatives sourced in their epistemic function, which neatly delivers the goods: shiftiness of proper assertoric speech in conjunction with independence of proper thought from practical considerations. The book does three main things: First, it puts forth a new, generalized dilemma for practical-contextindependent views of epistemic normativity. I argue that the foe’s argument against the independence of proper thought from the practical threatens to dangerously generalize in three ways: first, it threatens to generalize to all context-invariant norms of assertoric speech. Second, it threatens to generalize beyond mere assertives, to all constative speech acts. Third, it threatens to generalize beyond knowledge, to epistemic

Shifty Speech and Independent Thought: Epistemic Normativity in Context. Mona Simion, Oxford University Press (2021). © Mona Simion. DOI: 10.1093/oso/9780192895288.001.0001

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

xii



normativity as a whole. All in all, I argue, the foe’s argument is more dangerous for epistemic independence than we took it to be. Second, the book looks at extant views that attempt to preserve the independence of thought in the face of assertion’s practical shiftiness and argues they remain ultimately unsuccessful. First, Assertion Sensitivism— embracing a context-variant view of the normativity of assertion—is shown to be strongly incompatible with uncontroversial value theoretic claims, and thus to be rejected on prior plausibility grounds. Second, pragmatic warranted assertability manoeuvres are shown to fail to generalize beyond assertions featuring knowledge attributions. Last, views appealing to higher-order knowledge to explain away the data threatening the independence of proper thought are shown to fail on normative grounds. The third and central aim of the book is to go on a rescue mission on behalf of independence. To this effect, I first show that, against orthodoxy, the foe’s argument from practical shiftiness of proper assertoric speech against the independence of proper thought from the practical does not go through, for it rests on normative ambiguation. Second, I defend a non-shifty, functionalist account of the epistemic normativity of assertion. Third, I generalize my account to all constative speech and I develop the first integrated account of the epistemic normativity of constatives in the literature. Last, I defend my view for the particular cases of conjecturing, telling, and moral assertion. While the book hangs nicely together for someone with my knowledge-centric, purist taste in epistemic normativity, several results in this book also hold independently of it. For instance, one need not be a knowledge norm champion, nor a Classical Invariantist, in order to accept that the argument for the incompatibility of the two doesn’t work, due to normative ambiguation. Indeed, foes of Classical Invariantism can take this result as a challenge, while defenders of other non-shifty norms (e.g. justification norms) can just enjoy the good news, since the result generalizes. Also, one need not like the knowledge norm in order to like and embrace the functionalist picture of the normativity of constatives defended here: keeping the functionalist structure fixed, a different view of the function may deliver a different normative account. Last but not least, the results in the third part of the book can survive even

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi



xiii

readers who are not inclined to buy into anything that happened in the first two parts, since the accounts there are defended on largely independent grounds. As some readers may be interested in different parts of the book, in what follows I give a brief sketch of its contents.

Game Plan Chapter 1: The Context Shiftiness Dilemma Generalized. Most actors in the debate assume that the intuitive variability of proper assertion with practical stakes (the Shiftiness Intuition) motivates the following dilemma: either 1) we embrace the Knowledge Norm of Assertion (KNA), and are forced into a view that takes knowledge, or ‘knowledge’, to be sensitive to practical stakes, or 2) we stick to our Classical Invariantist guns, but then the knowledge norm goes out the window and we get practical sensitivity in the normativity of assertion (the Shiftiness Dilemma). This chapter brings this dilemma to centre stage and argues that it threatens to generalize in three ways: to all contextinvariant norms of proper assertoric speech, to all constative speech acts, and to all epistemic normative constraints. If this is right, we should be more worried about this dilemma than we’ve been so far. The following three chapters look at views that attempt to safeguard the independence of thought in the face of assertion’s practical shiftiness. Chapter 2: Epistemic WAMs. This chapter looks into the feasibility of preserving the independence of epistemically proper thought from practical considerations via the second horn of the Shiftiness Dilemma. For people who like Classical Invariantism about knowledge attributions, the jump from variation in assertability with stakes to contextualism or pragmatic encroachment seemed rushed. As such, these authors venture to account for the Shiftiness Intuition under a Classical Invariantist umbrella by arguing for context-sensitivity of proper assertability. This chapter argues that the view fails on prior plausibility due to being incompatible with the following highly uncontroversial value-theoretic claim: norms of type X are associated with values of type X.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

xiv  Chapter 3: Pragmatic WAMs. In light of embracing the Shiftiness Dilemma, the vast majority of philosophers accept Assertion Incompatibilism: given intuitive variability of proper assertion with practical stakes, Classical Invariantism is incompatible with a biconditional knowledge norm of assertion. There are also a few dissenting voices, however: some invariantists venture to preserve both Classical Invariantism and the Knowledge Norm of Assertion (Assertion Compatibilism). There are two varieties of Compatibilism available on the market. This chapter discusses the first of the two: Pragmatic Compatibilism employs a pragmatic warranted assertability manoeuvre to explain away the shiftiness data. The chapter argues that the view has difficulties with delivering epistemic normative independence for cases of shiftiness of assertions that do not involve knowledge attributions or explicitly tabled error possibilities. Chapter 4: KK Compatibilism. According to KK Compatibilism, the unassertability in the high stakes contextualist cases can be explained in terms of the subjects’ lack of higher-order knowledge: Although, strictly speaking, all that is needed for proper action—assertion included—is first-order knowledge, when the stakes are high, we tend to find people who act without knowing that they meet the condition for proper action blameworthy for so doing. This chapter argues that (1) the view misidentifies the epistemic deficit that is explanatorily salient in contextualist cases, in that the absence of second-order knowledge is not a difference maker, and (2) on closer look, the account also requires normative finessing for extensional adequacy. Chapter 5: Against the Shiftiness Dilemma. This chapter argues that the Shiftiness Dilemma is a false dilemma: The Knowledge Norm of Assertion is perfectly compatible with Classical Invariantism. Furthermore, it offers independent reason to believe that if KNA and Classical Invariantism are true, variation in proper assertability is exactly what we may expect. More precisely, the chapter advances the debate in two important ways: (1) it identifies a widely assumed assumption concerning epistemic norm individuation (Content Individuation), which gets the Shiftiness Dilemma off the ground; (2) it argues that Content Individuation is false, and that therefore the norm at stake in the debate need not be epistemic.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi



xv

Chapter 6: Assertion Functionalism and Context. This chapter shows how to reconcile Classical Invariantism with the Knowledge Norm of Assertion. My basic proposal is that we can combine invariantism with a functionalist account of assertion: according to the account I favour, assertion is governed by a knowledge norm in virtue of its epistemic function of generating testimonial knowledge. Requirements generated by other functions of assertion, though, such as its prudential function, can override the constraints imposed by the epistemic function, and render the knowledge requirement either too strong or too weak for all-things-considered permissible assertion. All-things-considered permissible assertion can vary with practical stakes; epistemically permissible assertion does not. Chapter 7: Knowledge Norms for Constative Speech Acts. This chapter takes up an ambitious task: it develops the first integrated account of the epistemic normativity of constatives. In order to do that, it argues for a generalized knowledge-based account of the epistemic normativity of constative speech acts, and it develops the corresponding accounts for, respectively, assertives, predictives, retrodictives, descriptives, ascriptives, informatives, confirmatives, concessives, retractives, assentives, dissentives, disputatives, responsives, suggestives, and suppositives. The chapter argues for a knowledge account from three different angles: (1) the nature of communicative speech acts, (2) the relation between assertion and other constatives, and (3) the normativity of belief together with constatives’ epistemic function. In the next three chapters, I look more closely into the epistemic normativity of three constatives—conjecture, telling, and moral assertions—and I examine and account for the contextual variation of their propriety. Chapter 8: The Epistemic Normativity of Conjecture. This chapter has two aims: (1) it breaks new ground in that it develops the first fully fledged account of the epistemic normativity of conjecture in the literature; (2) it goes sharply against orthodoxy, in arguing that the epistemic requirements placed on us by conjecture are stronger than those governing assertion. According to the view developed here, one’s conjecture that p is permissible only if one knows that one has warrant, but not sufficient warrant to believe that p.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

xvi  Chapter 9: No Special Treatment for the Epistemic Normativity of Telling. While recent years have featured a vast amount of literature concerned with the epistemic norm for assertion, comparatively little attention has been paid to the corresponding norm governing acts of telling. One plausible explanation for this is that people have generally taken assertion and telling to fall under the same normative constraints. Recent work, however, ventures to show (1) that this assumption is false, and (2) that the epistemic propriety of instances of telling partly depends on what’s at stake for the hearer. This chapter argues that the case against normative commonality for assertion and telling fails due to speech act theoretic and value theoretic inaccuracies. In a nutshell, the chapter argues that there is nothing special about the epistemic normativity of telling. Chapter 10: A Special Case: Moral Assertion. This chapter is concerned with moral assertion. In recent years much attention has been given to the epistemic credentials of belief based on moral testimony. Some people think pure moral deference is wrong; others disagree. It comes as a surprise, however, that while the epistemic responsibilities of the receiver of moral testimony have been closely scrutinized, little to no discussion has focused on the epistemic duties of the speaker. This chapter aims to supply this lack: it defends a functionalist account of the normativity of moral assertion. According to this view, in virtue of its function of reliably generating moral understanding in the audience, a moral assertion that p needs be knowledgeable and accompanied by a contextually appropriate explanation why p.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

PART I

THE THOUGHT/SPEECH SHIFTINESS DILEMMA

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

1 The Context Shiftiness Dilemma Generalized There are good and bad assertions. It often feels natural to say things like ‘Good that you said so!’ but also to criticize speakers by asking ‘Why would you say such a thing?’ when their assertions seem inappropriate. On top of this, it looks as though assertions can be good or bad in a distinctively epistemic way: notions such as knowledge or justified belief figure prominently in our evaluations of each other’s speech acts. We do, for instance, challenge assertions by ‘Do you know that?’ questions, when we suspect the speaker does not find herself in a strong enough epistemic position to assert. In the light of this, it comes as no surprise that epistemic normative constraints on assertion are of central concern in recent literature. Also, we should expect questions pertaining to the nature of the corresponding epistemic standings to carry significant weight when it comes to the normativity of assertion. Now, here are two attractive theses, both of which enjoy a great deal of popularity:¹ The Knowledge Norm of Assertion (KNA): epistemically permissible iff one knows that p.

One’s assertion is

And Classical Invariantism (CI): The truth value of knowledge claims is insensitive to practical matters.

¹ For support for KNA, see e.g. DeRose (2002), Hawthorne (2004), Kelp and Simion (Forthcoming), Slote (1979), Turri (2010), Unger (1975), Williamson (2000, 2005). Shifty Speech and Independent Thought: Epistemic Normativity in Context. Mona Simion, Oxford University Press (2021). © Mona Simion. DOI: 10.1093/oso/9780192895288.003.0001

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

4

    

I, alongside many other epistemologists, care a lot about these two claims. Plausible (and popular) as they might be, however, CI and KNA are widely taken to be incompatible. Here is one prominent argument for their incompatibility, by Keith DeRose (2002): intuitively, standards for proper assertability vary across practical contexts: in highstakes but not in low-stakes scenarios, more warrant seems to be needed for one to be in a good enough epistemic position to assert (henceforth also the Shiftiness Intuition). Given this, we cannot have an invariant standard for proper assertability: a knowledge norm of assertion is incompatible with Classical Invariantism. Since we need to place the source of the felt sensitivity somewhere, one of the two has to go. In the light of this, most² actors in the debate assume that the intuitive variability of proper assertion with practical stakes motivates the following dilemma: either 1) we embrace KNA, and are forced into a view that takes knowledge, or ‘knowledge’, to be sensitive to practical stakes (henceforth, Knowledge Sensitivism, or SK), or 2) we stick to our Classical Invariantist guns, but then KNA goes out the window and we get practical sensitivity in the normativity of assertion (henceforth Assertion Sensitivism, or SA). Let us dub this the Shiftiness Dilemma. This chapter brings this dilemma to centre stage and argues that it threatens to generalize in three ways: to all speech act epistemic norms with epistemic content, to all constative speech acts, and to all epistemic normative constraints. If this is right, we should be more worried about this dilemma than we’ve been so far.

1.1 Classical Invariantism and the Knowledge Norm of Assertion I mentioned that CI and KNA enjoy quite a bit of popularity; in fact, if ever there have been ‘mainstream’ views in epistemology, it is fair to say that CI and KNA are amongst them. Here is why, very roughly: ² For an explicit statement of this dilemma, see DeRose (2002). For people embracing the first horn thereof, see e.g. DeRose (2002), Hawthorne (2004). For champions of the second horn, see e.g. Brown (2010), Gerken (2012), Goldberg (2015), Greenough (2011). For work resisting the dilemma see Williamson (2005), Brown (2006), Reed (2010, 2012), Hazlett (2009), Rysiew (2001), Simion (2016, 2018, 2020a), Kelp and Simion (Forthcoming).

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

    

5

For CI: we have good reason to care about the independence of thought from practical concerns: first, on a light note, I am not the first to notice that there’s something amiss about the idea of coming to know/ coming to be the appropriate subject of a knowledge attribution more easily just in virtue of you or others not caring much about the subject matter. Second, a lot hinges on CI being true for our practice of forming beliefs, as well as for our epistemic endeavours more generally. CI is not merely a claim about the semantics of ‘know that’; it is the core claim at the heart of the independence of thought from practical considerations: if CI is false, practical normativity threatens full encroachment on epistemic normativity (more on this in the next section). If we care about independence, we care about CI. Third, friends and foes alike agree that CI is the default epistemological position: we need to be argued out of it. Here are, for instance, Jeremy Fantl and Matt McGrath on this: ‘Anyone who has even passing knowledge of analytic epistemology in the last fifty or so years knows that moderate invariantism is the orthodoxy. It is the view to beat’ (Fantl and McGrath 2009, 37). Given its centrality to epistemological affairs, the argument against CI on offer had better be good. Last, going back to our practice of inquiry: arguably, the denial of CI would make inquiry into an odd practice in the normative landscape. Here is, very quickly, why; even if we agree with the pragmatist that the ultimate aim of all of our practices is prudential—something like survival, or success—direct reduction of epistemic success to prudential success is implausible. First, it fails extensionally: for instance, uncontroversially, wishful thinking is often prudentially great, and always epistemically bad. Second, it fails on prior plausibility: think of other practices more obviously aimed at our survival, and how their normativity works. Take a paradigmatic case: the practice of medicine. Plausibly, medicine aims at generating health. In turn, plausibly, health is good for our survival. Importantly, though, although the ultimate aim of engaging in the practice is survival, it is not the case that anything goes insofar as the ultimate aim is achieved. Luckily true diagnoses, for instance, are in breach of the norms of the practice of medicine: doctors are not engaging in proper medical behaviour if they toss a coin to determine what’s wrong with us, even if they get it right. What seems

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

6

    

to be the case is that the practice of medicine is internally normatively organized: moves in the practice are regulated by internal norms, which are reliably conducive to success in the practice: generating health. In turn, reliably generating health is a reliable path towards the ultimate prudential aim: survival. This picture is reductive, but indirectly so: norms of medicine do not reduce to survival norms directly, in that not any move in medicine that generates survival is a permissible move in medicine. Rather, for a move in medicine to be permissible, it needs to obey norms internal to the domain. Similarly, we should expect that epistemic normativity is not directly reductive: believing what’s prudentially good to believe is not to believe epistemically permissibly. Rather, our epistemic practice is normatively organized around its internal, epistemic aim—generating knowledge— which, in turn, is good for our survival. The norms governing our epistemic practices, then, will be norms ensuring the internal aim of the practice is reliably achieved: they will be directly sourced in the goal of achieving knowledge, not in prudential goals.³ For KNA: several people think⁴ (and I have defended these claims in several places, e.g. Simion (2019c, 2019a, 2020a)) that knowledge is the goal of our practice of inquiry. That is because it is (uncontroversially) more valuable than lesser epistemic standings, and, at the same time, an easy get: for the kind of creatures we are, both lesser epistemic standings (e.g. lucky true belief) and stronger epistemic standings (e.g. certainty) are very hard to acquire. In a nutshell, knowledge is just our way of representing the world: it is the easiest and most valuable epistemic standing that is achievable for the kind of creatures we are. Now, plausibly enough, assertion is a move in our practice of inquiry, in much the same way in which moving the rook is a move in the practice ³ Ernie Sosa (2007) has a similar value-theoretic picture: in Sosa’s view, there are such things as critical domains, which are associated with goods or values for that domain. Any critical domain has at least one central value. This central value is fundamental to the domain in the sense that the central value is valuable for its own sake, relative to the domain. Crucially, central values organize the evaluations in that domain in the sense that all other domain-specific values and norms will be derivative from the central ones: they can be explained fully in terms of the central values. I think of practices and their intrinsic functions along similar lines: practices are critical domains and their intrinsic functions are values for their own sake relative to the domain in question. ⁴ E.g. Kelp (2014), Williamson (2000).

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

    

7

of chess.⁵ In turn, moves in practices are aimed at fulfilling the goal of the practice:⁶ moves in chess aim at checkmate. Since moves in practices aim at fulfilling the goal of the practice and assertion is a move in the practice of inquiry, and on the assumption that the aim of the practice of inquiry is to generate knowledge, it follows that assertion aims at generating knowledge. Last, since the most reliable way for assertion to do so is by being knowledgeable, we have good reason to believe the norm governing assertion is a knowledge norm. Second, if knowledge is both necessary and sufficient for epistemically permissible assertion, we have a very straightforward and elegant way of explaining quite a few otherwise puzzling linguistic data, such as: (1) the paradoxical soundingness of Moorean statements of the form ‘p but I don’t know that p’. If knowledge is the norm of assertion, on the plausible assumption that knowledge distributes across conjunctions, one is in a good enough epistemic position to assert ‘p but I don’t know that p’ only if one knows both conjuncts. However, since knowledge is factive, one only knows the second conjunct if it is true that one does not know that p. But that contradicts knowing the first conjunct. Thus, according to KNA, one cannot meet the conditions for making a proper assertion of the form ‘p but I don’t know that p’ (Williamson 2000, 253). Furthermore, if KNA is true, it is plausible that when the speaker asserts that p, the hearer is led to believe that she ⁵ Williamson (2000) relies on the chess model to argue that KNA is uniquely constitutive of the practice of assertion. I have discussed this claim in several places, including at book length (e.g. Kelp and Simion 2020, Kelp and Simion Forthcoming). For the purposes of this book, I set the constitutivity claim aside. For excellent discussion see also Cappelen (2011), Maitra (2011), and Garcia-Carpintero (2004, 2019). ⁶ To be more precise, moves in practices aim either directly or indirectly to fulfil the function of the practice. The difference between direct and indirect aiming lies with achievability: for some moves in practices, the goal of the practice is only indirectly rather than directly reachable. Diabetological consults are moves in the practice of medicine, and they aim directly at fulfilling the goal of the practice of medicine: curing diseases. In turn, some moves aim at this final practice goal only indirectly, while aiming directly at intermediate goals: performing glycemy tests aims (directly) at informing the diabetologist as to how well the patient is doing, which, in turn, aims directly at her diagnosing the patient correctly, and, further, at curing her disease. In this, glycemy tests aim indirectly at the goal of the practice of medicine: curing diseases. They aim at making progress towards it. Similarly, baking cakes is a move in the practice of cooking, and it aims directly at fulfilling the aim of the practice: producing tasty, nourishing food. My getting the flour off the shelf aims indirectly at the general aim of the practice by aiming directly at adding flour to the cake mix. It aims at making progress towards producing tasty, nourishing food. And so on. See Simion (2020a) for a more detailed picture.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

8

    

knows that p;⁷ therefore, when one asserts the second conjunct of a Moorean statement, one denies what one has led one’s hearer to believe by asserting the first conjunct (Moore 1962, 277). Thus, KNA explains why the conjunction ‘p but I don’t know that p’ is not only unassertable, but it also sounds paradoxical. KNA also neatly explains the intuitive unassertability of lottery propositions (Williamson 2000); after all, lottery propositions cannot be known. Last but not least, KNA makes sense of the ubiquitous and intuitively permissible ‘How do you know?’ challenges (i.e. the necessity direction) and intuitively permissible criticisms such as ‘Why didn’t you say so? You knew it all along!’ (i.e. the sufficiency direction).

1.2 Contextualist Cases and the Shiftiness Dilemma In spite of the popularity of CI and KNA, those who want to embrace both claims⁸ are widely believed to face the Shiftiness Dilemma. The Shiftiness Dilemma arises from the Shiftiness⁹ Intuition: The Shiftiness Intuition: Intuitively, permissible assertion requires more epistemic support in high-stakes contexts than in low-stakes contexts. Proper assertion is harder when the stakes are higher in the sense that it appears to require more warrant. The Shiftiness Intuition is not exactly breaking news in epistemology. Austin (1946), for instance, observes that, while in normal contexts the fact that your hat is in the hall seems to be good enough reason for me to say that you are in, when a

⁷ In what sense does asserting p lead the hearer to believe that the asserter knows that p? The received view is that it is a matter of ‘presenting oneself ’ as knowing. See Simion (2020a) for a more detailed picture. ⁸ The ambition of this book is limited to discussing KNA taken as a distinctively epistemic norm. Also, again, it falls outside of the scope of this book to inquire into whether said epistemic norm is also constitutive of assertion. See, however, Kelp and Simion (Forthcoming) and Simion (2020a) for this. ⁹ To my knowledge, the term was coined in Fantl and McGrath (2012).

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

    

9

lot hinges on it, I would be reticent to do the same. That said, it came into popularity when epistemic contextualists started using it in support of their view. Epistemic contextualism is a semantic thesis about attributions of ‘knowledge’: it holds that the propriety of ‘knowledge’ attributions varies with context. It is typically supported by pairs of cases like the following ‘bank cases’ by Keith DeRose (1992, 913): BANK CASE LOW. My wife and I are driving home on a Friday afternoon. We plan to stop at the bank on the way home to deposit our pay cheques. But as we drive past the bank, we notice that the lines inside are very long, as they often are on Friday afternoons. Although we generally like to deposit our pay cheques as soon as possible, it is not especially important in this case that they be deposited right away, so I suggest that we drive straight home and deposit our pay cheques on Saturday morning. My wife says, ‘Maybe the bank won’t be open tomorrow. Lots of banks are closed on Saturdays.’ I reply, ‘No, I know it’ll be open. I was just there two weeks ago on Saturday. It’s open until noon.’ BANK CASE HIGH. My wife and I drive past the bank on a Friday afternoon, as in Case A, and notice the long lines. I again suggest that we deposit our pay cheques on Saturday morning, explaining that I was at the bank on Saturday morning only two weeks ago and discovered that it was open until noon. But in this case, we have just written a very large and very important cheque. If our pay cheques are not deposited into our chequing account before Monday morning, the important cheque we wrote will bounce, leaving us in a very bad situation. And, of course, the bank is not open on Sunday. My wife reminds me of these facts. She then says, ‘Banks do change their hours. Do you know the bank will be open tomorrow?’ Remaining as confident as I was before that the bank will be open then, still, I reply, ‘Well, no, I don’t know. I’d better go in and make sure.’ Let’s assume that the bank will, in fact, be open on Saturday and so DeRose’s corresponding belief is true in both BANK CASE LOW and BANK CASE HIGH. And while DeRose’s attribution of knowledge that the bank is open tomorrow is intuitively fine in BANK CASE LOW, the

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

10      same would not hold for BANK CASE HIGH. Contextualism has a straightforward explanation of these data: the truth conditions of knowledge attributions vary with context with the result that, in BANK CASE LOW, DeRose’s attribution of knowledge is true, whereas in BANK CASE HIGH it is false. Now, here is one important difficulty for epistemic contextualism, which DeRose himself recognizes and dubs the Generality Objection: the Shiftiness Intuition does not only arise for knowledge attributions. Rather, it is a much more general phenomenon. For instance, in BANK CASE LOW, it is not only fine for DeRose to assert that he knows that the bank will be open tomorrow, it would also be fine for him to assert that the bank will be open tomorrow. Similarly, in BANK CASE HIGH, it would not only seem wrong for DeRose to assert that he knows that the bank will be open tomorrow, but also to assert that the bank will be open tomorrow (DeRose 2002, 177). What these considerations suggest is that the Shiftiness Intuition is not limited to cases of knowledge attributions. As a result, the worry is that epistemic contextualism won’t give us the right explanation of the Shiftiness Intuition. In order to address this worry, contextualists have appealed to the Knowledge Norm of Assertion. Roughly, here is the thought: according to KNA, one’s assertion that p is epistemically proper if and only if one knows that p. If that is the case, however, it follows that the standards for knowledge go hand in hand with the standards for proper assertion. But now it looks as though epistemic contextualism is well positioned to avoid the Generality Objection. After all, the objection suggests that the standards for proper assertion vary in general (that is, not just in cases featuring knowledge attributions). But if KNA and contextualism are both true, this is just what we’d expect anyway. In fact, KNA and contextualism in conjunction promise to offer an appealing explanation of the relevant data. To see this, consider first a variation of BANK CASE LOW in which DeRose asserts that the bank is open. Suppose that the bar for ‘knowledge’ is low here, low enough that DeRose ‘knows’ what he asserts and so his assertion comes out proper. In contrast, suppose DeRose were to assert that the bank is open in a variation of BANK CASE HIGH. Suppose that the bar for ‘knowledge’ is high here, high

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

    

11

enough that DeRose doesn’t ‘know’ what he asserts and so his assertion would come out improper. In this way, KNA and contextualism together can offer an appealing solution to the generality problem. To see this more clearly, consider the following simple assertion cases, not involving knowledge attribution: ASPIRIN-LOW. On the way to your place, your sister lets you know that she has a minor headache and asks you: ‘Do you have aspirin at home or should we stop at the pharmacy?’ You remember having bought aspirin last month and tell her: ‘Don’t worry, we have aspirin at home.’ ASPIRIN-HIGH. On the way to your place, your sister lets you know that her two-year-old baby urgently needs an aspirin on doctor’s orders and asks you: ‘Do you have aspirin at home, or should we stop at the pharmacy?’ You remember having bought aspirin last month. Even so, you think it best to have your partner double check and so you call. When you find out that s/he is not at home you say to your sister: ‘Well, let’s drop by the pharmacy, just in case.’ In ASPIRIN-LOW you assert that you have aspirin at home. What’s more, your assertion is intuitively entirely proper. In contrast, you are unwilling to do so in ASPIRIN-HIGH. And, crucially, had you made the same assertion here, intuitively it wouldn’t have been proper. In this way, there is variation in the intuitive propriety of assertion. In turn, DeRose argues, this Shiftiness Intuition gives rise to the Shiftiness Dilemma. Here is DeRose: If the standards for when one is in a position to warrantedly [i.e. properly] assert that P are the same as those that constitute a truth condition for ‘I know that P,’ then if the former vary with context, so do the latter . . . What of the advocate of the knowledge account of assertion who does not accept contextualism? Such a character is in serious trouble. Given invariantism about knowledge, the knowledge account of assertion is an untenable attempt to rest a madly swaying distinction upon a stubbornly fixed foundation. (2002, 182)

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

12      DeRose’s argument promises to establish the entailment between KNA and the denial of Classical Invariantism. In other words, DeRose does appear to establish at least the following: Conditional Incompatibility. If the Shiftiness Intuition is correct, then KNA is incompatible with Classical Invariantism. Since we have already seen that there is excellent reason to think that the Shiftiness Intuition is correct, we have equally good reason to think that KNA is indeed incompatible with Classical Invariantism. And that, of course, is all we need to force the Shiftiness Dilemma on those who embrace both KNA and Classical Invariantism: they’ll have to give up one of the two. Many, if not most, contributors to the debate accept the Shiftiness Dilemma and grasp one of its two horns: the view that grasps the first horn of the dilemma (Assertion Sensitivism, SA) allows for variability in what it takes to comply with the epistemic norm of assertion. This view comes in more than one variety; first there are people thinking that assertion is governed by one norm that stipulates that the appropriate amount of warrant for proper assertion varies with contextual features (e.g. Brown 2010, Gerken 2012, Goldberg 2015, McKinnon 2013, Rescorla 2009). Another way to be a sensitivist about assertion is to stipulate several norms governing assertion depending on the context (e.g. Greenough 2011, Levin 2008, Stone 2007). In contrast, we have seen that the view that goes for the second horn (Knowledge Sensitivism, SK) allows for variability in what it takes to know or ‘know’. Pragmatic encroachment (e.g. Hawthorne 2004, Fantl and McGrath 2009, Stanley 2005) takes practical sensitivity to be a feature of knowledge itself, and thus the relevant stakes to be those associated with the subject of the knowledge claims. In contrast, we have already seen that contextualists (e.g. Cohen 1999, DeRose 2002) argue that it pertains to knowledge attribution, and thus it varies with the stakes of the attributor. In light of embracing the Shiftiness Dilemma, then, the vast majority of philosophers accept Assertion Incompatibilism: given intuitive variability of proper assertion with practical stakes, Classical Invariantism is incompatible with a biconditional knowledge norm of assertion.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

    

13

1.3 The Shiftiness Dilemma Generalized The dilemma identified in the previous section worries many epistemologists. This is evidenced by the enormous literature published on the issue in the last two decades, as well as by the fact that, arguably, the Shiftiness Dilemma has given rise to a revival of epistemic contextualism, as well as pragmatic views of the epistemic. I will now argue that we should be even more worried than we are about the Shiftiness Dilemma. In order to do that, I will argue that the Shiftiness Dilemma (SD) threatens to generalize in three dangerous ways. First, it threatens to generalize beyond KNA to all context-independent views of the epistemic normativity of assertion that feature epistemic content. SD-Generalized 1 (henceforth SDG1): If SD holds for KNA, then it holds for all context-independent norms of assertion with epistemic content, i.e. norms that stipulate that an epistemic condition is necessary and sufficient for proper assertion.¹⁰ To see why this is so, remember the main source of SD: the Shiftiness Intuition in conjunction with KNA. Note, though, that the Knowledge Norm is not required to generate SD: all context-independent norms with epistemic content will do in conjunction with the Shiftiness Intuition. Consider, for instance, the main competitor of KNA on the market, the Justification Norm of Assertion (henceforth JNA, defended most notably by Igor Douven (2006) and Jennifer Lackey (2007)). Crucially, on JNA, the relevant J is meant to stand for the level of justification necessary for knowledge. On a JNA picture, what the Shiftiness Intuition about proper assertability suggests is that either (1) knowledge-level justification varies with practical stakes, or (2) JNA is false.

¹⁰ What is meant here by norms with epistemic content is norms that stipulate an epistemic condition for proper assertion (such as knowledge, or justification) rather than, e.g., merely an alethic condition (i.e. the Truth Norm of Assertion, defended e.g. in Weiner 2005 and Whiting 2013).

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

14      Second, the Shiftiness Dilemma threatens to generalize beyond mere assertoric speech to all constatives: SD-Generalized 2 (henceforth SDG2): holds for all constative speech acts.

If SD holds for assertion, then it

Constative speech acts are, broadly speaking, speech acts that characteristically express a (particular degree of ) belief that the content holds. Bach and Harnish (1979) notably offer an excellent taxonomy of constatives¹¹ that will be very useful to illustrate the issue (see also Chapter 7 for a more detailed discussion). For our purposes here, it will be enough to only discuss a couple of examples, in order to see how this works. Take, for instance, the retrodictive constative speech act of reporting. According to Bach and Harnish, reporting expresses the belief that it was the case that p. If so, we can think of corresponding epistemic norms for reporting. For instance, a knowledge norm for reporting will read as follows: Knowledge Norm for Reporting (KNR): One’s reporting that p is epistemically permissible iff one knows that it was the case that p. Intuitively, though, proper reporting is shifty with practical context: it seems intuitively right that more epistemic support is needed when reporting on, say, whether the bridge in the centre of town was found at risk to collapse or not than, say, for reporting on whether a particular movie star was wearing red at the opening or not. It is then easy to see how the Shiftiness Dilemma will thus generalize to the speech act of reporting: given intuitive shiftiness of proper reporting with practical stakes, either KNR is false, or knowledge is practically shifty. Take, also, a weaker variety of a constative speech act, the speech act of conjecturing. On Back and Harnish’s view, conjecturing falls in the category of suggestive constatives, which express the belief that there is

¹¹ Bach and Harnish (1979, 40). Bach and Harnish think that constatives also express an intention on the part of the speaker that the hearer form or continue to hold the corresponding belief. This is of little consequence for us here. I will bypass discussing it for now.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

    

15

reason, but not sufficient reason, to believe that p. If this is so, it looks as though we can identify a corresponding knowledge norm for conjecturing along the following lines: Knowledge Norm for Conjecturing (KNC): One’s conjecturing that p is epistemically permissible iff one knows that there is reason, but not sufficient reason, to believe that p.¹² It is easy to see, again, how, if something in the vicinity of KNC holds, SD threatens to resurface. We first note the Shiftiness Intuition for conjecturing: it looks as though what is appropriate to conjecture hinges in important ways on what is at stake practically at the context. More epistemic support is needed for conjecturing that the patient has cancer than for conjecturing that it will rain tomorrow. If this is so, we are faced by a dilemma between denting KNC and endorsing a shifty view of knowledge. Before we move on to the third way in which SD threatens to generalize, one crucial thing to note is that, if SDG1 and SDG2 are correct, their conjunction would make our prospects of escaping the Shiftiness Dilemma at any point in the epistemic normative landscape for constative speech acts look dim. After all, the conjunction of SDG1 and SDG2 just is the claim that any context-invariant epistemic norm with epistemic content for any constative speech act is a proper source of the Shiftiness Dilemma. Third, and most dangerously I believe, SD threatens to generalize beyond knowledge or talk of knowledge to all epistemic normativity/ epistemic normative talk: SD-Generalized 3 (henceforth SDG3): If SD holds for knowledge/ knowledge attributions, then it holds for all epistemic normativity/epistemic normative attributions.

¹² This formulation is imprecise, but it will do for now. Chapter 7 offers a full account of the epistemic normativity of conjecture.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

16      To see the worry, ignore for a second the non-KNA horn of the Shiftiness Dilemma, and focus on the horn that denies CI. If CI is false, we get a view on which knowledge or ‘knowledge’ is sensitive to practical stakes. For the sake of brevity, I will only discuss this point with regard to knowledge and epistemic normativity, but everything below applies mutatis mutandis to ‘knowledge’ and epistemic normative attributions as well. There are two ways in which the practical sensitivity of knowledge can occur that are of epistemological interest: either (1) practical normativity encroaches directly on epistemic normativity, i.e. on the ‘justification’ component of knowledge, such that what counts as knowledge-level epistemic justification varies with practical stakes (e.g. Fantl and McGrath 2002), or (2) it does not, in that the level of justification required by epistemic normativity remains fixed, but practical stakes come and raise the support requirements for knowledge (Stanley 2005). If (1) is the case, all epistemic normativity of belief is pragmatically encroached; given belief ’s centrality to epistemological affairs, as well as its intimate relationships with assertion, reasoning, and other plausible targets of epistemic normativity, the threat of uniform practical encroachment on epistemic normativity looms large. On (2), the danger is less straightforward, but not to be ignored. If knowledge is pragmatically encroached upon, given the popularity of (epistemic) knowledge norms for belief, assertion, reasoning, action, etc., and the intimate relation between the relevant acts and mental states, again, practical encroachment threatens to affect the vast majority of epistemically normative affairs. Last but not least, I would like the reader to consider the effects of the conjunction of SDG1, SDG2, and SDG3: if all these generalizations of the Shiftiness Dilemma obtain, what we get is one horn on which all epistemic norms with epistemic content for constatives are affected by practical context-variance, and another on which most, if not all, epistemic normativity is affected by pragmatic encroachment. What is worst, as I am about to argue in Chapter 2, embracing the first horn of the Shiftiness Dilemma is value-theoretically problematic and threatens to collapse into the second horn. If all this is so, arguably, not much room is left for the independence of thought from practical considerations.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

    

17

There are, of course, ways around these generalizations; I most definitely did not attempt to give a conclusive argument in their favour here. Rather, my purpose is to signal the danger, such that the reader is in a better position to appreciate the stakes involved in the discussion.

Conclusion This chapter has argued that the widely assumed Shiftiness Dilemma for Classical Invariantism and the Knowledge Norm of Assertion threatens to generalize in three dangerous ways: to all constatives, to all contextually non-variant epistemic norms with epistemic content, and to all epistemic normativity. In the light of this, the prospects are looking dim for the champion of the independence of thought from practical considerations. In the following chapters I will look at escape routes for the defender of Classical Invariantism.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

2 Epistemic WAMs In the previous chapter, we have seen that most actors in the debate assume that the intuitive variability of proper assertion with practical stakes motivates the following dilemma: either 1) we embrace a knowledge norm of assertion, and are forced into a view that takes knowledge, or ‘knowledge’, to be sensitive to practical stakes, or 2) we stick to our Classical Invariantist guns, but then the knowledge norm goes out the window and we get practical sensitivity in the normativity of assertion. I have argued that this Shiftiness Dilemma threatens to generalize in three ways: to all context-invariant norms of proper assertoric speech, to all constative speech acts, and to all epistemic normative constraints. If this is right, we should be more worried about this dilemma than we’ve been so far. We have also seen that this dilemma is mainly motivated by the intuition that, in high-stakes practical contexts, assertability does not come cheap: intuitively, more warrant is required in ASPIRIN-HIGH but not in ASPIRIN-LOW, for being in a position to properly assert that you have aspirin at home. I have dubbed this the Shiftiness Intuition. The next three chapters look at extant views that attempt to preserve the independence of thought in the face of assertion’s practical shiftiness and argue they remain ultimately unsuccessful. In this chapter, I look at views that aim to safeguard the independence of thought via embracing the second horn of the Shiftiness Dilemma (Assertion Sensitivism, or SA). For people who like Classical Invariantism about knowledge attribution, the jump from the Shiftiness Intuition to contextualism or pragmatic encroachment (Knowledge Sensitivism, SK) seemed rushed. As such, these authors venture to account for the Shiftiness Intuition under a Classical Invariantist umbrella by arguing for context sensitivity of proper assertability. Here it is:

Shifty Speech and Independent Thought: Epistemic Normativity in Context. Mona Simion, Oxford University Press (2021). © Mona Simion. DOI: 10.1093/oso/9780192895288.003.0002

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

   19 Assertion Sensitivism (SA): The degree of warrant necessary for epistemically¹ proper assertion varies with contextual features, while the degree of warrant necessary for knowledge stays fixed. Now, here is a fairly uncontroversial value-theoretic claim concerning the relation between the axiological and the deontic:² The Norms/Goods Type Association Claim (the Association Claim, or AC): Norms of type X are associated with values of type X. In what follows, I will argue that the two claims above are incompatible. If this is so, given the widespread support and exceptional high plausibility of AC, prior plausibility requires us to abandon SA.

2.1 Assertion Sensitivism and Value/Norm Type Association The thought behind Assertion Sensitivism is, roughly, to explain the intuitive variability in propriety from one ASPIRIN case to the other by keeping the standards for knowledge fixed, and allowing that the degree of warrant for epistemically proper assertion varies with context. In some contexts, less than knowledge is required for epistemically proper assertion, while more warrant may be needed in others.³ This view has

¹ Crucially, the results of this chapter only concern SA as a claim about the distinctively epistemic normativity of assertion. ² See Schroeder (2016) for an excellent overview. ³ SA also comes in more than one variety; first, there are people thinking that assertion is governed by one norm that stipulates that the appropriate amount of warrant for proper assertion varies with contextual features (e.g. Brown 2010, Gerken 2012, Goldberg 2015, McKinnon 2013, Rescorla 2009). Another way to be a sensitivist about assertion is to posit several norms governing assertion, and to suggest that it is a contextual matter which of them applies in a given case (e.g. Greenough 2011, Levin 2008, Stone 2007). The subtle differences between the above views are, however, to a large extent irrelevant for now (but see the next section for refinements). That is because this chapter mostly dwells at a higher level of generality: what I am concerned with is the claim that epistemically proper assertability varies with practical stakes, no matter what triggers the variation in propriety in question. Insofar as these authors stand by this claim, they are the proper target of this chapter.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

20      become known in the literature as an epistemic Warranted Assertability Manoeuvre (WAM)⁴ against Knowledge Sensitivism. In this respect, thus, according to SA, although the speaker’s epistemic status remains unchanged in the two ASPIRIN cases, the assertion ‘I have aspirin at home’ would not be epistemically proper in ASPIRIN-HIGH due to change in the relevant contextual features,⁵ most likely related to the relevant stakes. That is, while the speaker does know that he has aspirin at home in both ASPIRIN-LOW and ASPIRIN-HIGH, due to changes in context it is only in the former but not in the latter that his relevant assertion would be epistemically proper. Now, here is one plausible, widely endorsed⁶ thought: if there’s such a thing as an epistemic norm for assertion out there in the first place, it is there to make it likely that assertion delivers the epistemic goods⁷ we are using it for. And here is a fairly innocent value-theoretic claim to capture this thought: it looks as though a norm’s pertaining to one type or another has to do with the type of goods it is associated with. Thus, prudential norms will be associated with prudential goods, moral norms will be associated with moral goods, etc. Epistemic norms will thus come together with epistemic goods. Peter Graham puts the point succinctly: ‘Epistemic norms in this sense govern what we ought to say, do or think from an epistemic point of view, from the point of view of promoting true belief and avoiding error’ (Graham 2012). What we get, then, is the following easy way to individuate normative constraints: The Association Claim (AC): Norms of type X are associated with values of type X.

⁴ There are two ways one can pull a WAM: one can place the source of context sensitivity at the level of the epistemic norm itself (SA), or, to the contrary, defend a fixed norm and argue that pragmatic, Gricean considerations influence propriety in context (e.g. Rysiew 2001). This chapter is only concerned with the first incarnation thereof. The next chapter will look at pragmatic WAMs. ⁵ Defenders of SA go on separate ways when it comes to listing the relevant contextual determiners; that is, for some of them practical concerns figure higher on the list (e.g. Gerken 2012), while others (e.g. Goldberg 2018) focus more on non-practical context sensitivity. ⁶ See e.g. Brown and Cappelen (2011), Goldberg (2018), Kelp (2018), Simion (2018a, 2019c, 2020a). ⁷ I use ‘goods’ and ‘values’ interchangeably throughout. Nothing hinges on this.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

   21 Notice that AC is fully innocent from a value-theoretic perspective. That is because the mere association claim between norms and goods of the same type does not imply any substantial value-theoretic commitment; it holds on both the most notable views regarding the relationship of the good to the deontic.⁸ The teleologist explains the ‘ought’ in terms of the ‘good’; according to this philosopher, the following is true: AC-Teleology: Norms of type X are there to guide us in reaching values of type X. The deontologist reverses the order of explanation: according to ‘Fitting Attitude’ accounts of value: AC-Deontology: Values of type X are only valuable because norms of type X give us reasons to favour them. Anyhow, one way or another, the mere association claim holds. Let us now take a closer look at the SA proposal concerning the normativity of assertion and at how it fares in conjunction with the Association Claim. First, what we are talking about is the epistemic norm of assertion. The question, then, becomes: what is the relevant epistemic good? Many authors (e.g. David 2005) regard truth as the fundamental epistemic good. The most prominent counter candidate in the literature is knowledge (Williamson 2000). For our purposes here, in order to stay on the safe side, we will test the plausibility SA for both candidate goods.⁹ Note, also, that the epistemic interest at stake can be thought to be both at the speaker’s and at the hearer’s end. As such, we will have to look on both sides. ⁸ For support of AC: for a good general overview of the relevant literature in value theory, see, for instance, Schroeder (2012); for champions of the teleological direction of explanation, see e.g. Moore (1993), Portmore (2005). Sidwick (1907), and Slote (1989). For the deontological direction, see e.g. Scanlon (1998), Ewing (1947), and Rabinowicz and Rönnow-Rasmussen (2004). ⁹ Note, also, that the argument can be run in a parallel fashion for a justification goal (and the results are likely to coincide with the results for the knowledge goal, insofar as what is meant is knowledge-level justification). Also, see section 2.2 for a discussion of what is the case on the assumption of an epistemic goal that itself varies with practical stakes, such as ‘providing actionable information’.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

22      Let us start with teleological order of explanation. By AC, then, SA proponents will also be committed to: SA-Teleology: goods.

The SA norm is there to guide one in reaching epistemic

Spelling out the norm, and on a truth-goal assumption, then, we get: SA-Teleologytruth: One should proportion the degree of warrant supporting one’s assertion to contextual features to the aim of making a true assertion/generating true belief in one’s hearer. This formulation is clearly false: pragmatic factors do not influence truth-conduciveness. Let us turn to knowledge¹⁰ as the main epistemic good, then plug it into the SA-Consequentialism and see what happens to the framework: SA-Teleologyknowledge: One should proportion the degree of warrant supporting one’s assertion to contextual features to the aim of making a knowledgeable assertion/generating knowledge in one’s hearer. Unfortunately for the SA proponent, this formulation, although not strikingly false, amounts to what she was trying to avoid in the first place: Knowledge Sensitivism (SK). Here is how: in the speaker’s case, the route to SK is pretty straightforward: in order to come to know, the speaker is in need of more epistemic support in high-stakes contexts than in lowstakes ones. While on the hearer’s side, what the claim amounts to is that the hearer needs an epistemically better source in high-stakes scenarios than in low-stakes ones in order to gain knowledge. This is an unacceptable result, since, once again, it collapses the view into SK; what the SA claim would amount to, under this formulation, would be a view according to which one needs a degree of warrant that is suitable to one’s practical context in order to be knowledgeable. ¹⁰ Note that prominent defenders of SA (e.g. Goldberg 2018) explicitly support generating testimonial knowledge as the main epistemic role of assertion.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

   23 If that is the case, SA seems to not be very nicely compatible with a teleological value-theoretic framework. On one hand, this is rather unfortunate; after all, ideally, one does not want one’s preferred account of the normativity of assertion to commit one to very substantive valuetheoretic claims. On the other hand, given that several people in this literature dislike consequentialism on independent grounds, maybe this will not worry the SA champion too much. Let us change the framework, then, and go for the deontological incarnation of the Association Claim. Consider, first: SA-Deontologytruth: Truth is an epistemic good because the SA norm gives us reason to favour it. It is a bit mysterious, however, in virtue of what exactly does SA give us reason to favour truth rather than falsehood. After all, it looks as though, independently of whether I am right or wrong about whether p is the case, according to SA, the important thing is that I don’t assert it unless I have a contextually appropriate amount of warrant. As such, SA seems to be completely indifferent when it comes to whether I am in possession of the truth or not, and therefore fail to favour it in any way. SA-Deontologyknowledge: Knowledge is an epistemic good because the SA norm gives us reason to favour it. Again, this formulation is either false or it collapses SA into SK. Recall that SA asks for less warrant in low-stakes scenarios and more warrant in high-stakes; as such, it gives us no particular reason to favour Classical Invariantist knowledge over other epistemic standings characterized by less, respectively more warrant. If, however, knowledge itself is sensitive to practical context, as SK would have it, the SA norm is able to provide us with reason to favour it. To sum up, then: if the (value-theoretically innocent) Association Claim between norms and goods of a particular type holds, SA comes out untenable for the main candidates for the central epistemic good in the literature. On both available AC directions of explanation, in a truthgoal framework, its claims turn out either false or highly implausible. In a

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

24      knowledge-goal framework, the position collapses into Knowledge Sensitivism, which was what its proponents were reacting against in the first place.

2.2 Sensitive Epistemic Goals What SA seems to need is a complementary pluralistic account regarding the epistemic goal, tightly connected to contextual practical determiners. That is, roughly, a view on which the epistemic goal varies with practical stakes, such as: the goal of assertion is providing actionable information.¹¹ On such a view, variation in warrant for proper assertion would just track the variation in epistemic goal, which, in turn, would track the variation in epistemic needs given the practical context. Now, to my knowledge, this view is still in need of defence in its own right; thus, as things stand, it cannot easily be employed in support of SA, given that its theoretical up- and downsides are still underexplored. This aside, a few serious worries arise even from just this rough sketch for the view. First, note that holding this practical-context variant view about the epistemic goal in general might get SA into trouble when it comes to the normativity of belief; that is, if some variety of (the very popular) norm commonality assumption is true about assertion and belief—i.e. the claim that they are governed by one and the same epistemic norm—SA will be in danger of collapsing into SK, if they will hold that the epistemic goal—and thereby the epistemic norm—of belief also varies with practical factors. To see why this is the case, note that many (if not most) people¹² in the debate stand behind something like the following deontic thesis for belief (DTB): DTB:

A belief is epistemically permissible iff epistemically justified,

¹¹ Jessica Brown and Sandy Goldberg (in personal communication) suggested the epistemic WAM-er might want to take this route. ¹² See e.g. Williamson (2000), Simion, Kelp, and Ghijsen (2016).

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

   25 where the justification at stake is taken to be knowledge-level justification. Since goals and norms bear the association relation, it would seem to follow that if the epistemic goal of belief varies with stakes, so does its norm. If epistemically permissible belief varies with stakes, however, on DTB, so does knowledge-level justification, and therefore knowledge itself. We are back to SK. As far as I can see, there are two ways to go at this point in order to avoid this result:¹³ either (1) restrict the view to assertion and deny the commonality assumption, i.e. have a view on which permissible assertion varies with context but permissible belief does not, or (2) hold that epistemically permissible assertion goes hand in hand with practically permissible belief, not with epistemically permissible belief. The problem with both (1) and (2) is that they will allow for intuitively strange situations whereby a speaker’s assertion that p will count as epistemically permissible (in virtue of its degree of warrant being good enough for the hearer’s practically permissible belief ), although they do not believe that p themselves, in virtue of not having enough warrant to epistemically permissibly believe that p. While many people (e.g. Brown 2010) have put forth cases in which it seems intuitively permissible to believe that p, but not to assert it (due to contextual high stakes), the reverse is much more problematic. Indeed, many people think that assertion without belief is always criticizable (‘Why would you say this? You don’t believe it yourself!’). Some disagree: Jennifer Lackey (2008) notably puts forth cases in which the speaker (intuitively permissibly) asserts warrantedly but not knowledgeably due to lack of belief. That being said, even if we think Lackey is right,¹⁴ it’s one thing to think that permissible assertion does not require belief, and quite another to believe permissible assertion does not require enough warrant to believe. However, that is exactly the view that the SA champion will end up with if she chooses the route of denying commonality in the envisaged way: there will be cases in which assertion is permissible but belief is not.

¹³ Many thanks to an anonymous referee for pressing me on this point. ¹⁴ I agree with Lackey’s intuition about these cases, and Kelp and Simion (Forthcoming) discusses them and mounts a functionalist defence of the knowledge norm.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

26      Alternatively, one can (3) deny that the status at stake in DTB is knowledge-level justification, rather than some practically sensitive variety of epistemic justification for belief. This is an epistemically normative pluralistic picture for belief: a belief might be epistemically justified even if it is not a justification that is strong enough as the one that is required for knowledge. The standards of epistemic justification/ permissibility for belief might therefore be context-sensitive, even though the standards for knowledge-level justification/permissibility are not context-sensitive. The former track actionability, the latter do not. A few things about this: while I grant that this picture occupies a position in the logical space, it is not defended anywhere in the literature, and it is hardly a straightforward, theoretically neutral way to go. Therefore, I take it, it requires very serious defence in order to be taken seriously. Here are a few reasons for this: first, because it needs to deny the widely accepted DTB. Second, because it needs to stipulate normative pluralism where all the competing views do not—so it scores worse on simplicity. Third, because on this view one can have a knowledgeable belief that one should, epistemically, not hold which is rather counterintuitive. Fourth, most crucially, the defender of such an account will want to avoid the following results: on her view, given that the epistemic permissibility of belief varies with practical stakes, believing a falsehood, or something one has no justification whatsoever for, when nothing hinges on it, or in return for one million dollars, would be epistemically perfectly fine. That seems like a serious theoretical cost. In a similar vein, when nothing of importance is at stake for the hearer— say, for instance, we are just making conversation about the weather— it should turn out to be epistemically fine to assert with no warrant whatsoever. All this, of course, does not seem right. One can maybe try to address this problem by setting a minimal threshold for (epistemically) permissible assertion/belief.¹⁵ One way to do this in a non-ad-hoc manner would be by arguing for some pragmatic considerations in favour of the relevant threshold. For instance,

¹⁵ On Greenough’s (2011) view, for instance, knowledge is the minimal threshold for permissible assertion.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

   27 one could think that something like the Gricean maxim of Relevance would recommend against making assertions devoid of any practical importance.¹⁶ Or, alternatively, one could think that, by the maxim of Quality, asserting that p carries the implicature that there is some reason to believe that p. As such, on this view, in no-stakes cases, while asserting in the absence of any warrant is strictly speaking epistemically proper in virtue of it being practically proper, it comes across as intuitively inappropriate due to considerations pertaining to the pragmatics of language. Alas, though, this move will not get the champion of the variant goal view too far either. After all, one can easily imagine cases where the amount of warrant is problematically raised rather than lowered. Take, for instance, a case where I am offered one million dollars to withhold belief unless I am certain (as in Cartesian certainty) that p. In this case, the defender of the variant goal view will have to say that, if I see that there is a table in front of me, and therefore I believe that there’s a table in front of me, my belief is epistemically impermissible. Again, this does not sound right. What the defender of this account seems to be in need of, then, would be a principled way to separate the ‘good’ prudential considerations from the ‘bad’ ones; I submit that there is reason to believe there is no easy, non-question-begging answer for this problem in sight.

2.3 Sensitive Discursive Justification One possible reply from the SA camp goes along the following lines: the stipulated variability in warrant is required for belief generation. In highstakes scenarios, the thought would go, the hearer might be extremely cautious and ask the speaker to back her assertion. In this case, being in possession of an amount of warrant appropriate to the situation would put the speaker in a position to be able to meet this demand, and thus successfully generate the relevant belief in her hearer.

¹⁶ Mikkel Gerken (p.c.) suggested this might be the way he might want to go about this issue.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

28      The first problem with this move is that, on the present formulation, SA only asks speakers to be in the possession of the relevant degree of warrant, not to also have access to it so as to be able to back their assertion if needed. Second, I worry that adding the necessary access requirement would render the view implausibly strong; after all, surely small children can produce epistemically proper assertions, in spite of the fact that they don’t have very well developed reflective capacities. Furthermore, most of our knowledge is stocked in memory and, for most of it, we do not really remember how we came to acquire it to begin with. I, for instance, surely do not remember how I got to know that Berlin is the capital of Germany. Does that mean I cannot make the relevant assertion? The answer, according to this enhanced version of SA, will have to be ‘no’. Two options are still available to the SA defender at this point: first, she could make the need for discursive justification context-dependent also, such as to only encounter the cognitively unsophisticated asserters’ problem when the stakes are high. This more restricted version seems to enjoy more plausibility. Gerken’s view, for instance, explicitly requests that it is only in some contexts that one should be able to back one’s assertion with appropriate support. Alternatively, she could argue that the need for more warrant in highstakes than in low-stakes scenarios pertains to hearers not believing what the speaker says unless they not only know the content of their assertion, but they also know that they know—which, in turn, requires more warrant than mere knowledge.¹⁷ This picture, in turn, would have nothing to do with the standards for knowledge; quite to the contrary, it explicitly allows that those standards remain fixed. There are, however, good reasons to believe that the prospects for this sort of move are rather dim. To see this, let us take another look at the envisaged SA champion’s reply: the variability in warrant is required for belief generation, not for its truth. In high-stakes scenarios, the hearer might be extremely cautious and ask the speaker to either back her assertion with the contextually appropriate discursive justification or, alternatively, to know that they know. Were the speaker not able to do ¹⁷ See Williamson (2005) for an account along these lines and Chapter 4 for discussion.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

   29 so/not to have knowledge of knowledge, the hearer would not believe the content of the assertion, and, as such, the aim of generating true belief/ knowledge would be missed. Now, note that, for all that is said above, we are dealing with a descriptive, empirical claim: the thought is that, as a matter of fact, there is a chance that the hearer requests discursive justification/knowledge of knowledge for believing. But, of course, this cannot be what is meant to be relevant to the normative claim of SA; after all, maybe hearers are not in their epistemic right to do so, in which case no obligation for the speaker should follow. Just because hearers might, for instance, require speakers to wear red hats if they want to be believed, it does not follow that we will have a red-hat-wearing norm of proper assertion thereby, at least surely not an epistemic such norm. Similarly, just because, in high-stakes cases, hearers usually require speakers to know that they know or, alternatively, to offer contextually appropriate discursive justification, it does not follow that we will have any such requirement on the speaker’s side. The SA defender does not want to say that any absurd claim hearers might have is going to affect the content of the epistemic norm governing the speaker’s speech acts. What is needed is a normative claim alongside the empirical one; for any obligation to follow on the speaker’s side, it must be the case that, on top of them being in the habit to do so, hearers are also epistemically permitted to ask for discursive justification/knowledge of knowledge. What is needed, then, is a norm that makes the requirement for knowledge of knowledge/discursive justification permissible. Furthermore, given that we are interested in the epistemic norm of assertion, the relevant norm on the part of the hearer also needs to be epistemic, rather than prudential or moral. In a nutshell, then, what we need is an epistemic norm that makes it permissible for hearers to only believe what the speaker says if the latter has knowledge of knowledge/contextually appropriate discursive justification. This, however, will easily threaten, once again, to drive the SA defender back to the trouble she was trying to avoid to begin with—i.e. collapsing into Knowledge Sensitivism. Here is how: again, it is widely accepted that a belief is epistemically justified (where what is at stake is the justification required for knowledge) if and only if it is epistemically

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

30      permissible (DTB).¹⁸ Therefore, a stakes-variant epistemic norm of belief will readily result in stakes variation for knowledge. Of course, one can have a stakes-variant prudential norm of belief, for instance. However, again, this prudential norm on the part of the hearer would only be able to generate a prudential norm on the part of the speaker; what we are searching for, though, is the distinctively epistemic norm of assertion. The two will, of course, often come apart: it might be epistemically perfectly fine, for instance, to tell your boss that he’s bald if you know it to be the case, but, prudentially speaking, it is definitely better to keep quiet (Brown 2010). One last option for the SA defender that still needs to be discussed is her possible retreat from direct to indirect practical stakes sensitivity. According to this account, the reason why we need more warrant in high-stakes than in low-stakes is because more error possibilities become salient. As such, proper assertability is only indirectly sensitive to practical stakes, through its being sensitive to the (genuinely) epistemic need for dismissing salient error possibilities. Patrick Greenough’s (2011) view affords this way out. According to Greenough, assertion is governed by different norms in high-stakes and low-stakes scenarios. That is, in highstakes, but not in low-stakes, the speaker must also be able to cite explicit evidence against all those not-p possibilities which are salient in the high standards in play. The assumption that needs be discussed here, however, is the claim that this need generates a genuinely epistemic norm. To what epistemic aim does one need to be able to dismiss the relevant error possibilities? One plausible answer is that the latter constitute themselves in normative defeaters (Lackey 2008) and, as such, the hearer (epistemically) should not believe the speaker’s assertion unless suited defeater defeaters are offered. This reply, indeed, seems to be innocent of any pragmatic normative consequences for belief. Note, however, that this reply will not do its job in supporting SA’s claim against KNA, i.e. the claim than more than knowledge is needed in high-stakes contexts for proper assertability. After all, plausibly enough, the same normative defeaters that forbid the hearer from believing will also (normatively) act on the ¹⁸ Williamson (2000), Simion, Kelp, and Ghijsen (2016).

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

   31 speaker’s epistemic standing. As such, the defender of KNA can easily help herself to the same explanation of the Shiftiness Intuition here: the reason why the speaker needs to be able to dismiss relevant error possibilities constituting themselves in normative defeaters is because, otherwise, her knowledge is defeated, and therefore, by KNA, it is not permitted to assert.

Conclusion This chapter has identified a strong incompatibility between a very popular response to the Shiftiness Dilemma—what I have dubbed Assertion Sensitivism—and a fairly uncontroversial value-theoretic thesis concerning the association between norms and values of the same type. I have argued that Assertion Sensitivism, as a thesis about the epistemic normativity of assertion, is untenable in conjunction with the uncontroversial Association Claim between norms and values of the same type. To show this, I have picked several candidates for the main epistemic goods in the literature, and showed how SA’s claims either turn out false or collapse the view into Knowledge Sensitivism, i.e. the position champions of SA were trying to avoid to begin with. If this is so, given the widespread support and exceptional high plausibility of AC, prior plausibility requires us to abandon Assertion Sensitivism.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

3 Pragmatic WAMs We have seen that, in light of embracing the Shiftiness Dilemma, the vast majority of philosophers accept Assertion Incompatibilism: given intuitive variability of proper assertion with practical stakes, Classical Invariantism is incompatible with a biconditional knowledge norm of assertion. We have also seen, in the previous chapter, that views attempting to preserve Classical Invariantism by embracing the Assertion Sensitivist horn of the dilemma are value-theoretically problematic: the view fails on prior plausibility due to being incompatible with the following highly uncontroversial value-theoretic claim: norms of type X are associated with values of type X. If all this is so, it would seem we are stuck with a shifty view of knowledge (or ‘knowledge’). This is not good news for the Classical Invariantist, nor for the purist about epistemic normativity more generally. There are also a few dissenting voices in the literature, however, who deny the Shiftiness Dilemma altogether:¹ some invariantists venture to explain the sensitivity data for proper assertion in a fashion that preserves both CI and KNA (Assertion Compatibilism). There are two varieties of Compatibilism available on the market: the first one employs a pragmatic Warranted Assertability Manoeuvre (henceforth, Pragmatic Compatibilism); the second variety appeals to higher orders of knowledge to explain away the Shiftiness Intuition (henceforth KK Compatibilism). I will look at these views in turn. This chapter discusses Pragmatic Compatibilism and argues that the view has difficulties with delivering epistemic normative independence for cases of shiftiness of assertions that do not involve knowledge attributions.

¹ See Williamson (2005), Brown (2006), Reed (2010, 2012), Hazlett (2009), Rysiew (2001), Simion (2016, 2018, 2020a), Kelp and Simion (Forthcoming). Shifty Speech and Independent Thought: Epistemic Normativity in Context. Mona Simion, Oxford University Press (2021). © Mona Simion. DOI: 10.1093/oso/9780192895288.003.0003

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

   33

3.1 Pragmatic Compatibilism According to Pragmatic Compatibilists, the source of variability in contextualist cases pertains to what is pragmatically conveyed by the assertion in question rather than by what is, strictly speaking, said (e.g. Brown 2006, Hazlett 2009, Rysiew 2001). On this view, in BANK CASE HIGH, for instance, DeRose does, in fact, know that the bank will be open on Saturday. However, given the presumption of relevance and informativeness, saying that he knows pragmatically conveys that he is able to eliminate all contextually salient error possibilities—such as, for instance, the possibility that the bank changed its hours in the last two weeks. But of course, this is just an error possibility that DeRose cannot eliminate. In this way, then, DeRose’s self-attribution of knowledge would carry a false implicature. And it is just this false implicature that, according to Pragmatic Compatibilists, explains why DeRose’s self-attribution of knowledge would seem wrong. Strictly speaking, then, on this view, Classical Invariantism is perfectly compatible with KNA: DeRose knows in both BANK CASE LOW and BANK CASE HIGH and is, therefore, in a good enough epistemic position to assert that the bank will be open. Now, a lot of ink has been spilled on whether the Pragmatic Compatibilist account of the bank cases is ultimately satisfactory for accounting for the shiftiness of proper knowledge attributions.² I will not press in that direction here. In contrast, what I will focus on is whether Pragmatic Compatibilism can explain the Shiftiness Intuition in its full generality. After all, we have seen that this intuition is not restricted to knowledge attributions, but rather it also arises in cases of assertions of propositions that do not feature epistemic concepts. By way of illustration, recall once more the variation of CASE B in which were DeRose to assert simply that the bank is open tomorrow, his assertion would be improper. Note that Pragmatic Compatibilists will have a hard time extending their account of the original CASE B to this variation. The thought that the knowledge attribution, in conjunction with

² See e.g. DeRose (2009).

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

34      contextually salient error possibilities, generates an implicature to the effect that DeRose can eliminate the salient error possibilities has little traction here. After all, there is no knowledge attribution being made. To the best of my knowledge, Rysiew and Brown have not addressed this issue in print. Here is Hazlett’s account of these cases (2009, 31): say that, in the relevant variation of CASE B, DeRose were to assert that the bank is open tomorrow. Now, by the Gricean maxim of Quality (alternatively, if knowledge is the norm of assertion), this generates the implicature that DeRose knows that the bank is open tomorrow. According to Hazlett, this, in turn, will imply that the speaker can eliminate all the contextually salient error possibilities. For instance, in BANK CASE HIGH, were DeRose to assert that the bank will be open tomorrow, his assertion would generate the implicature that he knows this, which, in turn, would generate the implicature that he can eliminate the contextually salient error possibility that the bank changed its hours in the last two weeks. Since DeRose can do no such thing, the assertion does generate a false implicature, i.e. in virtue of generating an implicature that itself has a false implicature.

3.2 Shiftiness without Error Possibilities One immediate worry with this line is that it is easy to come up with cases that are just like the variation of BANK CASE HIGH we have been considering in which there are no contextually salient error possibilities being tabled. The ASPIRIN cases we have been looking at are precisely like that. Recall the cases: ASPIRIN-LOW. You remember having bought aspirin last month. You are heading together with your sister towards your place for dinner, and she lets you know she has a minor headache. Your sister asks you: ‘Do you have aspirin at home, or should we go to the pharmacy?’ You flat out assert: ‘Don’t worry, I have aspirin at home, I remember having bought some.’ ASPIRIN-HIGH. You remember having bought aspirin last month. Your sister’s two-year-old baby is having a fever, and needs an aspirin as

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

   35 soon as possible. Your sister asks you: ‘Do you have aspirin at home, or should we go to the pharmacy?’ You give the matter a bit of thought, and answer: ‘Well, let’s drop by the pharmacy, just in case.’ In these cases, in contrast to the Bank Cases, no error possibilities are raised; still, the Shifty Intuition survives. Perhaps the Pragmatic Compatibilist could respond that the set of contextually salient error possibilities need not be made explicit. It will do if there is an implicit set of such possibilities. For instance, in ASPIRIN-LOW, the Pragmatic Compatibilist could perhaps claim that the possibility that you misremember having bought aspirin becomes implicitly salient. The trouble with ‘going implicit’ is that the solution threatens to be too easy. For this move to not deliver ad hoc explanations of the data, we will need an independently plausible account of how the set of implicitly salient error possibilities is generated. Given that what we are dealing with is a context-sensitive phenomenon, it is quite plausible that it is facts about the participants to the conversation that generate the set, such as their interests, what they are attending to, etc. It is hard to deny that these facts vary from one speaker to the next, which begs the question as to just how a single set is settled on. One obvious proposal is that it is the lowest common denominator, as it were, i.e. the intersection of the set of error possibilities that each speaker would generate individually. However, this proposal is tricky because it is just not clear why all the error possibilities the Pragmatic Compatibilist needs will always be generated. For instance, why should we think that, in ASPIRIN-HIGH, the possibility of misremembering should always be generated as implicitly salient? After all, maybe you—but not your sister—know about yourself that you have an excellent memory. Note that if it is possible that the possibility of misremembering is not generated as implicitly salient for all the participants in the conversation, it looks as though the Pragmatic Compatibilist account of the data is in trouble. An alternative proposal that would take care of this problem claims that what matters is the highest common denominator, i.e. the union of the set of error possibilities that each speaker would generate individually. This proposal is also tricky because it is not clear why there couldn’t be a participant to a low-stakes conversation whose interests unduly raise the bar for

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

36      ‘knowledge’. But given all this, the Pragmatic Compatibilist who wants to solve the problem posed by cases like ASPIRIN-HIGH shoulders a substantive explanatory burden. Unless this burden has been discharged, we have every reason to remain sceptical about the Pragmatic Compatibilist line here.

3.3 Shiftiness without Stakes Cases like ASPIRIN-HIGH suggest that tabled error possibilities are not necessary for triggering the unassertability intuition. Even if it turns out that this problem can be circumvented, perhaps along the lines just outlined, the Pragmatic Compatibilist faces a further problem: tabled error possibilities seem not to be sufficient for doing the job either. That is, it looks as though, at least in some cases, tabling error possibilities will not do the trick, at least not in the absence of accompanying high stakes.³ If that is the case, however, it starts looking as though proper assertability does not have all that much to do with tabled error possibilities to begin with. Rather, it seems more and more plausible that the only thing that matters are variations in stakes. Thus consider the following variation of ASPIRIN-LOW: ASPIRIN-LOW*. You remember having bought aspirin last month. You are heading together with your sister towards your place for dinner, and she lets you know she has a minor headache. You flat out assert: ‘Don’t worry, I have aspirin at home, I remember having bought some.’ Your sister remarks: ‘Maybe you’re mistaken; you do sometimes misremember things.’ ‘I have aspirin at home,’ I reply. ‘All right,’ she says. I take the intuition here to be that you and your sister are right to put the matter to rest. If that is true, however, tabled error possibilities by themselves—that is, in the absence of raised stakes—won’t do the work

³ Among Brown’s concerns in her (2006) is to argue that Rysiew doesn’t sufficiently recognize this in his (2001).

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

   37 Pragmatic Compatibilists take them to do. This, of course, should hardly come as a surprise: just because sceptical worries are ‘in the air’ while I write an epistemology paper, for instance, it hardly follows I do not know and cannot assert that there is a computer in front of me. More is needed to trigger unassertability. And we still have not reached the end of the obstacles the Pragmatic Compatibilist will have to overcome. Recall that Hazlett wheels in some pretty heavy theoretical machinery, i.e. the idea of an implicaturegenerating implicature: for the account to work, the implicature that the speaker knows what she says in itself must be able to generate the further implicature that she can eliminate all contextually salient error possibilities. Again, there is an explanatory burden to be discharged. To begin with, one might wonder whether there is such a thing as an implicature generating implicatures in the first place. Moreover, even if it turns out that the phenomenon exists, it would be good to know whether what we are dealing with here is a fully general phenomenon, or whether it is restricted to particular types of implicatures. And then there is the issue of just how implicatures generate implicatures. And finally the account had better work at least for the type of implicature that the Pragmatic Compatibilist wants to invoke. In this way, the burden the Pragmatic Compatibilist shoulders gets weightier and weightier. Here is another, more direct problem for the Pragmatic Compatibilist. It is agreed by many that implicatures must be calculable. And, of course, whether a given implicature is calculable will depend on the cognitive capacities of the agents who are facing the calculation task. But given that this is so, it is surely possible for there to be communities of speakers that are only able to generate simple implicatures, and not implicatures of implicatures. It is also surely possible for some such communities (i) to operate the concept of knowledge and have a word expressing it in their language, and (ii) to have variations in stakes. If so, we can generate analogues of DeRose’s bank cases for members of these communities. Note that the Pragmatic Compatibilist account will work just fine for the analogues of DeRose’s original bank cases. After all, all the Pragmatic Compatibilist needs here is a simple implicature. However, the envisaged account for the generalized cases like ASPIRIN-HIGH will just not work.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

38      After all, according the Pragmatic Compatibilist line, what’s going on here is an implicature of an implicature. Since members of the community cannot calculate implicatures of implicatures and since implicatures must be calculable, it follows that there is no such thing as an implicature of an implicature for this community. By the same token, the Pragmatic Compatibilist account is bound to fail here.

3.4 Shiftiness and Relevance Last, I would want to look at an alternative possible reply on behalf of Pragmatic Compatibilism that does not rely on the problematic notion of implicature-generating implicature,⁴ and which, one might think, is closer to the spirit of the view to begin with. What seems key for people like Brown (2006) and Rysiew (2001) is that in the relevant cases (e.g. DeRose’s BANK CASE HIGH) what’s practically important is that the subject be in a very good/strong epistemic position vis-à-vis the proposition in question. Due to considerations of relevance (it’s alleged), a positive claim to knowledge in BANK CASE HIGH would then pragmatically impart that S’s epistemic position is good enough relative to those heightened standards, and so that S can rule out any salient alternatives. But this doesn’t clearly mean that some such alternatives must be mentioned in advance for the former proposition, which is plausibly false, to be communicated. Why would it be improper for DeRose to assert, in HIGH, simply that the bank is open tomorrow? Because he’d be representing himself (as per the maxim of Quality) as having ‘adequate evidence’, where such adequacy (as per the maxim of Relevance) is context-, and so stakes-sensitive. Given this, in asserting that p, DeRose would be pragmatically communicating that his epistemic position vis-à-vis p is much stronger than it plausibly is. The worry for this reply is that the notion of ‘contextually-sufficient’ adequate evidence is itself typed: it can be sufficient e.g. prudentially—i.e. for practical purposes; morally—i.e. for moral purposes; or epistemically— for epistemic purposes. What the champion of Pragmatic Compatibilism ⁴ I would like to thank an anonymous referee very much for this suggestion.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

   39 needs, and is missing, is an argument that it is the former and not the latter sense of enough that is at stake here—on pain of either collapsing into a shifty view of knowledge (if the former obtains), or begging the question against it (if no argument is given). Furthermore, as I am about to argue in Part II, once we recognize and have an argument in favour of prudential normativity being at work in these cases, we no longer need to rely on the pragmatics of language to do the work in explaining the Shiftiness Intuition.

Conclusion This chapter has argued that the Pragmatic Compatibilist faces a number of difficult obstacles. The Pragmatic Compatibilist line relies crucially on contextually salient error possibilities. However, we have seen that there is reason to think that such error possibilities are neither necessary nor sufficient in generating unassertability. In contrast, stakes alone seem to do the trick. If that is the case, however, a plausible Pragmatic Compatibilist explanation of the generalized data does not seem to be forthcoming.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

4 KK Compatibilism We have seen in the previous chapters that, in light of embracing the Shiftiness Dilemma, the vast majority of philosophers accept Assertion Incompatibilism: given intuitive variability of proper assertion with practical stakes, Classical Invariantism is incompatible with a biconditional knowledge norm of assertion. There are also a few dissenting voices, however: some invariantists venture to preserve both Classical Invariantism and the knowledge norm (Assertion Compatibilism). There are two varieties of Compatibilism available on the market. The previous chapter discussed the first of the two: Pragmatic Compatibilism, employing a pragmatic warranted assertability manoeuvre to explain away the shiftiness data. I have argued that the view has difficulties with delivering the goods when it comes to epistemic normative independence for cases of shiftiness of assertions that do not involve knowledge attributions or explicitly tabled error possibilities. The competing Compatibilist view is championed by Tim Williamson (2005) and John Turri (2010). The key idea that both subscribe to is that the unassertability in the high-stakes contextualist cases can be explained in terms of the subjects’ lack of higher-order knowledge (henceforth, KK Compatibilism).

4.1 Knowledge of Knowledge While there is agreement on the central idea between Williamson and Turri, their accounts differ in the details. According to Williamson, although, strictly speaking, all that is needed for proper action—assertion included—is first-order knowledge, when the stakes are high, we tend to

Shifty Speech and Independent Thought: Epistemic Normativity in Context. Mona Simion, Oxford University Press (2021). © Mona Simion. DOI: 10.1093/oso/9780192895288.003.0004

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

 

41

find people who act without knowing that they meet the condition for proper action blameworthy for so doing: How harshly should we judge practical reasoning in which the agent relies on an appropriate premise without being in a position to know that it is appropriate? A natural answer is: it depends on how much is at stake. If not much, then it seems unreasonably pedantic to condemn the reasoning. But if matters of life and death are at stake, the charge that the agent was not in a position to know that the premise was appropriate becomes more serious. (Williamson 2005, 232)

While Turri also wants to explain the unassertability in high-stakes cases in terms of lack of knowledge of knowledge, he claims that what is going on is that speakers in high-stakes cases do not make mere assertions but perform speech acts of a different type, to wit, guaranteeing. According to Turri, different speech acts come at different ‘epistemic costs’: guessing is cheap, conjecturing requires some epistemic support, assertion asks for knowledge and swearing and guaranteeing are the most ‘epistemically expensive’ speech acts in that they require secondorder knowledge. Furthermore, participants in a conversation are usually sensitive to contextual variations in the speech acts that are being performed. As such, what happens in contextualist high-stakes cases is that, as soon as the context changes, the speech act that is assumed to be performed is not mere assertion anymore, but the more ‘expensive’ speech act of guaranteeing. This is said to explain the intuitive variability of speech act propriety with stakes, while leaving both CI and KNA unaffected. For our purposes, the differences between the views defended by Williamson and Turri are of little consequence. The reason for this is that my main target will be KK Compatibilism’s central claim that the Shiftiness Intuition can be explained in terms of the subjects’ alleged lack of higher-order knowledge, which both Williamson and Turri accept and employ. In what follows, then, I will be focusing on Williamson’s account. However, the vast majority of the results below—in particular the discussion in section 4.3—apply, mutatis mutandis, to Turri’s view as well.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

42     

4.2 The Normative Status of the Blameworthiness Claim Recall that, according to Williamson, we tend to blame people who, when the stakes are high, act without knowing that they have what it takes to properly act, i.e. that they meet the relevant norm. Here is one question that arises at this point: what is the normative source of the blameworthiness at stake in Williamson’s account? One option that Williamson considers and rejects—with what I take to be good reason—is that it is the norm for action/assertion itself: The insensitive invariantist could try to build variation in the required number of iterations of knowledge into appropriateness itself, with a corresponding revision of [the knowledge norm of action/assertion]: in some cases q would be appropriate iff one knew q, in others iff one knew that one knew q, and so on, depending on the stakes. If such a move is acceptable, it provides the insensitive invariantist with a systematic response to arguments from practical differences to shifting semantic standards for epistemic terms. However, the move does not guarantee the epistemic accessibility of appropriateness, for the antiluminosity argument is quite general. (Williamson 2005, 233)

In sum, the thought is that the non-luminosity of knowledge generates cases where you obey the norm but you fail to know that this is the case— and are, therefore, by Williamson’s lights, blameworthy for making the corresponding assertions in high-stakes scenarios. However, if one modifies the norm of assertion as to stipulate that you not only have to know, but you also need to have the higher-order knowledge appropriate to your stakes, that condition can fail to be luminous too—like any other non-trivial condition. That is, you can know that p, and have the contextually appropriate higher-order knowledge also, but fail to know that you do. In this case, again, Williamson will have to say that you are blameworthy for asserting, given that you didn’t have the right number of iterations of knowledge. And so on. I agree with Williamson on this count. However, the question is whether leaving the requirement for contextually appropriate higher-

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

 

43

order knowledge out of the norm for action/assertion helps avoid the above problem. To see the worry, note that the following principle is extremely plausible: BLAME: Ceteris absentibus, norm N-blameworthiness implies norm N-violation.¹ BLAME² is an extremely plausible principle: after all, it is plausible to believe that one cannot be blameworthy if one has not done anything wrong; furthermore, trivially, one cannot do something wrong without breaking a norm, i.e. wrongness implies norm violation of sorts. Together, these two claims imply BLAME.³ To be clear, several authors have argued that moral blameworthiness need not imply one has done something objectively wrong (see e.g. Capes 2012). I agree with this claim. BLAME is not restricted to objective wrongdoing. You can intend to do something wrong—in breach of a certain norm N—and fail, in which case you are not N-blameworthy, but blameworthy for breaking a ¹ This view falls squarely into what Ishtiyaque Haji (1988) has dubbed the ‘Objective View’ of blameworthiness: necessarily, if S is blameworthy for A-ing, then it was objectively morally wrong for S to A. The Objective View is assumed in much recent work on moral responsibility. Notable exceptions include Capes (2012), Haji (1998) and Zimmerman (1997). The classical type of case thought to constitute a problem for the Objective View is one that features badly intended agents who end up doing morally good things by mistake. The thought is that these agents are blameworthy even though they do not do anything wrong. Note also that it is entirely compatible with Norm-Specific Blamelessness that an agent who intends to break N and fails is not blameless relative to N. Of course, the failed attempt will be blameless. Compatibly with that, however, the agent’s may (and very often will) not be blameless for intending to break N. Another way of making the same point is that, when assessing the principle, we need to be careful that we substitute consistently for N and f. The ceteris absentibus clause is meant to secure consistent substitution via restriting to one and the same action and norm. ² See Kelp and Simion (2017) for a full normative picture for blameworthiness. ³ Again, to be clear, BLAME needs to be carefully indexed to the same norm throughout: the claim here is not that, just because you are blameworthy for acting/asserting, you are in breach of the norm for action/assertion we are interested in. That is because, first of all, for any norm N governing φ-ing non-uniquely, blameworthiness for acting/asserting can be sourced in a complementary norm governing φ-ing. To use Jessica Brown’s example once again, I can be prudentially blameworthy for telling my boss he is bald, in spite of the fact that I am perfectly well warranted to do so from an epistemic perspective: I know he’s bald. Furthermore, for any norm N governing φ-ing, if one is blameworthy for φ-ing in breach of N, it need not be that the source of blameworthiness is the breach of N: one can be indirectly blameworthy for φ-ing in breach of N by being blameworthy for φ-ing, of which φ-ing is the consequence. For more on this, see Kelp and Simion (2017).

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

44      different norm N*, asking one not to intend to do the wrong thing. To see this, note that if it had been perfectly permissible to form such intentions, the intuition of blameworthiness would also disappear in these cases. All this is perfectly compatible with BLAME, insofar as some norm is violated. In other words, it is important to keep the norm in question fixed throughout BLAME: blameworthiness by norm N implies N-violation. If that is the case, however, one question Williamson needs to answer on pain of ad-hoc-ness is: what is the normative source of the blameworthiness claim involved in his explanation of the contextualist data? What norm makes it such that one should proportion one’s higher-order knowledge to one’s stakes? Of course, it need not be the norm we are interested in. After all, more than one norm can—and usually will— govern any one token action. Take assertion: it is plausible that, on top of KNA, it will also be governed, say, by Gricean relevance requirements. As such, insofar as Gricean relevance requirements are norms proper, one can be blameworthy for making an otherwise knowledgeable, but irrelevant, assertion. Still, the stipulated blameworthiness needs a specified, plausible normative source, on pain of ad-hoc-ness. Now, if Williamson does not want the source to be the main norm at stake, presumably, what will need to be the case is that he stipulates a supplementary norm which triggers the blameworthiness in question. As such, the normative picture that we get is one where, on top of KNA, assertion is also governed by a second norm, call it the Luminosity Norm, asking for an appropriate level of higher-order knowledge relative to the stakes in place: The Luminosity Norm for Assertion (LNA): One’s assertion that p is permissible iff one has the contextually appropriate number of iterations of knowledge that p. Unfortunately, though, once the normative structure is made clear, two fairly serious problems arise for this view: first, consider Williamson’s further lack of luminosity worry. Recall that the reason why Williamson did not just add the requirement for contextually appropriate higher-order knowledge to the norm for assertion itself was that, like any other

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

 

45

non-trivial condition, this condition too can fail to be luminous. That is, you can know that p, and have the contextually appropriate higher-order knowledge also, but fail to know that you do. In this case, again, Williamson will have to say that you are blameworthy for asserting, Note, though, that it’s not clear what will be gained by having two separate norms. After all, insofar as LNA is also a norm-governing assertion, on top of KNA, one can fail to know that the conditions thereby specified obtain, and arguably be blameworthy for asserting under these circumstances in high-stakes scenarios. Second, and more importantly, the plausibility of the sufficiency claim at issue in KNA would appear to be in jeopardy; after all, if LNA also governs assertion, it looks as though, as a matter of fact, being knowledgeable is not sufficient for proper assertion. As far as I can see, there are two options for Williamson at this point: indexing (i) to targets, or (ii) to types of norms. I will look at them in turn. On the first option, LNA governs a different entity than KNA: while the latter is a norm for the speech act, i.e. stating what it takes for an assertion to be a good assertion, the former stipulates what it takes for a speaker to be a good speaker. Williamson (Forthcoming) puts forth a normative framework for blameless believers meant to address the New Evil Demon problem for externalism,⁴ which might prove useful here. According to Williamson, a given norm N typically generates derivative norms, which render subjects who violate the primary norm blameless for so doing: For example, there is the secondary norm DN of having a general disposition to comply with N, of being the sort of person who complies with N . . . There is also a tertiary norm ODN of doing what someone

⁴ This case has been put forth in Lehrer and Cohen (1983) against reliabilist accounts of justification, but they generalize to many externalist views, including Williamson’s JB = K account. Suppose that an evil scientist kidnaps Philip, a normal adult, places his brain in a vat, and hooks it up to a supercomputer that is programmed to induce the kind of sensory experiences he would have had if he had not been kidnapped. The thought is that, intuitively, Philip is as justified after envatment as he was beforehand, although, in fact, his cognitive processes hardly ever deliver truths, not to mention knowledge.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

46      who complied with DN would do in the situation at issue (‘O’ for ‘occurrent’). (Williamson Forthcoming, 6–7)

Correspondingly, maybe LNA is a derivative norm of KNA and what is going on in the contextualist cases is that we have a correct assertion—an assertion that meets KNA, the primary norm—with a blameworthy speaker, in virtue of her being in breach of the derivative norm LNA. One problem with this proposal is, as Williamson also agrees, that one can break norms blamelessly in various ways; the classical ways discussed in the literature are through lack of control or lack of awareness. LNA, of course, will make no exception. In fact, Williamson’s own framework for blamelessness itself can be easily employed to vindicate this thought: one can fail to comply with LNA but still do well by the corresponding derivative norms, call them DLNA and ODLNA. What’s worse, there is excellent reason to think that DLNA may well be satisfied in contextualist cases. To see this, consider ASPIRIN-HIGH once more. You had a choice between performing a proper and an improper speech act. Note that here you chose not to make the improper assertion. You do not assert that you have aspirin at home. Why not? One possible answer is that because you satisfy DLNA. So let’s assume that you do. In that case, it’s plausible that you also satisfy DLNA even if you were to violate LNA on one particular occasion. But then it’s also plausible that had you chosen to make the improper assertion instead, you would still have satisfied DLNA. Finally, note that Williamson endorses that satisfying DLNA entails satisfying ODLNA (Forthcoming, 8). If so, then had you chosen to make the improper assertion instead, you would have satisfied not only DLNA but also ODNLA. But then, by his own lights, we’d expect you to come out violating LNA blamelessly anyway, in which case the blameworthiness just cannot be explained in the way envisaged. Perhaps Williamson should retract the entailment between DLNA and ODLNA and argue that even though, had you chosen to make the improper assertion, you would have complied with DLNA, you wouldn’t have complied with ODLNA, and so you come out blameworthy. Unfortunately, it’s not clear that even this move will work. To see why not, let’s ask what would have been the case had you chosen to make the improper assertion, keeping firmly in mind that you satisfy DLNA. Very

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

 

47

plausibly, the answer is that you would have been in a case in which (i) you know what you assert, and (ii) you are in a borderline case of knowledge, i.e. you don’t know that you know what you assert. Most importantly, given that you satisfy DLNA it is also plausible that you are in a borderline case that is indistinguishable for you from a clear case of knowledge. (After all, had it been distinguishable, since you satisfy DLNA, you would have refrained from asserting.) But pretty much anyone who satisfies DLNA would assert in a case that to them looks just like a clear case of knowledge. And given that this is the case, it is now very hard to deny that you comply with ODLNA also. So, again Williamson’s framework leads us to expect that you to come out violating LNA blamelessly. Another problem with the above proposal is that LNA just doesn’t appear to be plausible as a norm for being a good speaker. To see why, note that LNA is a fairly strongly externalist norm: whether one complies with it in a given case is, to a significant extent, dependent on the cooperation of the world. This, however, makes it ill-suited as a speaker norm. After all, it just doesn’t seem right to blame a speaker when she fails to comply with a norm just because the world happened to not cooperate. As far as I can see, there is only one option available for Williamson at this point: indexing to types of norms. He could do this in two ways. One way would be to argue that, while KNA is constitutive of assertion, LNA is not.⁵ The issue with this reply, though, is that, on Williamson’s own understanding of constitutivity, it’s not clear why that should be so: in Williamson’s (2000) view, a norm is constitutive of a speech act iff it essentially governs it: non-knowledgeable assertions, on this view, are criticizable at all possible worlds. The question then is: why would KNA, but not LNA, satisfy this constraint? Maybe the thought could be that practical stakes are a contingent factor of this world, while knowledge is not. As such, in a possible world where we are practically maximally well, there are no stakes to affect the contextual propriety of assertion. Note, though, that just because there are no stakes in the practically perfect world, it does not follow that LNA does not hold. It may just hold ⁵ Thanks to an anonymous referee for this suggestion.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

48      vacuously, in that more than knowledge is never needed for assertion, since there are never high stakes in place. In a nutshell, this way to go requires more motivation and clarity if it is to work. Another, I think better, option for normative type-indexing would be to argue that KNA is an epistemic norm for assertion, while LNA, i.e. the source of blameworthiness, is a practical norm governing this speech act. On this picture, knowledge is epistemically speaking sufficient for proper assertion; however, practical considerations ask for contextually appropriate higher-order knowledge on top of the epistemic requirement. This reply seems, indeed, quite plausible, and also helps to explain the otherwise somehow weird status of LNA: when your life is at stake, prudentially, it is good to check more on the status of your basis for asserting.⁶ I think this way to go will do the work in explaining the source of the blameworthiness in contextualist cases: it is prudential, rather than epistemic, blameworthiness that is at stake. That being said, I am about to argue that, contra the KK Compatibilist, the prudential blameworthiness in question is not plausibly sourced in a lack of higher-order knowledge. Indeed, as I’m about to show, whether one has knowledge of knowledge in contextualist cases is orthogonal to the Shiftiness Intuition.

4.3 Knowledge of Knowledge of Knowledge Recall that KK Compatibilism’s central claim is that the Shiftiness Intuition can be explained in terms of the subjects’ alleged lack of higher-order knowledge, which both Williamson and Turri accept and employ. One question that immediately arises for KK Compatibilists is just why DeRose fails to know that he knows in the bank cases. Turri does not address this question. In Williamson’s (2000) view, any non-trivial condition is non-luminous, in the sense that it allows for borderline cases where it barely obtains, such that one is not in a position to know that it ⁶ In both print (2005, 233) and conversation, Williamson suggested a prudential reading of the KK requirement; according to him, someone of prudent character will have a disposition to check whether he has as many iterations of knowledge as the stakes require, that is, to ask herself the question: ‘Am I in a position to say that p? Do I know that p? Do I know that I know . . . ?’

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

 

49

obtains. Since knowing is a non-trivial condition, there will be cases in which one barely makes the threshold for knowledge and so is bound to fail to know that one knows. Crucially, according to Williamson (2005), the cases the contextualist exploits pass muster. In other words: No Luminosity: Contextualist cases are borderline cases of knowledge, and therefore exhibit failures of luminosity. While agents in contextualist cases know, they do not know that they know. For instance, on Williamson’s view, strictly speaking, you know that you have aspirin at home in both the ASPIRIN cases. As such, according to KNA, you are permitted to assert that you have aspirin at home. However, your memorial belief barely makes the threshold for knowledge, and thus, in both cases, you fail to know that you know, and therefore you fail to know that your assertion is appropriate. Now, in Williamson’s view, it would be rather unreasonable to insist that you should have second-order knowledge in the low-stakes ASPIRIN-LOW case. However, given that you don’t know that you are meeting the condition for properly asserting, it seems unacceptable to do so in ASPIRIN-HIGH, where your nephew’s health hinges on it. With this point in play, let us look more closely at No Luminosity: note that, crucially, it will adequately explain the Shiftiness Intuition only if it is taken to be a necessity claim: necessarily, all pairs of cases that exhibit the Shiftiness Intuition are borderline cases of knowledge. To see this, just suppose that there are cases exhibiting the Shiftiness Intuition that are not borderline cases of knowledge. In that case, there is no reason to think that the agent lacks second-order knowledge. The intended explanation does not get off the ground. And Williamson will have failed to explain the Shiftiness Intuition in its full generality. The trouble is that, understood as a necessity claim, No Luminosity loses a lot of its initial appeal. In fact, it is quite plausible that the Shiftiness Intuition survives in clear cases of knowledge too. Thus consider the following pair of cases due to Jessica Brown: LO: [S]uppose that Lo truly believes that the seaweed in front of her is correctly classified as of type F, on the basis of the testimony of an

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

50      accompanying expert. She has no reason to doubt the expert’s competence and the expert is in fact reliable. (Brown 2005, 323) According to Brown, LO’s warrant is well above the ordinary standards for knowledge. Therefore, on a non-sceptical invariantist view, Lo knows that the seaweed is of type F. Moreover, her knowledge is not borderline. In fact, it is also plausible to think that she knows that she knows this. It also seems correct for her to say ‘I know that the seaweed is of type F’ and to use the relevant proposition in practical reasoning. HI: Hi is in the same epistemic position as Lo; she truly believes that the seaweed in front of her is correctly classified as of type F on the basis of the testimony of an accompanying [reliable] expert . . . However, [in her context] . . . seaweed F could rapidly come to dominate the local seaweed population, leading to loss of the marine diversity for which the area is internationally renowned. The only way to prevent this loss would be a hugely expensive clean-up programme which would require closure of nearby tourist resorts. Further, in Hi’s context, various error possibilities have been raised, such as the possibility that the expert is mistaken (‘Experts do sometimes make mistakes’). (Brown 2005, 323) It looks as though in HI, but not in LO, it is inappropriate to assert ‘The seaweed is of type F’ or to rely on the proposition that the seaweed is of type F in, say, deciding whether to close the local resort. Furthermore, it looks as though Hi should make further checks by asking one or more other experts for their opinion. Cases like Brown’s suggest that the Shiftiness Intuition can be generated with clear cases of knowledge in which it is plausible that the agent knows that she knows. As a result, the unassertability in high-stakes cases cannot be explained in terms of the absence of second-order knowledge. The KK Compatibilist’s explanation of these cases fails. In response, Williamson ventures to handle such cases by employing contextual variability when it comes to the needed number of iterations of knowledge: the higher the stakes, the higher the order of knowledge required for permissible action. Accordingly, in Brown’s cases, in spite of being in a fairly nice epistemic situation, Hi still misses the contextually

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

 

51

appropriate number of iterations of knowledge: ‘if the stakes are high enough, prudent human agents will engage in third-order reasoning about whether to trust their second-order reasoning about whether to trust their first-order reasoning’ (2005, 233). Furthermore, according to Williamson, this need not be as heavy a burden as it might seem, since in practice we often have enough iterations of knowledge to withstand the interrogations to which it is reasonable to respond. Now, one question that immediately arises at this point concerns the empirical plausibility of this picture; that is, how plausible is it that humans really have such an extended capacity for higher-order thought, as to cover the full possible range of stakes variability?⁷ Is forming the corresponding higher-order beliefs something that we are even capable of? Perhaps Williamson can argue that, even if we might not be able to host these higher-order attitudes explicitly, we can do so implicitly, or dispositionally.⁸ Even if we can, though, the question remains as to whether the processes producing these beliefs are reliable enough to generate the corresponding knowledge or whether we aren’t rather prone to make mistakes at some point (say because we tend to get confused about the exact number of iterations we are looking at). With these points in play, suppose it turns out that our cognitive capacities are limited in the sense that we can wrap our minds around only a fairly limited number of iterations of knowledge. In that case it is just not clear that the Williamsonian picture is rich enough to cover the entire scale of stakes. For instance, suppose that the maximum number of iterations most of us can process is four. The question is why there couldn’t be pairs of cases that generate the Shiftiness Intuition in which the subject has fourth-order knowledge. In fact, it’s not clear why Brown’s cases aren’t cases in which the subjects have the relevant number of iterations of knowledge. I take these considerations to show that the jury is still out on whether the KK strategy is viable at the end of the day. And there are further problems. One of the advantages Williamson claims for his account is that it is but a special case of a more general phenomenon: the idea is that, in general, we tend to harshly judge people

⁷ For discussion, see Kluwe (1982).

⁸ Thanks to an anonymous referee for this point.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

52      who, when a lot is at stake, act without knowing that they meet the condition for proper action. Consider, for instance, the following case: LOCK. The company you work for operates a norm that the door be locked after 10 p.m. Moreover, the last employee is in charge of this. Today you are the last employee and you know this. You also know that it’s of special importance today that the door is locked as your company just received some valuable goods that are likely to be stolen unless the door is locked. In this case, it’s plausible that Williamson is right in that you’ll be subject to blame if it turns out that you didn’t know that the door was locked after 10 p.m., even if, as a matter of fact, it was. By way of evidence, notice that if it does transpire that you didn’t know that the door was locked, we can legitimately criticize you by saying things like: ‘The door was locked. But you didn’t know that. You should have known.’ What’s less plausible is that, as the stakes go up, you need further iterations of knowledge. Consider: LOCK*. The company you work for operates a norm that the door be locked after 10 p.m. Moreover, the last employee is in charge of this. Today you are the last employee and you know this. You also know that it’s of vital importance today that the door is locked. Your company invested all its capital in certain goods that just arrived and are nearly certain to be stolen and result in the company’s bankruptcy unless the door is locked. Here, too, it’s plausible that you’ll be subject to blame if it turns out that you didn’t know that the door was locked, even if, as a matter of fact, it was locked. What’s less plausible is that you’ll be subject to blame if the door is locked and you knew that (you knew that you knew that . . . ) it is locked but you didn’t know that (you knew that you knew that . . . ) you knew that the door is locked. By way of evidence, notice that if all this transpires, it doesn’t look as though we can legitimately criticize you by saying: ‘The door was locked. And you knew that (you knew that you knew that . . . ) it was locked. But you didn’t know that (you knew that

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

 

53

you knew that . . . ) you knew. You should have known that (you knew that you knew that . . . ) you knew.’ What comes to light, then, is that while in general we legitimately blame those who don’t know that they comply with a certain norm, we can’t legitimately blame them in cases in which they comply with the norm and have some number of iterations of knowledge, but don’t attain some higher number of iterations for knowledge. But given that the case of assertion was meant to be but a special case of the general phenomenon, we’d expect assertion to behave in this way also. By the same token, we have reason to think that the KK account remains unsuccessful. Let us take a moment to take stock. We have seen that, on the most plausible way to construe the Williamsonian proposal, the reason why speakers are blameworthy for asserting based on mere knowledge in high-stakes cases is that there is a practical norm that requires more epistemic support and that overrides the epistemic norm. We have also, however, seen that there are serious problems with the higher-order knowledge requirement that, on Williamson’s picture thus conceived, would be featured by this prudential norm. In addition, the fact that Williamson needs to stipulate a further norm of assertion remains a fly in the ointment, as does the lack of luminosity problem for this new norm. But now note that all these difficulties are sourced in the KK Compatibilist’s claim that (prudential) blamelessness requires higherorder knowledge in the contextualist cases. Fortunately, however, the idea that the Shiftiness Intuition is explained in terms of practical norms overriding the epistemic norm is entirely independent of the story about higher-order knowledge. In fact, as I am about to argue, there is a perfectly adequate explanation of the data along these lines available, which remains safely at the first order and so promises to escape all the problems of KK Compatibilism.

Conclusion I have argued that there are problems with all views on the market that attempt to defend the independence of thought from the phenomenon of

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

54      shiftiness of proper speech. I have first argued that Assertion Sensitivism is to be abandoned on prior plausibility grounds. I then looked at the two extant Compatibilist views on the market. Pragmatic Compatibilism, attempting to explain the Shiftiness Intuition in Gricean terms, was shown to fail to adequately account for the Shiftiness Intuition in its full generality. The alternative view, KK Compatibilism, was shown to misidentify the epistemic deficit present in contextualist high-stakes cases. In a nutshell, for now, it’s not looking great for epistemic independence. In the next part of the book, I will embark on a rescue mission on behalf of the independence of epistemically proper thought and speech, and defend an alternative Compatibilist account that relies on the epistemic function of constatives.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

PART II

THOUGHT INVARIANTISM AND SPEECH FUNCTIONALISM

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

5 Against the Shiftiness Dilemma In the first part of the book, we have seen that the vast majority of the literature has embraced the Shiftiness Dilemma: given that, intuitively, proper assertoric speech varies with practical stakes, CI and KNA are incompatible: one of the two needs to go. We have also seen that this dilemma threatens to generalize in three ways: to all context-invariant norms of proper assertoric speech, to all constative speech acts, and to all epistemic normative constraints. If this is right, I have argued, we should be more worried about this dilemma than we’ve been so far. I have also looked at three different attempts to preserve the independence of epistemically proper thought in the face of this threat: first, accounts that grasp the horn of the dilemma that preserves Classical Invariantism by defending context variance for the norm of assertion were shown to be problematic on prior plausibility grounds. Furthermore, Compatibilist views, trying to escape the Shiftiness Dilemma altogether, we found to fail to give a satisfactory explanation of the shiftiness data. All in all, for now, it’s not looking good for epistemic independence. This part of the book develops a novel defence of the independence of proper thought from practical considerations. In order to do that, it starts, in this chapter, by arguing that the Shiftiness Dilemma is a false dilemma: KNA is perfectly compatible with Classical Invariantism. Further on, in Chapter 6, I offer independent reason to believe that if KNA and Classical Invariantism are true, variation in proper assertability is exactly what we may expect. More precisely, this chapter advances the debate in two important ways: (1) it identifies a widely held assumption concerning epistemic norm individuation (Content Individuation), which gets the Shiftiness Dilemma off the ground; (2) it argues that Content Individuation is false, and that therefore the norm at stake in the debate need not be epistemic.

Shifty Speech and Independent Thought: Epistemic Normativity in Context. Mona Simion, Oxford University Press (2021). © Mona Simion. DOI: 10.1093/oso/9780192895288.003.0005

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

58     

5.1 Against Conditional Incompatibility Recall the Shiftiness Intuition: Shiftiness Intuition. Intuitively, permissible assertion requires more epistemic support in high-stakes contexts than in low-stakes contexts. DeRose’s argument for the Shiftiness Dilemma rested on the claim that, given the Shiftiness Intuition, we cannot combine KNA with Classical Invariantism—what I have dubbed the ‘Conditional Incompatibility’ claim: Conditional Incompatibility. If the Shiftiness Intuition is correct, then KNA is incompatible with Classical Invariantism. Since we have already seen that there is excellent reason to think that the Shiftiness Intuition is correct, Conditional Incompatibility gives us excellent reason to think that KNA is indeed incompatible with Classical Invariantism. While Conditional Incompatibility may indeed seem plausible at first glance, I will now argue that it is false. To see why, I’d first like to remind the reader of some relevant and fairly straightforward normativity theory. First, there are many types of action that are governed by a specific type of norm: chess moves are governed by norms of chess, driving by traffic norms, etc. (henceforth I will use norm N as a shorthand for such a specific norm). Now, everyone agrees that assertion is a type of action and the epistemic norm governing it is just such a norm N. Second, all types of actions are governed by certain types of norm simply in virtue of being actions. For instance, all types of action are governed by prudential and moral norms. Crucially, a given action may be improper in one respect whilst being proper in another. When you move the bishop diagonally in chess, your move is proper by the norms of chess. However, if someone also threatens to kill you if you move the bishop diagonally, your so doing will be practically improper. And the same goes, mutatis mutandis, for asserting that 2 + 2 = 4 at gunpoint, say. Your assertion is proper in the sense that it satisfies KNA but practically improper.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

   

59

Third, actions can be assessed not only with respect to whether they are proper in the sense that they comply with the various specific norms governing them, but also whether they are all-things-considered proper. Crucially, it may be that even though a given action conforms to one particular norm, e.g. norm N, it is all-things-considered improper. This is just what happens in the above gunpoint cases: it is all-thingsconsidered improper to move the bishop diagonally and to assert that 2 + 2 = 4 even though the relevant norms N are satisfied here. The reason for this is that requirements of moral and practical norms override the requirements of the norms N. What comes to light, then, is that many norms—plausibly, including the epistemic norm of assertion— are defeasible: they can be overridden by other norms. With these fairly trivial normative points in play, let’s return to the Shiftiness Intuition. Recall that DeRose argues from the intuitive variability in proper assertion to the incompatibility of KNA and Classical Invariantism. Crucially, however, KNA is widely taken to be a distinctively epistemic norm of assertion,¹ that is, a norm that specifies conditions for epistemically proper assertion in particular. But now recall that norms N, including the epistemic norm of assertion, can be overridden by other norms. So suppose for a moment that our propriety intuitions in cases like the ASPIRIN cases do not track distinctively epistemic propriety but some other kind of propriety—plausibly enough, all-things-considered propriety. Since it is entirely possible that an assertion is improper in some other way whilst being proper in the distinctively epistemic sense associated with the epistemic norm for assertion and vice versa, it simply does not follow from the Shiftiness Intuition that KNA and Classical Invariantism are incompatible. By the same token, Conditional Incompatibility is false. What DeRose needs in order to establish the incompatibility of KNA and Classical Invariantism is that there is intuitive variability in the

¹ See Weiner (2005) and Whiting (2013) for defences of a truth norm of assertion, which may be construed as non-epistemic.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

60      distinctively epistemic propriety of assertion, Let’s call this thesis henceforth Epistemic Shiftiness: Epistemic Shiftiness. Intuitively, epistemically permissible assertion requires more epistemic support in high-stakes contexts than in lowstakes contexts. And here is the revamped version of DeRose’s incompatibility claim: Conditional Incompatibility*. If Epistemic Shiftiness is true, then KNA is incompatible with Classical Invariantism. Conditional Incompatibility* is correct. Note, though, that in order for the incompatibility of KNA and CI to follow, it still needs to be the case that Epistemic Shiftiness is correct. The Shiftiness Intuition alone, in conjunction with Conditional Incompatibility*, will not serve to establish the incompatibility between KNA and Classical Invariantism. What DeRose’s claim needs is the following assumption to hold: Epistemic Tracking. The intuitive variation in propriety of assertion with stakes tracks the distinctively epistemic propriety of assertions. In other words, according to Epistemic Tracking, the Shiftiness Intuition tracks Epistemic Shiftiness. Unfortunately, Epistemic Tracking is not all that easy to defend. To see this, consider the following question: given the fact that actions, including assertion, are governed by many norms the requirements of which may conflict in a given case,² how are we to distinguish the intuition triggered by the requirements of the norm we are interested in—that is, the distinctively epistemic norm governing assertion—from the intuition triggered by the requirements of other norms governing it, especially when the requirements are in conflict? ² Jennifer Lackey expresses a similar worry regarding excuse manoeuvres brought in defence of KNA: ‘For now, whenever evidence is adduced that concerns the epistemic authority requisite for proper assertion, it may bear on the norm of assertion or it may bear on these other . . . norms . . . [I]t will be extremely difficult, if not impossible, to tell which is being defended’ (Lackey 2011, 277).

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

   

61

Note also, again, that it is fairly plausible to think that, if anything, our intuitions track all-things-considered propriety, which may conflict with the requirements of pretty much any of the norms that govern the relevant assertion. Given that this is so, however, the prospects of making a convincing case for Epistemic Tracking are dim. There is precious little reason to think that we can get to the condition for distinctively epistemic propriety simply by registering our intuitions about cases. By the same token, DeRose will need to do more in order to get the argument from the Shiftiness Intuition and Conditional Incompatibility* to the incompatibility of KNA and Classical Invariantism off the ground. The trouble lies with making Epistemic Shiftiness plausible. What has transpired is that there is little reason to think that merely appealing to our intuitions about when assertions are proper will do the trick, as there is little reason to think that our intuitions track distinctively epistemic propriety here. Rather, DeRose will need to supplement his methodology with a principled way to distinguish the requirements of the distinctively epistemic norm from requirements of other norms, and show that there is variation in just these requirements in the relevant cases. This will give him Epistemic Shiftiness. And, since Conditional Incompatibility* is fine, it will enable him to establish the incompatibility of KNA and Classical Invariantism.

5.2 Epistemic Norms and Norms with Epistemic Content Now, here is a view about individuating epistemic normative requirements that enjoys a considerable amount of popularity among epistemologists: Content Individuation. If a norm affects the amount of epistemic support needed for permissible ϕ-ing, then it is an epistemic norm.³

³ One can find Content Individuation implicitly assumed in most of the literature on epistemic normativity of belief, assertion, or action in the last decade (see e.g. Benton 2014 for a nice overview). Furthermore, Content Individuation is frequently explicitly endorsed as in the following passages:

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

62      Note that if Content Individuation is a viable way of distinguishing between genuinely epistemic norms and norms of different sorts, Epistemic Shiftiness is true. If this isn’t immediately obvious, consider the following variation of ASPIRIN-HIGH: ASPIRIN-HIGH*. On the way to your place, your sister lets you know that her two-year-old baby urgently needs an aspirin on doctor’s orders and asks you: ‘Do you have aspirin at home, or should we go to the pharmacy?’ You remember having bought aspirin last month. Even so, you think it best to have your partner double check and so you call. When s/he confirms that there is a full pack in the medical box you say to your sister: ‘Don’t worry, we have aspirin at home.’

Here, your assertion that you have aspirin at home is again intuitively entirely proper. Crucially, the only thing that differs between ASPIRINHIGH and HIGH* is that in the latter you have more epistemic support for your assertion than in the former: you not only remember having bought it a month ago, but you also have confirmation from your partner that it is there. But then, of course, it is hard to deny that the reason why the very same assertion would have been improper in ASPIRIN-HIGH is that you do not have enough epistemic support for it. By Content Individuation, the norm that you are violating in ASPIRIN-HIGH is an epistemic norm and so concerns distinctively epistemic propriety. This gives us Epistemic Shiftiness. Since Conditional Incompatibility* holds, the incompatibility between KNA and Classical Invariantism will be established. The trouble with this line is that there is excellent reason to believe that Content Individuation is false, and therefore does not serve to support Epistemic Shiftiness. To see why, I’d like to return once again to the general theory of normativity. Consider, once more, a chess case.

[T]he problem with the agents in the above cases is that it is not epistemically appropriate for them to flat-out assert that p . . . One reason this is clear is that the criticism of the agents concerns the grounds for their assertions. (Lackey 2013, 38) Assertions are governed by an alethic or an epistemic norm—that is, a norm that specifies that it is appropriate to assert something only if what is asserted is true, or justifiably believed, or certain or known. (Maitra 2011, 277)

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

   

63

You are in a situation in which someone will kill you if you make any move. Here doing what is proper by norm N—say, moving the bishop diagonally—is all-things-considered improper. Overriding moral and practical norms render relevant action altogether improper in the sense that you shouldn’t play at all in the circumstances. Crucially, however, rendering the action altogether inappropriate is not the only way in which overriding norms can affect the all-thingsconsidered status of particular actions governed by it. To see this, consider traffic norms: suppose you are driving down a road for which the relevant traffic norm (norm N) prohibits driving faster than 50 km/h. Suppose next that a terrorist group placed a bomb in the town centre and that you are the only person who is able to defuse it. In this case you clearly have a moral obligation to (do your best to) defuse the bomb. But now suppose that, in order to get there in time, you have to break the traffic norm and drive at 90 km/h. Clearly, the moral requirement does not render driving altogether improper. Rather, what happens is that the moral norm overrides the traffic norm and renders driving at 90 km/h the all-things-considered proper thing to do. Notice that, in the two cases above, overriding norms affect the normative profile of the cases in two importantly different ways. First, in the chess case, overriding moral and practical norms render actions that are proper by norm N all-things-considered improper altogether. In contrast, in the driving case, overriding moral and practical norms modify the requirement for all-things-considered proper action: the bomb threat moves the all-things-considered proper speed up to 90 km/h. In sum, overriding norms can affect actions in both of the following two ways: Override1. They can render an action all-things-considered improper altogether. Override2. They can modify the requirements for all-thingsconsidered propriety up or down.

It may be worth noting that Override2 does not only concern the case of driving. On the contrary, similar examples can be construed for many types of action. What matters is that the norms in question regulate how

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

64      much of a gradable property one’s action needs to enjoy in order to be proper. The case of driving is but one instance of this phenomenon. If this isn’t immediately obvious, just note that it can be morally and/or practically required to have a better or a worse grade average, to wear a lighter or a darker dress, to speak louder or more quietly. These considerations suggest that, when a certain type of action is governed by a norm (e.g. a norm N) that requires a certain degree of a gradable property G, a norm overriding this norm can move the threshold for all-things-considered proper action up or down on the G spectrum. Now here is the crucial point: when a certain type of action is governed by a norm (e.g. a norm N) that requires a certain degree of a gradable property G and an overriding norm modifies the threshold for all-thingsconsidered proper action up or down, this does not turn the overriding norm into a norm of the type that it overrides. For instance, in the traffic case, when a moral norm modifies the requirements for all-thingsconsidered proper driving up to 90 km/h, it does not follow that what we are now dealing with is a traffic norm that requires driving at 90 km/h. On the contrary, the traffic norm remains unaffected—it continues to require you to drive no faster than 50 km/h—as does the fact that the norm requiring you to drive at 90 km/h is a moral norm. Rather, what is going on here is that we are dealing with a moral norm that has traffic-related content and the requirements for all-things-considered proper driving coincide with those of the moral norm.⁴ This means that there is a noteworthy difference between a norm of type X (e.g. a traffic norm) and a norm that has X-related content (e.g. a moral norm that tells you how fast you may drive). In particular, and most importantly for present purposes, the fact that a norm has X-related content does not imply that it is a norm of type X.

⁴ Again, the other cases follow suit: just because a norm regulates the appropriate colour of one’s clothes for a funeral, it need not follow it is a fashion norm: it will likely regard social appropriateness. It can be practically appropriate to speak louder or more quietly, no matter what the norms of etiquette require—if one is talking to someone who is hard of hearing, say.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

   

65

5.3 No Shiftiness Dilemma With these points in play, let’s return to the case of assertion. Recall that assertion is a type of action.⁵ According to KNA, epistemically proper assertion requires knowledge. Now, it is widely agreed that knowledge in turn is true belief such that one is in a good enough epistemic position towards the target proposition (henceforth also warranted true belief for short). And, according to Classical Invariantism, the threshold for warrant is of course fixed. Crucially, however, warrant is a gradable property: one can have more or less warrant for a given proposition. This means that, if KNA and Classical Invariantism are both true, assertion is a type of action that fits the normative profile of Override2. To repeat, KNA requires a certain (according to Classical Invariantism fixed) amount of a gradable property (i.e. warrant). We may thus expect overriding moral, practical, etc. norms to be able to modify the degree of warrant needed for all-thingsconsidered proper assertion up or down. Moreover, the fact that an overriding moral, practical, etc. norm so affects the degree of warrant required for all-things-considered action does not turn the norm into an epistemic norm. Rather, here too, what is going on is that the overriding moral, practical, etc. norm has epistemic content and that the requirements for all-things-considered proper assertion coincide with the moral, practical, etc. norm. Here too, we find the difference between an epistemic norm (e.g. KNA) and a norm with epistemic content (e.g. a moral, practical, etc. norm that tells you how much warrant you need to assert properly). And, again most importantly for present purposes, here too, the fact that a norm has epistemic content does not imply that it is an

⁵ One legitimate question at this point is: where does this discussion leave the epistemic norm for action and/or practical reasoning? A full answer to this question falls outside the scope of this book. I have argued elsewhere (Simion 2018b, 2020a) in quite some detail that, in virtue of the distinction between epistemic norms and mere norms with epistemic content, it is a mistake to lump together the epistemic propriety of practical reasoning with that of action. Practical reasoning will be governed by genuine epistemic norms; action will not. See also Fantl and McGrath (2002).

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

66      epistemic norm.⁶ By the same token, there is excellent reason to think that Content Individuation is false. Now, recall that we are in the business of assessing the argument from Epistemic Shiftiness and Conditional Incompatibility* to the incompatibility of KNA and Classical Invariantism. We had already seen that there was little hope that Epistemic Shiftiness could receive the needed support from intuitions about the propriety of assertions in cases like the ASPIRIN cases. And while Content Individuation may at first glance appear to offer a more promising alternative, we have now seen that there is excellent reason to think that there is an important difference between epistemic norms and norms with epistemic content and, as a result, Content Individuation is false. By the same token, it doesn’t serve to provide the needed support for Epistemic Shiftiness either. In sum, DeRose’s argument that KNA and Classical Invariantism are incompatible remains unsuccessful. The original argument from Intuitive Shiftiness and Conditional Incompatibility fails because Conditional Incompatibility is false. This problem can be fixed by replacing Conditional Incompatibility with the more plausible Conditional Incompatibility*. The trouble now is that the uncontentious Shiftiness Intuition will not serve to establish the desired result in conjunction with Conditional Incompatibility*. Rather, what is needed is the stronger Epistemic Shiftiness. And we have now seen that there is no reason to think that Epistemic Shiftiness is true. Either way, the argument does not go through.

Conclusion This chapter has argued that the Shiftiness Dilemma is a false dilemma, resting on an unmotivated normative assumption, i.e. the assumption

⁶ Here is one natural worry that might arise at this point: is the distinction concerning epistemic norms vs. norms with epistemic content merely terminological? What hinges on this distinction? Even if I am right, couldn’t we just continue using ‘epistemic norm’ as standing for ‘norm with epistemic content’? The answer is ‘no’; this would get us in trouble when it comes to fit with the general normative landscape, if we do individuate other types of norms along different lines. This chapter argues that we do. For more discussion, see Simion (2018b, 2020a).

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

   

67

that the felt variation in assertability with stakes is epistemic in nature. As soon as we give up this assumption, Classical Invariantism is perfectly compatible with a knowledge norm of assertion. The next chapter defends a functionalist account of the normativity of assertion that vindicates this compatibility claim.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

6 Assertion Functionalism and Context The previous chapter has argued that the Shiftiness Dilemma is a false dilemma: KNA is perfectly compatible with Classical Invariantism. More precisely, I have first identified a widely assumed assumption concerning epistemic norm individuation (Content Individuation), which gets the Shiftiness Dilemma off the ground. I have then argued that Content Individuation is false, and that therefore the norm generating the Shiftiness Intuition need not be epistemic. This chapter builds on this result and shows how to reconcile Classical Invariantism with the Knowledge Norm of Assertion. My basic proposal is that we can combine invariantism with a functionalist account of assertion: according to the account I favour,¹ assertion is governed by a knowledge norm in virtue of its epistemic function of generating testimonial knowledge. Requirements generated by other functions of assertion, though, such as its prudential function, can override the constraints imposed by the epistemic function, and render the knowledge requirement either too strong or too weak for all-things-considered permissible assertion. All-things-considered permissible assertion can vary with practical stakes; epistemically permissible assertion does not.

6.1 Etiology The previous chapter showed that the normative individuation recipe implicit in DeRose’s argument for the Shiftiness Dilemma is faulty: norms with epistemic content need not be epistemic norms. Content Individuation fails. We need a better individuation recipe if we are to be ¹ I have defended a functionalist account of assertion in several places, e.g. Simion (2018a, 2019d, 2020a, 2020b), Kelp and Simion (Forthcoming). Shifty Speech and Independent Thought: Epistemic Normativity in Context. Mona Simion, Oxford University Press (2021). © Mona Simion. DOI: 10.1093/oso/9780192895288.003.0006

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

   

69

able to tell epistemic norms from practical, moral, etc. norms, and, in turn, if we are to be able to diagnose the normativity that’s efficacious in triggering the Shiftiness Intuition. Now, recall the following fairly uncontroversial value-theoretic claim concerning the relation between the axiological and the deontic that was discussed in Chapter 2: The Association Claim (AC for short). Norms of type X are associated with values of type X. Note that if AC holds, we can individuate values by the type of norms they are associated with, and the other way around: norms of a particular type will be those norms that are associated with values of the relevant type. Here it is: Value Individuation (VI). A norm N is of type T if and only if it is associated with values of type T. Again, just like AC, VI is theoretically neutral, since it does not imply one direction of explanation or another between the axiological and the deontic. I have defended VI extensively in several places (e.g. Simion 2018a, 2018b, 2019c, 2020a). Of course, VI can only go so far in helping us in our individuation efforts: first, if we don’t have an independent grasp of what the values of the relevant type are, VI is not particularly informative for norms individuation. I gather, though, that we have a clearer grasp on which goods are epistemic goods² (e.g. knowledge, justification) rather than, e.g., prudential goods (e.g. pleasure, money). Still, a theoretically loaded choice for some goods over others will have to be made before putting VI to work. Another limitation of VI is that, in virtue of its theoretical neutrality, it does not yet make clear predictions: depending on whether one is a consequentialist or a deontologist, for instance, different norms might be diagnosed as epistemic by VI. We need to take a theoretical stance on the issue.

² Again: I use goods and values interchangeably throughout. Nothing hinges on this.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

70      In previous work, I have defended extensively a right-to-left direction of explanation for VI, on functionalist grounds. Here is the VI incarnation and normative picture I favour:³ in traits, artefacts, and actions alike, functions generate norms. There is such a thing as a properly functioning heart, a properly functioning can opener, etc. We should expect, then, that if there is such a thing as a proper assertion, it will somehow be associated with the function it serves. It would be useful, then, given that what we are after is the epistemic norm of assertion, to begin by identifying the epistemic function thereof. To this effect, in what follows, I will be employing an etiological account of functions.⁴ On the etiological theory of functions, functions turn on histories that explain why the item exists or operates the way it does. Take my heart; plausibly, tokens of the type pumped blood in my ancestors. This was beneficial for my ancestors’ survival, which explains why tokens of the type continued to exist. As a result, my heart acquired the etiological function (henceforth also e-function)⁵ of pumping blood. Acquiring an etiological function is a success story: traits, artefacts, and actions get etiological functions of a particular type by producing the relevant type of benefit. My heart acquired a biological etiological function by generating biological benefit. Through a positive feedback mechanism—the heart pumped blood, which kept the organism alive, which, in turn, ensured the continuous existence of the heart—our hearts acquired the etiological function of pumping blood. Now, not all functional items follow the model of the heart: there will be cases where a requirement of selection over generations for function

³ For a book-length defence of this normative picture, see (Simion 2020a). ⁴ Defended by people like e.g. David J. Buller (1998), Ruth Millikan (1984), Karen Neander (1991). The etiological theory of functions is, by far, the most widely endorsed view in the literature. Its main competitor is the ‘systemic’ theory of functions, notably defended in Cummins (1975). Systemic functions describe how something works or operates—what it does—as a part of a larger system. Functions, in this sense, are the causal role capacities of parts that contribute to some capacity of the containing system. Systemic functions are widely believed to lack normative import, which is what explains, to a large extent, both the popularity of the competing, etiological account, and why the latter is thought to be much better suited for applications to normative domains like epistemology. ⁵ For applications of the etiological account of functions to the normativity of belief, see e.g. Millikan (1984) and Graham (2012). For a knowledge-first incarnation thereof, see Simion (2019a, 2019c, 2020a).

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

   

71

acquisition will seem implausibly strong (Sosa 1993).⁶ The paradigmatic case is that of beneficial macro-mutations, also known as hopeful monsters (Graham 2014, 30). Most mutations are harmful (think of extreme birth defects); once in a while, though, a happy accident happens: someone is born with an almost entirely new trait or organ, very different in kind from its ancestral trait, which actually benefits the recipient. Since they are mutations, they don’t have an evolutionary history; they are ‘first-generation’ traits. Still, many biologists would say that they have functions. While etiology does require some history, it does not require an awful lot of it (Graham 2014); there are several ways to cash out the etiological requirement that do not presuppose directional selection, i.e. selection over generations. What is essential to function acquisition is the positive feedback loop—whereby the trait in question generates a beneficial effect that contributes to its own continuous existence. Thus, a trait can also acquire a particular function by ongoing maintenance selection, or through a learning process, or even by the metabolic activity of the organism itself. What it all amounts to, eventually, is explaining the existence/continuous existence of a trait through a longer or shorter history of positive feedback. Here is the account of e-functions that I favour and I have defended extensively in previous work:⁷ E-Function. A token of type T has the e-function of type B of producing effect E in system S iff (1) tokens of T produced E in the past, (2) producing E resulted in benefit of type B in S/S’s ancestors, and (3) producing E’s having B-benefited S’s ancestors contributes to the explanation of why T exists in S.

⁶ Take also Davidson’s Swampman, for instance (1987, 443). Swampman comes into existence as a result of a lightning bolt hitting a swamp and strangely rearranging gas molecules into the exact duplicate of some ordinary person. Now, according to traditional proper functionalism, beliefs that enjoy epistemic entitlement are beliefs formed via cognitive processes that have the function of reliably producing true beliefs; since function acquisition requires history, and since Swampman’s processes lack any, it would seem as though he cannot have beliefs that enjoy epistemic entitlement. However, intuitively, at least after acquiring the necessary concepts, Swampman seems entitled to his belief that he is sitting in a swamp (Sosa 1993). See also Graham (2014) for an excellent response and a short-history proper functionalism. ⁷ E.g. Simion (2019a, 2019c, 2020a).

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

72      One thing that is important to note for our purposes here is that, in contrast to the traditional etiological functionalisms about the epistemic (by Peter Graham (2012) and Ruth Millikan (1984)), the account defended here takes functions to be typed by the corresponding benefit. As such, if a trait produces a benefit of type B in a system, the function thereby acquired will be a function of type B. If it is biological benefit that is at stake in function acquisition, what we get is a biological function. On this account, the heart’s function to pump blood is a biological function in virtue of the fact that the produced benefit is also biological—i.e. survival. The function of music is an aesthetic function in virtue of the fact that the produced benefit is an aesthetic benefit. And so on. Now, of course, aesthetic benefit might, and often will, also result in biological benefit. This, however, in no way renders the function at stake a biological function. What is important to keep in mind is that the benefit that is essential to aesthetic function acquisition is the aesthetic one. The fact that biological benefit is also associated with the latter is a mere contingent matter of fact. Now, note that the etiological account is an account of functions as purposes: by being selected for it, our hearts have acquired the purpose of pumping blood in our organisms (Graham 2012, 449). Reaching that purpose—i.e. successfully pumping blood—will amount to function fulfilment. But purposes will also come with associated norms prescribing the right way to proceed in order to reliably reach the corresponding purpose. Because its function contributes to the explanation of its very existence, the trait in question ought (in the relevant sense—e.g. in the case of the heart, biologically ought) to perform in a way that is associated with likely function fulfilment. According to the etiological theory of functions, this is but the way in which the trait functioned back in the day when it acquired its function. Your heart will pump blood in normal conditions, i.e. conditions similar to those in which it was selected, when functioning normally, that is, when functioning in the way in which it was functioning when it was selected for its beneficial effects. Plausibly, in normal conditions, a normally functioning heart will fulfil its function of pumping blood in your system by beating (at a particular rate etc.). According to the etiological theory, then, normal functioning is proper

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

   

73

functioning: a heart functions in the way it should (i.e. by the norm) when it functions in the way it did back in the day when it acquired its function: when it beats. The e-functionalist picture constitutes itself in a straightforward norm-identification machinery; first, what we need is to take a look at the relevant function plausibly served by the trait/artefact/action in question. Once the function is identified, the question we need to ask ourselves is: how did the trait/artefact/action plausibly fulfil its function at the moment of function acquisition? The answer to this question will give us normal functioning which, on the etiological account, corresponds to proper functioning; therefore, the answer to this question delivers the content of the norm we are after. Furthermore, on this picture, we also get an easy way to identify the type of norm at stake: norms will be typed by the corresponding functions, which, in turn, are typed by the produced benefit. The norm ‘hearts ought to beat’ is a biological norm in virtue of the fact that it serves a biological function—pumping blood—which, in turn, produces a biological benefit.

6.2 A Functionalist Account of Assertion To see how this works in practice, let us return to assertion: our speech acts too can, and most—hopefully!—will, serve etiological functions, which, again, will come with associated norms. Plausibly, if there is such a thing as an epistemic norm for assertion out there in the first place, it is likely there to ensure that assertion delivers the epistemic goods we are using it for (Goldberg 2015, Grice 1989, Kelp 2018, Reynolds 2002). What epistemic goods is assertion meant to deliver? Although not essentially—I can, say, make assertions in a diary, which are usually not intended to affect any audience in any way— characteristically, assertions will aim at generating testimonial knowledge in the audience. Plausibly, this is the main epistemic function of assertion (see, e.g., Kelp 2018, Kelp and Simion Forthcoming, Reynolds 2002, Turri 2016). Due to our physical and cognitive limitations, a lot of

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

74      the knowledge we have is testimonial; thus, assertion is one of our main epistemic vehicles. Just as hearts were selected for their reliability in generating biological benefit, I submit, the speech act of assertion has been selected for its reliability in generating epistemic benefit, i.e. testimonial knowledge. As such, our assertoric practice acquired the epistemic etiological function of generating testimonial knowledge in hearers. Because it generated testimonial knowledge in our ancestors, it enabled them to survive— find out about the whereabouts of dangerous predators, find food, and so on—and reproduce, thereby replicating the same practice with the same function in their descendants. The fact that assertion generates testimonial knowledge in hearers explains the continuous existence of the practice.⁸ Dan Sperber puts this point succinctly: ‘From the point of view of receivers, communication, and testimony in particular, is beneficial only to the extent that it is a source of genuine . . . information (2001, 404). However, if the practice stops being beneficial to the hearers, it will plausibly be discontinued: hearers will stop believing what they are being told, and, eventually, since their assertions remain inconsequential, speakers will stop asserting. Just like your heart’s pumping blood keeps you alive which, in turn, contributes to the explanation of the continuous existence of the heart, the function of assertion is a stabilizing one, for it ‘encourages speakers to keep using the device and hearers to keep responding to it with the same (with a stable) response’ (Millikan 2005, 94). The speech act of assertion will reliably generate testimonial knowledge in hearers in normal conditions, i.e. conditions similar to those in which it was selected, when functioning normally, that is, when functioning in the way in which it was functioning when it was selected for its beneficial epistemic effects. When functioning normally, the speech act of assertion will fulfil the epistemic norm constitutively associated with its epistemic e-function of reliably generating testimonial knowledge; it will work the way it is supposed to work, where the right way of working is partly constituted by reliably delivering the epistemic goods in normal ⁸ For a detailed argument in favour of a knowledge rather than truth function of assertion, see Kelp and Simion (Forthcoming) and Simion (2020a).

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

   

75

conditions. Thus, an epistemically proper assertion will be one that, in normal conditions, generates knowledge in the hearer in the way it did back when it acquired its epistemic function. The question to ask, then, if we want to know the content of the epistemic norm we are after, is: how did assertion fulfil its function of generating knowledge in hearers at the moment of function acquisition? I submit that the overwhelmingly plausible answer is: by being knowledgeable. And here are some reasons that, in my view, leave room for very little doubt that this is so: first, on most if not all accounts of testimony in the literature, in the vast majority of cases, the speaker needs to know in order to be able to generate knowledge in the hearer. Furthermore, knowledge is all the speaker needs to this effect when it comes to her epistemic standing vis-à-vis p.⁹ Also, exceptions to this rule describe extremely unusual scenarios,¹⁰ which render them highly unlikely to affect the argument from testimony to the knowledge norm in any way. After all, if the function of assertion is generating testimonial knowledge, and in the vast majority of the cases knowledge on the part of the speaker is both needed and enough for generating testimonial knowledge in the hearer, it makes sense to have a knowledge norm governing assertion. To see this, consider driving again: norms regulating speed limit within city bounds are presumably there to make it so that we arrive safely at our destination. Surely, though, driving at 50 km/h within city bounds is not always the ideal speed; there are instances when, for instance, overtaking at 80 km/h will avoid a major accident. However, presumably, the reason why the norm says ‘Drive at most 50 km/h within city bounds!’ is because, most of the time, that is the ideal speed for safety purposes. Due to our finite calculation capacities, leaving it flexible—i.e. having an ⁹ Williamson (2000, 256) makes a similar point, although he does not pursue this line any further: ‘Although there are special cases in which someone comes to know p by hearing someone who does not know p assert p . . . the normal procedure by which the hearer comes to know p requires the speaker to know p too.’ ¹⁰ Exceptions are few, and they roughly boil down to two types of cases: first, we have, again, ‘selfless asserters’, asserting on knowledge-level justification without belief (Lackey 2007). These speakers assert to what is best supported by evidence, although they cannot get themselves to believe it due to some rationality failures. Secondly, we have ‘Compulsive Liar’ cases (Lackey 2008). Roughly, what happens in these cases is that, although the speaker intends to lie on a regular basis, some external intervention makes it so that she safely asserts the truth.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

76      alternative norm along the lines of ‘Drive at most 50 km/h within city bounds, unless it seems to you that one can arrive safely and more quickly by driving 50+n km/h, in which case drive more than 50+n km/h’¹¹—would likely result in disaster; this is a familiar point from ethics, concerning the calculation problem for act utilitarianism (Williams 1973, 1981). On top of all this, note that, in the friendly epistemic environment we inhabit, knowledge is readily available (Kelp and Simion 2017).¹² All we have to do to acquire knowledge is open our eyes, listen to what other people tell us, attend to our feelings, etc. As such, KNA is not a very stringent norm; it amounts to a fairly user-friendly way to ensure reliable function fulfilment.¹³ If that is the case, again, it is plausible that, back at the time of function acquisition, assertion fulfilled its function of generating knowledge in hearers by being knowledgeable. In sum, things seem to stand as follows: in the friendly epistemic environment we inhabit, knowledge is readily available. On top of this, knowledge on the speaker side is needed for generating knowledge on the hearer’s side in the vast majority of cases. Very plausibly, the way in which assertion generated knowledge in hearers at the moment of function acquisition was by being knowledgeable. As such, KNA is vindicated: an epistemically proper assertion will be one that, in normal

¹¹ Thanks to an anonymous referee for pressing me on this point. ¹² Of course, the plausibility of this claim, as well as of several other empirical claims about the amount of knowledge had by our ancestors, will, to some extent, depend on how stringent one’s view of the nature of knowledge is. As a limit case, the sceptic will likely not be very convinced by this picture. That being said, I would like to think that I’m not losing the sceptic in the process of putting forth this view, for two reasons: first, while my favorite non-shifty norm of assertion is KNA, the functionalist picture, together with its explanation of the shiftiness data, survives even in the absence of this commitment. Second, most sceptics accept that, even though we don’t have a lot of knowledge, we have a lot of something weaker that we employ in our everyday life—call it shmoledge. If so, the sceptic can read this account as being about the shmoledge norm of assertion. ¹³ To be clear, while the prevalence of knowledge in our environment is sufficient for me to make this point, it is not necessary: a system or practice can acquire a function so long as it is successful enough of the time to confer some survival benefit on the system or practice having that function-constituting feature. This need not be reliable: Millikan (1984) emphasizes the unreliability of the process by which e.g. sperm fertilize eggs, rabbits’ danger-detecting systems actually identify danger, etc. Thanks to an anonymous referee for pointing out that I should clarify this.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

   

77

conditions, is able to fulfil its epistemic function of generating knowledge in hearers by being knowledgeable.¹⁴

6.3 Knowledge vs. Truth Contra the knowledge account defended here, Ruth Millikan (1984) and Peter Graham (2010) take the function of assertion to be generating true belief rather than knowledge in one’s audience. If they are right, one could wonder, would it not be reasonable to suppose that assertion is governed by a corresponding truth norm? The short answer is ‘no’: even if Graham and Millikan are right, and the function of assertion is generating truth, proper functioning will still amount to asserting knowledgeably; after all, given that knowledge is easily available, very plausibly, at the moment of function acquisition, assertion generated true belief in hearers by being knowledgeable. If so, knowledge is the condition for assertion’s normal functioning. That being said, I think there is reason to doubt the true belief function claim to begin with. To see this, let’s start by envisaging how the argument for a true belief function, and against a knowledge function, would go. Recall the account of etiological functions we are working with: E-Function. A token of type T has the e-function of type B of producing effect E in system S iff (1) tokens of T produced E in the past, (2) producing E resulted in benefit of type B in S/S’s ancestors, and (3) producing E’s having B-benefited S’s ancestors contributes to the explanation of why T exists in S. It may seem as though true belief would do better on clause (3) than knowledge. That is, it may seem as though what genuinely contributes to the explanation of the continuous existence of the practice of assertion is ¹⁴ Simion (2020a) argues that the functionalist schema serves to construct a better model of the nature of assertion as well as of its normativity. In a nutshell, I argue that the knowledge norm is constitutive of assertion insofar as it contributes to this speech act’s continuous existence. I also argue that this light-touch constitutivity model compares favourably with the mainstream model on the market, the games model defended by Williamson (2000).

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

78      its success in generating true belief. After all, the argument would go, true belief (about the whereabouts of food, predators, etc.) is what kept us alive and well: knowledge is not needed.¹⁵ Had assertion not produced true beliefs, our species may well have disappeared. The same does not hold for knowledge: true belief falling short of knowledge would have been enough to keep the species alive. Or so the thought would go. Note, though, first, that (3) only requires that the benefit produced by past tokens of a type contributes to the explanation of why the system continues to exist. Can it be the case that such a benefit contributes to the explanation of why the system continues to exist even though it is not the case that had tokens of the type not produced the benefit, the system would not have existed? As I am about to argue, the answer to this question is yes. In order to achieve this, I will argue that the satisfaction of (3) for traits in general does not generally require the truth of a counterfactual of the form had past tokens of the product not produced the benefit, the system would not have existed today. I will then offer an alternative account of what it takes to satisfy (3) and will argue that knowledge does satisfy (3).¹⁶ Consider your heart: past heart tokens reliably pumped blood in your ancestors’ circulatory systems, which kept them alive, and which contributes to the explanation of why your heart exists today. Now, let’s suppose that past hearts had not reliably pumped blood in your ancestors. Does this imply that your heart would not have existed today? The answer is no: it could have still been that the species would have been kept alive by some other, alternative blood pumping mechanism. Furthermore, it could have even been the case that your heart would exist today: after all, we do carry around useless vestigial traits for generations. By the same token, your heart does not satisfy the counterfactual had past tokens of the product not produced the benefit, the system would not have existed today. Does this mean that your heart does not have the function of reliably pumping blood in your circulatory system? The answer is ‘no’; rather, what transpires is that the proposed counterfactual reading of (3) is too strong. We need an alternative account of ¹⁵ Thanks to an anonymous referee for pressing me on this. ¹⁶ For more on this, see (Kelp and Simion Forthcoming).

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

   

79

what it is to contribute—in the relevant way—to the continuous existence of a trait. In previous work (Kelp and Simion Forthcoming, Simion 2020a), I suggested the following answer: (3) is satisfied if the benefit was actually produced and actually was part of the instantiation of the causal structure in virtue of which the explanation of why the trait continues to exist comes out true. If so, it is easy to see that the benefit of pumping blood in our circulatory systems does contribute to the explanation of why hearts continue to exist today: the benefit was actually produced by past hearts and is part of the instantiation of the causal structure in virtue of which the explanation comes out true. Now, I want to suggest that, analogously, assertions can have the e-function of generating knowledge in hearers even if the practice of assertion would still have existed today had they not generated knowledge in hearers. What matters to whether (3) is satisfied is whether the relevant benefit was actually produced by the trait and that it is actually part of the instantiation of the causal structure in virtue of which the explanation comes out true. And this is the case for both hearts and assertions. Just as past hearts actually pumped blood reliably in the circulatory systems of our ancestors, so past assertions actually generated knowledge in hearers. Moreover, just as the benefit of blood pumping is actually part of the instantiation of the causal structure in virtue of which the explanation of why hearts continue to exist today comes out true, so the benefit of generating knowledge in hearers is actually part of the instantiation of the causal structure in virtue of which the explanation of why the practice of assertion continues to exist today comes out true. If so, assertion does have generating knowledge in hearers as an epistemic e-function.

6.4 Overriding Functions Let us now return to DeRose’s argument for the Incompatibility Claim, and see how the functionalist account proposed here deals with it. To begin with, note that a trait/artefact/act can have several e-functions simultaneously, even several e-functions of different types; take, for

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

80      instance, the functions served by food for humans. One important such function will surely be a biological one, a nourishment function. Plausibly enough, though, on top of this, food also serves an aesthetic function, that of generating pleasant gustatory experiences. Now, normally, the aesthetic function complements the nourishing function. It serves, as it were, the greater goal of survival, by increasing the probability of us ingesting nourishing substances. This need not be the case, however; there can be situations where the two functions come in conflict, at which point the more survival-stringent requirement will take precedence and dictate what’s the all-things-considered good to observe. Think about a case where I am on a deserted island and all I can eat in order to stay alive are my boots; surely, against my aesthetic well-being, that is exactly what I should do, all-things-considered. How does this overriding mechanism work in cases of conflict? The literature in normativity theory offers several recipes for this. My preferred account, the ‘balance’ account (e.g. Lord 2018, Schroeder 2015), explains normative overriding in terms of weighing normative constraints. According to this account, very roughly, considerations in favour of different alternative actions at a context are balanced against each other: considerations against one option are reasons in favour of the alternative options, and vice versa: considerations in favour of one of the alternative options are considerations against the option under consideration.¹⁷ Now, recall that, on my view, the normative is generated by the valuable: norms come from functions. The more important the function for the survival of the organism, the weightier the normative constraints generated by it will be. This account will predict, indeed, that one should eat one’s boots on the deserted island: normative constraints generated by the prudential function are more survivalstringent than those generated by the aesthetic function in this case. Prudential constraints sourced directly in the survival function will count in favour of eating the boots, and thus will constitute considerations against not eating the boots, which will come in conflict with

¹⁷ The view is traditionally formulated in terms of weighing reasons. This will not make any difference for our purposes here. For a functionalist view on reasons, see Simion (2020a).

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

   

81

and override the considerations in favour of not eating the boots sourced in the aesthetic function.¹⁸ It is highly plausible that the epistemic function is the main function of assertion: that’s what we mainly do with assertions, we transmit information.¹⁹ That being said, it is also highly plausible that the epistemic function is merely one of the many functions served by assertion. Surely, when I tell you that the weather is nice while sitting with you in a café, generating testimonial knowledge in you with regard to meteorological states of affairs is hardly among my chief concerns. In this case, assertion rather serves a social bonding function. Just because the epistemic function of assertion is its main function, however, it does not follow that the norms generated by it cannot be overridden by norms sourced in its secondary functions: indeed, this will likely happen, for instance, when the normative constraints generated by the prudential function will be more survival-stringent and thus will outweigh the ones in favour of the epistemic function.²⁰ One of the most important functions of assertion, as with action in general, will be a prudential one, serving our survival. Again, just as in the case of food, the epistemic function will, in most cases, complement this prudential function. Generating testimonial knowledge in one’s hearer with regard to an imminent threat, or about the whereabouts of resources, are paradigm cases. However, again, this need not be the case. For instance, even if one knows that one’s boss is bald, it may not be polite, prudent, or relevant to point this out to him (Jessica Brown 2010, 550): the requirement triggered by the prudential function comes in conflict with the one corresponding to the epistemic function and overrides it.

¹⁸ Many thanks to an anonymous referee for pressing me to say more about the overriding mechanism. ¹⁹ Other functions—social, prudential, etc. functions—are plausibly secondary functions of assertion: one useful test to see this is the ‘defect assertion qua assertion’ test. A prudentially permissible lie is intuitively a defect assertion qua assertion. A prudentially impermissible knowledgeable assertion is not. This suggests that the epistemic function is the main function of assertion. ²⁰ Note that it is not my intention to suggest that practical concerns will always override epistemic ones (or aesthetic ones, for that matter). The claim made (and needed) here is weaker than that: practical requirements will do so when they will be more survival-stringent. I gather that this is a fairly plausible thought.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

82      Here, then, is my proposal: both Classical Invariantism and KNA hold. In virtue of the function of assertion of generating knowledge in hearers, one is in a good enough epistemic position to make an epistemically proper assertion that p if and only if one knows that p. In turn, whether one knows (or ‘knows’) that p is insensitive to practical matters. Crucially, on the account defended here, what KNA claims is mere epistemic propriety. The standards for all-things-considered propriety of assertion will vary with context, while standards for both epistemically proper assertion and knowledge will remain fixed. The way to ascertain whether the requirements at work in one case or another are genuinely epistemic requirements is by looking at the function that is plausibly being served. An epistemic function will be associated with an epistemic norm. To see how this works, consider, first, Williamson’s TRAIN case: TRAIN. Suppose that I, knowing that it is urgent for you to get to your destination, shout ‘That is your train’ upon seeing a train approach the station, although I don’t know that is the case; I merely believe there is some chance that is the case, and I think you should check it out. (Williamson 2000) According to the present proposal, the assertion in TRAIN is epistemically defective, but all-things-considered proper, due to prudential considerations stepping in and requiring a lower degree of warrant. In view of the epistemic function of generating testimonial knowledge, this assertion is not permissible. However, in view of the prudential function of raising your chances of catching the train, the assertion in TRAIN is perfectly fine. Given that the prudential constraint overrides the epistemic constraint in this case (as in most), by lowering the degree of (all-things-considered) necessary warrant, the assertion is all-thingsconsidered proper, which is what triggers our relevant intuitions. Similarly, in high stakes such as DeRose’s BANK CASE HIGH, strictly epistemically speaking, speakers are permitted to assert. They are in a position to make an assertion that, in normal conditions, will fulfil its epistemic function of generating testimonial knowledge in their hearer. However, prudential constraints referring to the tragic consequences

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

   

83

their assertion might have (the cheque might bounce!) override the epistemic constraint, raise the necessary degree of warrant, and make the assertion all-things-considered improper. As such, again, it is allthings-considered propriety that varies with context, not epistemic propriety.

6.5 Contextualist Rejoinders In what follows, I would like to take a brief look at what I take to be the main two objections that the champions of the Shiftiness Dilemma are likely to press when presented with my account, i.e. objections from knowledge denial in high-stakes cases and from shifty data for belief.

6.5.1 Objection 1: Knowledge Denial Data On the view defended here, the contextualist data are explained by normative overriding: while epistemically fine, it is prudentially impermissible for DeRose to assert ‘The bank will be open tomorrow’ in the high-stakes scenario. Recall, however, that, according to DeRose, not only can he not assert ‘The bank will be open’ or the corresponding ‘I know that the bank will be open’ in high-stakes cases, but it is even appropriate for him to deny knowledge to himself. Now, what does the functionalist picture have to say about this? Two things: first, several people (e.g. Pritchard 2005) have expressed doubts about the plausibility of the knowledge denial data. I tend to agree. Here is why: note that, if at all, it looks as though the natural way to deny knowledge to oneself in high-stakes scenarios goes along the lines of ‘All right, I guess I don’t really know the bank will be open tomorrow!’ rather than the bare: ‘I don’t know that . . .’. But it is fairly plausible that the epistemic standing at stake in this assertion—i.e. ‘really knowing’—may be a different epistemic standing than knowledge— something stronger, closer to certainty. To see the plausibility of this, compare, for instance, with ‘liking John’ and ‘really liking (in the sense of

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

84      having a crush on) John’. It looks as though ‘really liking’ is but a different state than liking. It is, I submit, not completely implausible that the same is true for ‘really knowing’, in which case DeRose’s denial does little to prove the invariantist wrong. Or to make the even weaker claim: the intuitive need for the ‘really’ modifier in this case is in need of explanation. Second, more importantly, even if we buy the knowledge denial data, they are hardly problematic for the functionalist. After all, recall that DeRose’s denial comes as a reply to his wife’s question: ‘Do you know that the bank will be open tomorrow?’ It looks as though, then, at that point in the conversation, DeRose is faced with a choice between (1) odd silence, (2) attributing knowledge to himself—which would likely, in the absence of any qualification, result in a risky course of action, (3) saying something along the lines of: ‘I guess we should check, just to make sure!’, or (4) denying knowledge to himself in order to stay on the safe side. Of course, one of the latter two options will be the (prudentially!) preferable one. To my ear, option (3) is the most natural one. Whichever it is, however, it will not hurt the view defended here, since, again, it is fairly clear that prudential constraints are at work in the choice of speech act. If so, the apparent propriety of (4) need not be generated by its being true: it can be prudential propriety alone that explains the intuition in favour. As such, more needs to be done to argue in favour of the contextualist claim that what generates the favourable intuition here is also the truth of the claim in question rather than its prudential propriety alone.

6.5.2 Objection 2: Shifty Data for Belief What about belief? After all, one could think that the relevant shiftiness seems also to be present at this level. It looks as though, one could argue, in high stakes, after his wife raises the possibility of the bank having changed its hours, DeRose is not warranted in holding an outright belief that he knows that the bank will be open on Saturday.²¹ Similarly for the ²¹ DeRose stipulates that he remains as confident as before after the error possibility is raised. The question remains, though: is this warranted at that point?

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

   

85

generalized data—an outright belief that the bank will be open also seems impermissible. What does the account defended here have to say about this? After all, while it is plausible that DeRose’s assertion might trigger a prudentially risky course of action, that seems less plausible about his belief. Also, if belief turns out to be governed by a shifty norm, this is fairly bad news for the invariantist picture defended here—insofar as we take belief to be justified if and only if permissible. It is less than clear to me however that, insofar as the intuition of impropriety extends to belief, it is not prudential impropriety that is at stake in this case also. To see why, note that, as the case is built, plausibly, DeRose’s outright belief that the bank will be open on Saturday in BANK CASE HIGH is as likely to be conducive to risky behaviour as is his corresponding assertion. Furthermore, it looks as though one easy way to ascertain whether my account gets things right here—when it goes to both assertion and belief—is to try to build a case where asserting/ believing the relevant proposition is not likely associated with any risky action. If the Shiftiness Intuition is preserved, this raises serious worries about the prudential nature of the permissibility at stake. Note, though, that building such cases will not be a trivial affair. One way to try and do this will be by removing practical stakes altogether from the picture, and relying on the mere tabling of error possibilities. It is not clear, however, that the Shiftiness Intuition survives this move. After all, again, I seem to be perfectly warranted in both my belief and the corresponding assertion that I am just about to make that there is a computer in front of me, in spite of, say, sceptical scenarios being tabled.²² Alternatively, one can try to build a case where the subjects falsely believe high stakes are present, but, by stipulation, no risky action is likely to follow from DeRose’s assertion/belief. Say that, unbeknownst to DeRose, his rich grandmother deposited a large sum of money in his account as a present for his birthday, which will be more than enough to cover the cheque. Doesn’t the Shiftiness Intuition survive this alteration of the case? First, I am not sure it does, as a matter of fact. I can get myself to think both ways. Second, more importantly, even if we find a way to build the cases such that the Shiftiness Intuition is preserved, the ²² See also Turri (2010) for a similar argument.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

86      explanation in terms of prudential propriety is still not out of the picture. After all, many people in the literature think of prudential propriety as a subjective affair. On a Subjective Bayesian picture, for instance, what matters for prudential propriety is not the actual state of the world, but the subjective probability assigned to the relevant outcome. As such, it is not clear that building such cases will do the intended job against my view to begin with.

Conclusion This chapter has argued that Classical Invariantism fits snugly with a functionalist knowledge account of assertion. According to this view, knowledge is the norm of assertion in virtue of assertion’s function of generating knowledge in hearers. Requirements generated by other functions of assertion, though, such as its prudential and moral functions, can override the constraints imposed by the epistemic function, and render the knowledge requirement either too strong or too weak for all-thingsconsidered permissible assertion. As such, all-things-considered permissible assertion can vary with practical stakes; epistemically permissible assertion does not.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

7 Knowledge Norms for Constative Speech Acts The previous two chapters showed that the Shiftiness Dilemma is a false dilemma, and defended a functionalist recipe for reconciling Classical Invariantism with the Knowledge Norm of Assertion. According to the account I defended, assertion is governed by a knowledge norm in virtue of its epistemic function of generating testimonial knowledge. Requirements generated by other functions of assertion, though, such as its prudential function, can override the constraints imposed by the epistemic function, and render the knowledge requirement either too strong or too weak for all-things-considered permissible assertion. Allthings-considered permissible assertion can vary with practical stakes; epistemically permissible assertion does not. Now, recall that I started off in Chapter 1 by worrying that the Shiftiness Dilemma threatens all constative speech. In this chapter, then, I will show how my optimistic results generalize beyond assertion, to the epistemic normativity of all constative speech acts. In doing so, I put forth the first integrated account of the normativity of constatives. Assertion is hip, and it has been so for a while now: a lot of ink has been spilled in the literature discussing its nature and normativity. This is hardly a historical contingency: in philosophy of language, we care a lot about assertion because the default use of declarative sentences is to make assertions (Williamson 2000, 258). In epistemology, it is somewhat unsurprising that we care so much about the speech act that is our main vehicle for testimonial knowledge acquisition, which, in turn, constitutes the vast majority of the knowledge we have: we are psychologically and physically limited creatures, we know most of the things we know from the epistemic labour of others.

Shifty Speech and Independent Thought: Epistemic Normativity in Context. Mona Simion, Oxford University Press (2021). © Mona Simion. DOI: 10.1093/oso/9780192895288.003.0007

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

88      In contrast, other constatives have enjoyed far less attention in the philosophy of language literature, and, it is fair to say, next to none in epistemology.¹ I think this is a lack that needs supplying for many reasons; most importantly for our purposes here, I think we need to understand the epistemic normativity of constatives in an integrated manner if we are to offer a unified response to the threat of the Shiftiness Intuition in its full generality. After all, we have seen that the Shiftiness Dilemma threatened to generalize beyond assertions, to all constatives. If this is so, a full answer to the Shiftiness Dilemma requires a full epistemic normative account for constatives. This chapter is up to two ambitious tasks: it argues for a generalized knowledge-based account of the epistemic normativity of constative speech acts, and it puts forth the corresponding accounts for, respectively, assertives, predictives, retrodictives, descriptives, ascriptives, informatives, confirmatives, concessives, retractives, assentives, dissentives, disputatives, responsives, suggestives, and suppositives.

7.1 Bach and Harnish’s Taxonomy For the reader that is well versed in speech act theory, it might have already become clear that I follow Bach and Harnish’s (1979, 42) excellent taxonomy of communicative speech acts.² Not a lot will hinge on this particular taxonomy eventually, but I will flag the theoretically loaded issues as they come up. For now, here it is: Communicative Speech Acts Constatives

Directives

Commissives

Acknowledgements

Assertives Predictives Retrodictives Descriptives

Requestives Questions Requirements Prohibitives

Promises Offers

Apologize Condole Congratulate Greet

¹ One notable exception is the speech act of telling. See Chapter 9 for discussion. ² Importantly, though, Bach and Harnish’s constatives correspond (roughly, but closely enough for our purposes) to e.g. Austin’s expositives and Searle’s representatives.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

      Ascriptives Informatives Confirmatives Concessives Retractives Assentives Dissentives Disputatives Responsives Suggestives Suppositives

Permissives Advisories

89

Thank Bid Accept Reject

For our purposes here, we will only be concerned with the taxonomy for constatives. Bach and Harnish individuate speech acts by the attitudes they express: according to them, a speech act is a constative just in case it expresses a belief and an intention that the hearer form or continue to hold the like belief. Here is the view in full: Constatives: expressions of a belief, together with the expression of an intention that the hearer form (or continue to hold) a like belief. (1) Assertives (affirm, allege, assert, aver, avow, claim, declare, deny (assert . . . not), indicate, maintain, propound, say, state, submit): In uttering e, S asserts that p if S expresses: i. the belief that p, and ii. the intention that H believe that p. (2) Predictives (forecast, predict, prophesy): In uttering e, S predicts that p if S expresses: i. the belief that it will be the case that p, and ii. the intention that H believe that it will be the case that p. (3) Retrodictives (recount, report): In uttering e, S retrodicts that p if S expresses: i. the belief that it was the case that p, and ii. the intention that H believe that it was the case that p. (4) Descriptives (appraise, assess, call, categorize, characterize, classify, date, describe, diagnose, evaluate, grade, identify, portray, rank): In uttering e, S describes 0 as F if S expresses: i. the belief that 0 is F, and ii. the intention that H believe that 0 is F. (5) Ascriptives (ascribe, attribute, predicate): In uttering e, S ascribes F to 0 if S expresses: i. the belief that F applies to 0, and ii. the intention that H believe that F applies to 0.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

90      (6) Informatives (advise, announce, apprise, disclose, inform, insist, notify, point out, report, reveal, tell, testify): In uttering e, S informs H that p if S expresses: i. the belief that p, and ii. the intention that H form the belief that p. (7) Confirmatives: (appraise, assess, bear witness, certify, conclude, confirm, corroborate, diagnose, find, judge, substantiate, testify, validate, verify, vouch for): In uttering e, S confirms (the claim) that p if S expresses: i. the belief that p, based on some truth-seeking procedure, and ii. the intention that H believe that p because S has support for p. (8) Concessives: (acknowledge, admit, agree, allow, assent, concede, concur, confess, grant, own): In uttering e, S concedes that p if S expresses: i. the belief that p, contrary to what he would like to believe or contrary to what he previously believed or avowed, and ii. the intention that H believe that p. (9) Retractives: (abjure, correct, deny, disavow, disclaim, disown, recant, renounce, repudiate, retract, take back, withdraw): In uttering e, S retracts the claim that p if S expresses: i. that he no longer believes that p, contrary to what he previously indicated he believed, and ii. the intention that H not believe that p. (10) Assentives (accept, agree, assent, concur): In uttering e, S assents to the claim that p if S expresses: i. the belief that p, as claimed by H (or as otherwise under discussion), and ii. the intention (perhaps already fulfilled) that H believe that p. (11) Dissentives (differ, disagree, dissent, reject): In uttering e, S dissents from the claim that p if S expresses: i. the disbelief that p, contrary to what was claimed by H (or was otherwise under discussion), and ii. the intention that H disbelieve that p. (12) Disputatives (demur, dispute, object, protest, question): In uttering e, S disputes the claim that p if S expresses: i. the belief that there is reason not to believe that p, contrary to what was claimed by H (or was otherwise under discussion), and ii. the intention that H believe that there is reason not to believe that p. (13) Responsives (answer, reply, respond, retort): In uttering e, S responds that p if S expresses: i. the belief that p, which H has enquired about, and ii. the intention that H believe that p.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

     

91

(14) Suggestives (conjecture, hypothesize, speculate, suggest):³ In uttering e, S suggests that p if S expresses: i. the belief that there is reason, but not sufficient reason, to believe that p, and ii. the intention that H believe that there is reason, but not sufficient reason, to believe that p. (15) Suppositives (assume, hypothesize, postulate, stipulate, suppose, theorize): In uttering e, S supposes that p if S expresses: i. the belief that it is worth considering the consequences of p, and ii. the intention that H believe that it is worth considering the consequences of p.

7.2 Normative Transfer Now, it is crucial to note that, on the Bach and Harnish taxonomy, all constatives are actually species of assertion. Several of them are species of assertion in a straightforward way that is not particularly interesting, nor should it be a matter of much controversy: predictives, for instance, are assertions about the future, while retrodictives are assertions about the past. Informatives are a particular species of assertion that is characterized by a narrower intention—that of bringing about the belief in question in the hearer (as opposed to the disjunctive intention of either bringing it about or maintaining it, in the case of assertion). Similarly, ascriptives and descriptives are species of assertion that are concerned with property attribution. And so on, from (2) to (13). Bach and Harnish themselves point out (1979, 46) that most of the specialized types of constatives satisfy the definition of assertives. This type then stands out as a higher category. The harder cases to see are those of less committing constatives, such as suggestives (14). On the Bach and Harnish model, suggestives are expressions of i. the belief that there is reason, but not sufficient reason, ³ Bach and Harnish include guessing in the category of suggestives. For the purposes of this book, I would like to stay neutral on this front. Here is why: if they are also right that suggestives express beliefs that there is reason but not sufficient reason to believe that p, and intentions to get the hearer to believe this as well, it would seem that small children cannot guess due to not being sophisticated enough for second-order thought. But this seems prima facie counterintuitive, and there are empirical studies that show that children have the concept of guessing from very early ages (e.g. Miscione et al. 1978). Many thanks to an anonymous referee for pointing this out.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

92      to believe that p, and ii. the intention that H believe that there is reason, but not sufficient reason, to believe that p. Interestingly, though, if the Bach and Harnish account is correct, suggestives will also be a species of assertion. To see this, let us dub q the proposition ‘There is reason, but not sufficient reason, to believe that p.’ The above account of suggestives then becomes: Suggestives are expressions of (i) the belief that q, and (ii) the intention that H believe that q. As such, suggestives with content p collapse into assertions with content q. In this, suggestives under Bach and Harnish’s taxonomy are, indeed, a species of assertion: assertions about there being reason, but not sufficient reason, to believe that p. The next chapter (Chapter 8) will take up the case of one important suggestive—the speech act of conjecture—in detail. For this reason, I will henceforth bracket worries related to the case of suggestives for the purposes of this chapter. If the Back and Harnish taxonomy is correct, then, we can easily spell out all constatives as species of assertion. Here it goes: (2)* Predictives:

Assertions that q: ‘It will be the case that p.’

(3)* Retrodictives: (4)* Descriptives:

Assertions that q: ‘It was the case that p.’ Assertions that q: ‘0 is F.’

(5)* Ascriptives: Assertions that q: ‘F applies to 0.’ (6)* Informatives: Assertions that p expressing the intention that H form the belief that p. (7)* Confirmatives: procedure.

Assertions that p based on some truth-seeking

(8)* Concessives: Assertions that p, where p is contrary to what he would like to believe or contrary to what he previously believed or avowed. (9)* Retractives: Assertions that non-p, contrary to what S previously indicated he believed. (10)* Assentives: Assertions by S that p, as claimed by H (or as otherwise under discussion). (11)* Dissentives: Assertions that non-p, contrary to what was claimed by H (or was otherwise under discussion).

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

     

93

(12)* Disputatives: Assertions that q: ‘There is reason not to believe that p’, contrary to what was claimed by H (or was otherwise under discussion). (13)* Responsives: Assertions that p, which H has enquired about. (14)* Suggestives: Assertions that q: ‘There is reason, but not sufficient reason, to believe that p.’ (15)* Suppositives: Assertions that q: ‘It is worth considering the consequences of p.’ Let’s take stock: we have seen that the etiological function of assertion of generating knowledge in hearers gives strong support to a biconditional knowledge norm of assertion. We have also seen that, on the Bach and Harnish taxonomy of communicative speech acts, all constatives are species of assertion. If that is so, we should expect some variety of normative transfer from assertion to all constatives. Analytically, species enjoy all properties of the corresponding type (plus further properties). What this means for normative correspondence between type and species is that the species will be governed by a norm that is at least as strong as that governing the type (e.g. Simion 2018b). To see this, think, for instance, about dancing: if there is a norm requiring one to wear light shoes for dancing, one must, thereby, also wear light shoes for waltzing, in virtue of the latter being a species of dancing. That being said, normative transfer need not imply normative commonality (Simion 2018b). To see why, note that the relevant norm for the kind of shoes required for waltzing may be more demanding than the one governing dancing simpiciter. After all, what distinguishes waltzing from just any dancing is the fact that it is governed by some distinctive norms on top of the ones governing the type it belongs to. Thus, it can be that the relevant norm for waltzing is stronger than the one for mere dancing; and, as a matter of fact, waltz shoes are not just any light shoes, but medium heel, pointy, cross-strapped light shoes. What follows from all this is that KNA, together with the Bach and Harnish taxonomy of constatives, delivers the necessity, but not the sufficiency, direction of a knowledge norm for constatives. Knowledge

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

94      will be necessary for the epistemic permissibility of all constatives in virtue of them being species of assertion, but it may not be sufficient in virtue of their respective specific features. Here are, respectively, the knowledge norms for constatives delivered by the framework so far: KN(2): KN-Predictives: One’s predictive with content p is epistemically permissible only if one knows that it will be the case that p. KN(3): KN-Retrodictives: One’s retrodictive with content p is epistemically permissible only if one knows that it was the case that p. KN(4): KN-Descriptives: One’s descriptive with content 0 is F is epistemically permissible only if one knows that 0 is F. KN(5): KN-Ascriptives: One’s ascriptive with content F applies to 0 is epistemically permissible only if one knows that F applies to 0. KN(6): KN-Informatives: One’s informative with content p is epistemically permissible only if one knows that p. KN(7): KN-Confirmatives: One’s confirmative with content p is epistemically permissible only if one knows that p based on some truthseeking procedure. KN(8): KN-Concessives: One’s concessive with content p is epistemically permissible only if one knows that p, where p is contrary to what he would like to believe or contrary to what he previously believed or avowed. KN(9): KN-Retractives: One’s retractive with content p is epistemically permissible only if one knows that non-p, contrary to what one previously indicated he believed. KN(10): KN-Assentives: One’s assentive with content p is epistemically permissible only if one knows that p, as claimed by H (or as otherwise under discussion). KN(11): KN-Dissentives: One’s dissentive with content p is epistemically permissible only if one knows that non-p, contrary to what was claimed by H (or was otherwise under discussion). KN(12): KN-Disputatives: One’s disputative with content p is epistemically permissible only if one knows that there is reason not to

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

     

95

believe that p, contrary to what was claimed by H (or was otherwise under discussion). KN(13): KN-Responsives: One’s responsive with content p is epistemically permissible only if one knows that p, which H has enquired about. KN(14): KN-Suggestives: One’s suggestive with content p is epistemically permissible only if one knows that there is reason, but not sufficient reason, to believe that p. KN(15): KN-Suppositives: One’s suppositive with content p is epistemically permissible only if one knows that it is worth considering the consequences of p. What we have now is the first proposal in the literature for an integrated view of the epistemic normativity of constative speech acts. To be clear, one can come and dispute the particulars of the above knowledge norms: that’s fine. We need to start somewhere, though. If we are to make progress towards the correct integrated account of constatives, having one integrated account like the one I developed above, which relies on one of the best available taxonomies of communicative speech acts on the market, is a first crucial step in this direction.

7.3 Two More Arguments for Knowledge Norms for Constatives The derivation above relies on Bach and Harnish’s taxonomy of communicative speech acts; that’s not particularly theoretically loaded: Bach and Harnish’s taxonomy is roughly translatable to e.g. Austinian and Searlian taxonomies as well. The derivation also relies, however, on Bach and Harnish’s preferred way to individuate speech acts, by the attitudes that they take these speech acts to express. Since some readers might not endorse the Bach and Harnish way to do things in speech act theory—in particular, their individuation recipe—this section will provide two arguments for an integrated knowledge account of constatives that do not rely on their

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

96      framework. In particular, I expect normativists about the nature of speech acts—who take speech acts to be constituted by norms in the same way in which games and languages are (e.g. Searle 1969, Williamson 2000)—will more likely to be convinced by the second argument below than by the derivation based on Bach and Harnish’s taxonomy. The first argument is pretty straightforward: it relies on the independently plausible claim that constatives are all species of assertion, however we choose to individuate speech acts. They are e.g. assertions about the past, assertions about the future, assertions about something having particular properties, assertions about some proposition being false against what the hearer claims, and so on. I find the type–species claim to be eminently plausible no matter one’s theoretical commitments with regard to the nature of speech acts for cases (2) to (13) and (15), and likely more controversial for (14), which I’m bracketing for now (but see Chapter 8). If that is the case, though, the corresponding knowledge norms will follow directly from the type–species claim as shown above, independently of the Bach and Harnish speech act individuation recipe. Second, we can derive the integrated knowledge account for constatives independently of both the Bach and Harnish individuation model and the type–species claim, on functionalist grounds. Here is a claim that I find eminently plausible: Characteristically, constatives aim to generate beliefs in the hearer (henceforth CGB for short). CGB is a weak and, I submit, extremely plausible empirical claim: it is not an analysis of constatives and has no individuation ambitions. Now, here is a popular thesis about the epistemic normativity of belief: Knowledge Norm of Belief (KNB): one knows that p.⁴

A belief that p is a good belief iff

Crucially, KNB is an evaluative, not a prescriptive, norm of belief: it is an ought-to-be, not an ought-to-do. It gives correctness conditions for ⁴ Williamson (2000).

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

     

97

belief. In that, the ‘good’ at stake pertains to the Geach-ian (1956 category of attributive goodness: it describes what it is for a token of a particular type to be a good token of its type. In that, KNB merely claims that beliefs that fall short of knowledge are not good qua beliefs, i.e. as tokens of their type. It does not, for instance, imply any commitments with regard to the nature of justification or the permissibility of belief formation.⁵ If CGB and KNB hold, I submit that the following principle is eminently plausible: Characteristically, constatives aim to generate knowledge in the hearer (henceforth CGK for short).

After all, if the characteristic aim of constatives is to generate beliefs in the audience, it is plausible that the aim is to generate good rather than bad beliefs. Otherwise, it’s a mystery why these speech acts are not discontinued: if hearers were not to be incentivized to take constatives up in virtue of their beneficial effects—knowledgeable beliefs—they would stop doing so, which, in turn, would render speakers less likely to engage in making these speech acts to begin with. The practice would disappear. Here is Ruth Millikan on this issue: Language devices will produce effects that interest speakers often enough to encourage continued replication only if hearers replicate hoped-for cooperative responses often enough. (Millikan 2005, 25)⁶

Note, also, the very high plausibility of the respective incarnations of CGK for other constatives, no matter what one’s view on speech act individuation may be, such as: predictives (characteristically, they aim to generate knowledge about the future), retrodictives (characteristically, they aim to generate knowledge about the past), etc. Again, CGK should ⁵ For a defence of KNB from the normativity of assertion see Williamson (2000). For a defence from the nature of practice of inquiry, see Simion (2019c). For discussion about KNB as an evaluative, but not a prescriptive, norm, and its relation to epistemic justification, see e.g. Simion et al. (2016), Simion (2019c, 2020a). ⁶ See also Sperber (2001) for a very similar point.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

98      be uncontroversial for cases (2) to (13) and (15). I expect, once again, the controversial case for CGK to be that of suggestives (see Chapter 8). If CGK is plausible, though, so is a functionalist view of the normativity of constatives that parallels the view of the normativity of assertion defended in the previous chapter. After all, one way to spell out CGK is in functionalist terms, i.e. where we take talk of characteristic aims to track functions: Constatives’ Knowledge E-Function (CKEF): Constatives have the etiological epistemic function to generate knowledge in hearers. Of course, the content of the knowledge claim will vary from one variety of constative to another: predictives will be in charge with generating knowledge about the future, descriptives with knowledge about properties, and so on. CKEF is, admittedly, stronger than CGK: as we have seen in the previous chapter, for an aim to become an etiological function it needs to be part of a positive feedback loop: it needs be the case that fulfilling said aim is epistemically beneficial to the system and thus contributes to the explanation of why the relevant practice is not discontinued. Note, though, that we can now develop parallel arguments to the one put forth in the previous section for the case of assertion, for all constatives (again, setting suggestives aside). For instance, plausibly, the practice of prediction (in the technical sense above) meets e-function: it has generated knowledge about future events in hearers in the past, which was epistemically beneficial, and which, in turn keeps the practice alive. Similarly, the practice of description has plausibly generated knowledge of something 0 being F in the past, which was epistemically beneficial, and which, in turn, keeps the practice alive. And so on. Last, if CKEF holds, KN(2) to KN(15) follow as straightforwardly as KNA dropped out from the knowledge-generating function of assertion. Furthermore, the biconditional versions of KN(2) to KN(15) also hold, on the same functionalist grounds that vindicate KNA. Constatives will reliably generate knowledge in hearers in normal conditions, i.e. conditions similar to those in which they acquired their function, when functioning normally, that is, when functioning in the way in which

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

     

99

they were functioning when they were selected for their beneficial epistemic effects. When functioning normally, constatives will fulfil the epistemic norm constitutively associated with their epistemic e-function of reliably generating testimonial knowledge; they will work the way they are supposed to work, where the right way of working is partly constituted by reliably delivering the epistemic goods in normal conditions. Thus, an epistemically proper constative will be one that, in normal conditions, generates knowledge in the hearer in the way it did back when it acquired its epistemic function. Since it is plausible that constatives fulfilled said function via being knowledgeable, the integrated knowledge account drops right out of constatives’ e-function. What we get is a derivation of knowledge norms for constatives that does not rely on individuating speech acts by the attitudes they express. Where does this leave us with regard to the generalized Shiftiness Dilemma? On the one hand, we have seen that the first derivation above (relying on Bach and Harnish’s 1979 taxonomy) only afforded us the necessity direction of KNA for all constatives. Since the Shiftiness Dilemma requires the biconditional version to get off the ground, strictly speaking, the account defended above does not expose Classical Invariantism to the practical stakes threat. That being said, (1) the functionalist derivation that we have looked at does deliver the corresponding biconditionals, and (2) I find the biconditionals quite appealing. For someone like me, then, the Shiftiness Dilemma matters. Fortunately, the arguments developed in the previous two chapters against the Shiftiness Dilemma and in favour of a conjunction of Classical Invariantism and assertion functionalism will apply, mutatis mutandis, to the case of all constatives insofar as one likes biconditional knowledge norms for them. Classical Invariantism combines snugly with constative functionalism. To see this, let’s go back to the example involving the speech act of reporting: plausibly, permissible reporting requires more epistemic support in high-stakes than in low-stakes contexts. This requirement, however, is sourced in its practical function of serving the practical interests of the audience, not in its epistemic function. As such, reporting from knowledge (as the biconditional version of (KN(6)) requires) is epistemically appropriate independently of context; compatibly with this,

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

100      reporting from knowledge may well be all-things-considered inappropriate in virtue of practical overriding at high-stakes contexts.

Conclusion This chapter has defended the first integrated account of the epistemic normativity of constatives. I have argued for a knowledge account from three different angles: (1) the Bach and Harnish account of the nature of communicative speech acts, (2) the plausible type–species relation between assertion and other constatives, and (3) the normativity of belief in conjunction with constatives’ epistemic function. The account proposed in this chapter merely sketches knowledge views of the normativity of constatives, without defending each of them in detail: it is aimed as a start of a conversation rather than a final say on the issue. We need to start somewhere. That being said, in the next chapters, I will look into the epistemic normativity of three constatives: conjecture, telling, and moral assertion—in more detail; in particular, I will examine and account for the contextual variation of their propriety.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

PART III

C O N S T A T I V E S I N CO N T E X T

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

8 The Epistemic Normativity of Conjecture The first two parts of this book were concerned with arguing that the Shiftiness Dilemma, if correct, is a major threat for the epistemic normativity of constative speech in general, and then with defending a knowledge account against this threat. The previous chapter, in particular, was up to an ambitious task: it developed the first integrated account of the epistemic normativity of constatives: a non-shifty, knowledge norm for constative speech. In the next three chapters, I look more closely into the epistemic normativity of three constatives in particular—conjecture, telling, and moral assertion—and I examine and account for the contextual variation of their propriety. I will start with the speech act of conjecturing. Philosophical orthodoxy has it that conjecture is cheap: not a lot of epistemic work is needed to perform it well. In fact, the epistemic normativity of conjecture is often the favourite contrast case for the normativity of assertion:¹ the latter requires knowledge; conjecture requires less. In a nutshell, I take the literature to date to feature the following universally shared assumption: Conjecture Is Cheap (CIC): demanding than assertion.

Conjecturing is less epistemically

Other than voicing this well-circulated assumption, not a lot of work has been put into figuring out the details when it comes to the epistemic normativity of conjecture. This is very surprising in more than one way.

¹ E.g. Williamson (2000), Turri (2010), Smithies (2012). Shifty Speech and Independent Thought: Epistemic Normativity in Context. Mona Simion, Oxford University Press (2021). © Mona Simion. DOI: 10.1093/oso/9780192895288.003.0008

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

104      First, it is surprising because it cannot be justified by a lack of centrality of this speech act to our linguistic affairs. To the contrary, conjecture is fairly ubiquitous in our linguistic practice. Second, the lack of interest in conjecturing cannot be justified by a lack of centrality of this speech act to our epistemological affairs. Philosophers and scientists alike agree that conjecture plays a crucial tool in our epistemic endeavours. It is, for instance, central to our scientific practice; historically, some philosophers even thought that scientific theory, and human knowledge generally, is irreducibly conjectural (Popper 1963). Furthermore, empirical studies show conclusively that conjecture is the crucial tool in learning and problem-solving.² This chapter has two ambitious aims: (1) it breaks new ground in that it develops a fully fledged account of the epistemic normativity of conjecture; (2) it goes sharply against orthodoxy, in arguing that the epistemic requirements placed on us by conjecture are stronger than those governing assertion.³

8.1 Conjecture Orthodoxy Philosophers think that conjecturing requires fairly minimal epistemic support, and less than outright assertion (CIC). This assumption is well spread and widely voiced both in the literature on epistemic normativity and in speech act theory. How much less epistemic support than assertion? A few things are worth getting out of the way from the start: first, conjecturing is to be distinguished from weaker neighbouring speech acts, such as the speech act of tabling a hypothesis and the speech act of assuming for the sake of the argument. These latter, suppositive speech acts are often put to good use with no epistemic support at all. Indeed, the history of science is peppered with cases of implausible hypothesis tabling, and the history of philosophy was arguably built on assuming just for the sake of the argument—and even for reductio. Conjecture, while sometimes ² E.g. Anthony (1996), Canadas et al. (2007). ³ See Simion (2020c) for a more sophisticated argument in support of this view.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

    

105

confused⁴ with one or the other of these speech acts, is a more highbrow affair. To see this, it is useful to think of cases of explicit conjecturing— i.e. cases in which one conjectures that p via the performative ‘I hereby conjecture that p.’ If I utter e.g. ‘I conjecture that the population of England is more numerous than that of Slovenia’ and you find out that I had no reason at all to believe so—or, even worse, that I had reason to believe it was the other way around—I am properly criticizable for my speech act. This seems to suggest that, even if CIC holds, conjecture requires, at least, some warrant to believe p over not-p. One can endorse CIC on speech act theoretic grounds (henceforth, the Speech Act Theoretic Argument, or SATA for short). That is, one could think that conjecturing expresses a degree of belief that is e.g. higher than .5 but lower than what is required for full belief. What would make this account attractive is the intuitive thought that what essentially distinguishes conjecturing from asserting is what Searle and Vanderveken (1985) dub the ‘degree of strength of the sincerity condition’. According to Searle and Vanderveken, two speech acts might be the same along other dimensions, but express psychological states that differ from one another in the dimension of strength. Requesting and imploring both express desires; however, the latter expresses a stronger desire than the former. If this view is right, it is plausible to conceive of conjecture as expressing a less strong belief than assertion. It is easy to see how one could think that there’s only a short step from a view like this to (something in the vicinity of) the following norm: The Credence Norm of Conjecture (CNC): One’s conjecture that p is epistemically permissible only if one has a warranted credence that p that is higher than .5 but lower than full belief. Since conjecturing expresses credence higher than .5 but lower than full belief, the thought would go, it lies in the felicity conditions of the speech ⁴ Indeed, in many historical writings in philosophy of science conjecturing and hypothesis tabling are used interchangeably. Karl Popper’s Conjectures and Refutations: The Growth of Scientific Knowledge (1963), for instance, takes conjecture to serve the purpose of tabling an idea and then trying to falsify it, with a background expectation that there might be mistakes to be found.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

106      act of conjecturing for the credence in question to actually be present when a conjecture is made; further on, since there is a warrant condition on permissible credence, all in all, something in the vicinity of CNC will follow; or so the thought would go. On the assumption that knowledge is the norm of assertion (KNA), and since CNC asks for less than knowledge for permissible conjecture, CIC is also thereby vindicated. Declan Smithies (2012), for instance, takes this route in arguing that conjecture is cheap: I will explain why the norms for assertion are more demanding than the norms for conjecture by appeal to the claim that the norms for belief are more demanding than the norms for high confidence. (Smithies 2012, 276)

One can quibble with the details, of course: for instance, one could think that CNC is too weak in two ways; first, while conjecture does not require a lot of epistemic work, it does require more than a level of warrant barely surpassing chance (if my evidence makes it 0.5000005 likely that p, I don’t seem to have enough support to conjecture that p; or so the thought would go⁵). So maybe the relevant minimal threshold should be somewhat higher than .5. Second, one could think that CNC is too weak if one likes factive norms for constative speech acts, and thereby thinks that a conjecture that is not true is epistemically defective (Turri 2010). Finally, one could, to the contrary, think CNC is too strong: how about cases in which I conjecture that p as I regard p as the most likely of the various candidates (but still less likely than .5, since there are say ten mutually incompatible options)? Can’t this sort of conjecture be permissible? I will take up this issue below. For now, these details will largely be inconsequential: CNC will do. What is important is that, since CNC asks for less than knowledge for permissible conjecture, CIC is also thereby vindicated. Alternatively, one might like Conjecture Is Cheap on epistemological grounds (henceforth, the epistemological argument, or EA). Although no ⁵ There is room for pushback here from conjectures that are being made at the frontiers of new scientific fields, which plausibly have barely above chance warranted given the evidence.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

    

107

definite proposal concerning the epistemic normativity of conjecturing has been defended in the literature to date, the issue does often come up in passing in the assertion debates. When it does, conjecturing is often used as a contrast case in defence of KNA, as the paradigmatic speech act that requires less than knowledge (i.e. weaker evidential support) for its epistemic propriety (e.g. Williamson 2000, Turri 2010). Here is Tim Williamson, for one: ‘There is, for example, a speech act of conjecturing p, for which the evidential norms are more relaxed than they are for assertion’ (2000, 244). According to John Turri, as well, different speech acts come at different ‘epistemic costs’: guessing is cheap, conjecturing requires some epistemic support, assertion asks for knowledge. They are ranked on a so-called ‘credibility index’ (2010, 85), depending on what degree of credibility is required by the speech act type. For assertion, the standard is knowledge. In contrast, the standard for conjecturing is lower. In the next two sections I will argue that, on closer scrutiny, both SATA and EA fail to support the assumption that Conjecture Is Cheap.

8.2 Against Conjecture Orthodoxy Let us first try to unpack what I have dubbed above the Speech Act Theoretic Argument for CIC: we have seen that the argument crucially relies on the assumption that, essentially, in contrast to assertion, conjecture does not express a full belief, but rather a particular (whether lower or higher, it is largely inconsequential) credence that p that is higher than .5 but falls short of full belief: (1) To conjecture that p is to utter a sentence that expresses a credence that p that is higher than .5 but falls short of full belief. Now, on top of necessary conditions, speech acts are subject to felicity conditions along the following lines: (2) A speech act is felicitous only if it expresses an attitude that the speaker has.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

108      (1) and (2) together give us: (3) One’s conjecture that p is felicitous only if one has a credence that p that is higher than .5 but falls short of full belief. Furthermore, one might think a plausible assumption concerning the normative status of felicity conditions is the following: (4) A speech act is epistemically permissible only if it is felicitous.⁶ In turn, (3) and (4) give us: (5) One’s conjecture that p is epistemically permissible only if one has a credence that p that is higher than .5 but falls short of full belief. Now here is a fairly widely endorsed claim about the epistemic permissibility of belief: (6) One’s belief is epistemically permissible iff it is epistemically warranted (alternatively, justified, etc., nothing hinges on the terminology). In line with orthodoxy about the relationship between epistemic permissibility and epistemic justification/warrant for belief,⁷ then, one could think that the following claim is plausible too: (7) A credence of n that p is epistemically permissible if and only if it is epistemically warranted. Further on, the defender of SATA takes (7) to ground normative inheritance: that is, she takes it to be the case that, if (5) and (7) hold, it follows that:

⁶ Indeed, one might think that this spells out a plausible sincerity epistemic condition on speech acts. More on this below. ⁷ E.g. Williamson (2000).

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

    

109

CNC: One’s conjecture that p is epistemically permissible only if one has a warranted credence that p that is higher than .5 but falls short of full belief. Since CNC in conjunction with KNA imply CIC, CIC is thereby vindicated on speech act theoretic grounds. In sum, the thought behind the SATA line is simple and elegant: since it is plausible that there is a particular credence norm on conjecture, whatever the epistemic norm for the relevant credence is, it will get inherited by conjecture. What I will argue next, however, is that CNC fails to follow from the Speech Act Theoretic Argument on normative grounds. If I am right, CIC will thereby also lose support from SATA. Here is why: again, in brief, according to SATA, (5) One’s conjecture that p is epistemically permissible only if one has a credence that p that is higher than .5 but falls short of full belief. (7) A credence of n that p is epistemically permissible if and only if it is epistemically warranted. (CNC) Therefore, one’s conjecture that p is epistemically permissible only if one has a warranted credence that p that is higher than .5 but falls short of full belief. (from 5 and 7) Unfortunately, when thoroughly unpacked, it becomes clear that the argument does not go through: since the first premise above (5) features mere credence, while the second (7) stipulates necessary and sufficient conditions for epistemically permissible credence, and since credence does not imply epistemically permissible credence, the SATA line fails: deontic ‘possible’ does not work like this.⁸ There will be worlds where you have an epistemically impermissible credence, but you still do have ⁸ I take the broad philosophical point to be enough here. Here is, though, the same point, but in more detail: on a narrow-scope interpretation of the deontic claims in the argument, the conclusion fails to follow. On a wide-scope interpretation, standard deontic logic vindicates the conclusion; however, this is widely taken to be a problem rather than a feature of standard deontic logic; very roughly, one core problem (but hardly the only one) for deontic inheritance is that ‘ought’ is context-sensitive. For discussion, see, e.g. McNamara (2019).

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

110      it, therefore you can epistemically permissibly conjecture. Permissibility does not transmit like the SATA line needs it to transmit. What the SATA champion needs for the argument to go through, then, is to modify (5) so as to feature epistemically permissible credence throughout: SATA* (5)* One’s conjecture that p is epistemically permissible only if one has an epistemically permissible credence that p that is higher than .5 but falls short of full belief. (7) A credence of n that p is epistemically permissible if and only if it is epistemically warranted. (CNC) Therefore, one’s conjecture that p is epistemically permissible only if one has a warranted credence that p that is higher than .5 but falls short of full belief. (from 5* and 7) This refurbished version of the SATA argument is, indeed, valid; however, it is not clear what in the SATA general story can be taken to offer support to (5*). Recall that (5) was sourced in a view of the nature of speech acts that individuates them by the attitudes they express. Conjecture, on this view, was individuated by expressing a credence of more than .5 but less than full belief. Furthermore, felicity conditions for speech acts ask for the conditions expressed to be instantiated: one should not express a credence one does not have. There is no reason, however, to impose a stronger felicity condition on conjecturing, i.e. asking for epistemically permissible credence to be in place (as per (5*)). After all, what is expressed by conjecture is merely credence, not epistemically permissible credence. As such, speech act theory—i.e. accounts of the nature of the speech act of conjecturing—will not have enough normative resources to deliver (5*). Mutatis mutandis, the SATA champion cannot afford the refurbished version of her argument. In sum: we have seen that the Speech Act Theoretic Argument— relying on considerations pertaining to the nature of the speech act of conjecturing—failed to support CNC—and thereby CIC—in virtue of falling short of normative resources.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

    

111

The obvious alternative is to turn to a normative domain for the needed resources: maybe CIC will follow on epistemological grounds. We have seen that, although not a whole lot of ink has been spilled in epistemology on discussing the normativity of conjecture, one thought that has been circulated a lot is that conjecture is a good contrast case for assertion, normatively: while the latter requires knowledge, conjecture is a weaker speech act—roughly, requiring less support to for the proposition expressed than the corresponding assertion. Recall, for instance, Turri’s thought that competent speakers can rank constatives along a spectrum of credibility requirements, which he dubs a credibility index. Guessing ranks lower than conjecturing, conjecturing ranks lower than asserting. Since less credibility is needed for conjecturing that p than for asserting that p, the thought would go, and since knowledge is the norm of assertion, the epistemic norm of conjecturing should feature a weaker epistemic standing than knowledge. Here is the argument unpacked, on a first approximation: The Epistemic Argument (EA) (1) If a speech act A is lower on the credibility index than another speech act B, then A is less epistemically demanding than B. (2) Conjecturing is lower on the credibility index than assertion. (3) (CIC) Therefore, conjecturing is less epistemically demanding than assertion. The problem with unpacking EA as above is that premise (1) is false: just because a speech act A is lower on the credibility index than speech act B, it does not follow that more epistemic work is required for B than for A. In particular, the claim only holds when indexed in a quantitative way to a particular proposition. To see this, let’s unpack this idea of credibility index further. Turri does not get into much detail about how this index is constituted, but what he does say is that it is intuitive that one needs to put less of one’s credibility on the line for conjecturing than for asserting, and that this is what determines a particular speech act’s position on the credibility index. Here is Turri:

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

112      Asserting that Q places more of your credibility on the line than conjecturing that Q. Swearing that Q places more of your credibility on the line than merely asserting that Q. Let’s call the amount of credibility you place on the line by making an alethic speech act its credibility requirement . . . Competent speakers can intuitively sort credibility requirements along a spectrum. Call the resulting ordering the credibility index for alethic speech acts. The lesser a speech act’s credibility requirement, the lower it ranks on the index; the greater the requirement, the higher it ranks. (Turri 2010, 84–5)

Leaving aside complicated issues pertaining to the nature of credibility at stake, and what it might take to place it on the line, it becomes clear that what Turri has in mind with this credibility index is a quantitative spectrum of epistemic warrant for p: the stronger the speech act, i.e. the higher it is on the spectrum, the more epistemic warrant for p is needed. This seems like the most plausible explanation of why one ‘places more of one’s credibility on the line’ in making a speech act that is higher on the credibility spectrum (on the assumption, of course, that the credibility of the source at least partakes in warranting p in testimonial exchanges). I agree that more warrant for p is needed for asserting that p than for conjecturing that p. That being said, I don’t think that this picture is enough to give support for CIC. To see this, here is the correspondingly refurbished EA: EA* (1*) If a speech act A is lower on the credibility index than another speech act B, then less warrant for p is required for performing A on p than for performing B on p. (2*) Conjecturing is lower on the credibility index than assertion. (3*) Therefore, less epistemic warrant for p is required for conjecturing that p than for asserting that p. EA* is a good argument. Unfortunately for the champion of CIC, it fails to establish it; (3*) does not imply CIC: the fact that less epistemic warrant for p is required for conjecturing that p than for asserting that

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

    

113

p does not imply that conjecturing is less epistemically demanding than assertion. It merely implies a restricted CIC*: CIC*: Conjecturing that p is less epistemically demanding with regard to the quantity of warrant for p than assertion. Compatibly with asserting requiring more warrant for p (so quantitatively superior epistemic support), conjecture may still be more epistemically demanding in qualitative terms: that is, it might be that the kind of support that is needed for permissible conjecture is harder to achieve. Here is Lackey (2019, 627) on this important distinction when it comes to the normativity of assertion: ‘the norm of assertion focuses only on the quantity of epistemic support for proper epistemic assertion, such as knowledge, justified belief, certainty, and so on. But the assertoric quality of the support, such as whether the belief in question has firsthand or secondhand grounding, is just as important.’ Importantly, also, the quality of support is not restricted to the type of source as in Lackey’s example: it can, e.g., concern accessibility conditions on warrant, higheror lower-order support, discursive justifiability, etc. Compatibly, then, conjecture might well be more demanding epistemically, in qualitative terms, not narrowly indexed to the quantity of warrant required for p. To take but a few examples, for instance, it might be that the norm of conjecture features an accessibility condition, in a way in which assertion does not; or it might be that one needs to also be able to discursively justify conjectures, but not assertions; or, indeed, to follow Lackey’s example, it might be that second-hand support is not good enough for conjecture. And so on.

8.3 The Knowledge Norm of Conjecture We have seen that Bach and Harnish (1979) individuate speech acts by the attitudes they express: according to them, a speech act is a suggestive (conjecture, hypothesize, speculate, suggest) just in case n uttering e, S suggests that p if S expresses: i. the belief that there is reason, but not

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

114      sufficient reason, to believe that p, and ii. the intention that H believe that there is reason, but not sufficient reason, to believe that p. We have also seen that, on this taxonomy, suggestives are a species of assertives. Let us dub q the proposition ‘There is reason, but not sufficient reason, to believe that p.’ The above account of suggestives then becomes: suggestives are expressions of (i) the belief that q, and (ii) the intention that H believe that q. As such, suggestives with content p collapse into assertions with content q. In this, suggestives under Bach and Harnish’s account of the nature of constatives are, indeed, a species of assertion: assertions about there being reason, but not sufficient reason, to believe that p. Here it goes: Suggestives: Assertions that q: ‘There is reason, but not sufficient reason, to believe that p.’ If suggestives are a species of assertion, in turn, we should expect normative transfer from assertion to suggestives: species will be governed by a norm that is at least as strong as that governing the type. Assume now that a biconditional knowledge norm is the right norm for assertion (KNA). What follows is that KNA, together with the Bach and Harnish taxonomy of constatives, delivers the necessity, but not the sufficiency, direction of a knowledge norm for suggestives. Knowledge will be necessary for the epistemic permissibility of suggestives in virtue of them being species of assertion, but it may not be sufficient in virtue of their respective specific features. Correspondingly, we get, on a first approximation, the following knowledge norm of conjecture: The Knowledge Norm of Conjecture#1 (KNC#1): One’s conjecture that p is epistemically permissible only if one knows that there is reason, but not sufficient reason, to believe that p. On the view above, the norm for conjecturing asks for knowledge of there being reason but not sufficient reason to believe that p. Let us now try to further refine the view. The first important question will be: what does ‘reason but not sufficient reason to believe’ stand for? How much epistemic support?

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

    

115

I take the answer to this question not to be straightforward at all, and to depend on the variety of suggestive we are talking about. Recall that, according to Bach and Harnish, the genus ‘suggestives’ includes conjecturing, hypothesizing, speculating, and suggesting. While it is far beyond the ambition of this paper to diagnose the amount of epistemic support required for every one of these speech acts, I do want to say, in line with Turri (2010), that I would expect them to lie on a credibility spectrum, whereby ‘reason but not sufficient reason to believe’ requires more warrant in the case of conjecture than in the case of speculating or hypothesizing (more on this below). To put some flesh on the bones of this, I want to suggest that the warrant at stake for the normativity of conjecture should be enough for credence that p that is higher than .5 but lower than full belief. In a nutshell, one should have warrant to believe that p rather than non-p, but not enough for full, knowledge-level belief that p. Also, while I am not in the overly ambitious business to provide a precise warrant threshold here, I take it, in line with the earlier discussion, that it is intuitive that the .5 threshold should be somewhat substantively surpassed—warrant for a 0.500000001 credence will likely not be enough support for conjecture (although I don’t take this to be a significant complication: our warrants don’t often come as finely grained as this). For the sake of simplicity, I will use ‘warrant to believe that p’ henceforth to stand for the amount of warrant thus specified: warrant that somewhat significantly supports p over non-p. Let’s have a second shot at stating the view: The Knowledge Norm of Conjecture#2 (KNC#2): One’s conjecture that p is epistemically permissible only if one knows that there is warrant, but not sufficient warrant, to believe that p. One problem with KNC#2 is that it is too strong:⁹ as stated, it asks me, the speaker, to know that there is no sufficient warrant to believe that p. Of course, though, since knowledge is factive, that is a lot to ask: consider a case in which a colleague scientist on the other side of the world has just discovered more (indeed, enough for full belief) evidence that warrants p. ⁹ Thanks so much to Adam Carter for pressing me on this.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

116      In that case, unbeknownst to me, it is false that there is not enough warrant to believe that p, therefore I cannot know it. I am thus in breach of the norm if I conjecture that p. Intuitively, this is the wrong result: what seems to matter for permissible conjecture is one’s knowledge of warrant one has. Here it goes: The Knowledge Norm of Conjecture (KNC): One’s conjecture that p is epistemically permissible only if one knows that one has warrant, but not sufficient warrant, to believe that p. KNC is the formulation I want to stick with. I will, of course, stay theoretically neutral on how one must conceive of the relevant notion of warrant; just feel free to plug in your favourite view. Similarly, what the ‘having’ relation consists in is a hotly debated notion in the literature, and it falls way beyond the scope of this paper to dive into this debate. I will say one thing that does matter for the view, though: I take the warrant at stake to be propositional rather than doxastic: one need not have any belief or credence with p-content in order to permissibly conjecture that p. This seems right: plausibly, the main function of conjecture in our epistemic life is to make progress in inquiry. It is plausible, however, that while inquiry is still ongoing, it is permissible for the subject to withhold on p altogether. KNC neatly accommodates this intuition. It is important to note that KNC implies the denial of the Conjecture Is Cheap orthodoxy: if KNC is correct, conjecturing that p is more epistemically demanding than asserting that p in qualitative terms, since KNC requires knowledge of one’s warrant for p, therefore (1) access to one’s warrant and (2) a second-order doxastic attitude. Contra orthodoxy, conjecture, in contrast to assertion, is a sophisticated affair. Finally: the derivation above relies on Bach and Harnish’s taxonomy of communicative speech acts. Again, though, we have seen in the previous chapter that one need not buy into the Bach and Harnish speech act individuation model in order to accept knowledge norms for constatives: the attitudes constatives characteristically express, and their epistemic function, will also support them, and thereby KNC.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

    

117

First, I find it plausible that conjectures characteristically express the belief that the conjecturer has warrant, but not sufficient warrant, to believe that p. After all, recall that conjecture is mostly employed in scientific practice: it seems important, for such contexts, to have a speech act that communicates our epistemic circumstances in this way. If that is so, again, characteristically, conjectures will amount to assertions that q, where q is: ‘there is warrant but not sufficient warrant to believe that p’. If so, in virtue of this type–species claim, in conjunction with KNA, and via normative correspondence, we get the result that, characteristically, one must know that one has warrant but not sufficient warrant to believe that p in order to permissibly conjecture that p. This latter claim does not entail KNC, of course. But it does make it eminently plausible via inference to the best explanation: the best explanation of why characteristically one must know that one has warrant but not sufficient warrant to believe that p in order to permissibly conjecture that p is that KNC is the norm of conjecture. Second, we can derive the knowledge account for conjecture independently of both the Bach and Harnish speech act individuation model, and the type–species claim above, on functionalist grounds. Recall the functionalist claim I argued for in the previous chapter: Constatives’ Knowledge Function (CKF): Constatives have the epistemic function of generating knowledge in hearers. Here is the corresponding claim for conjecture: Knowledge Function of Conjecture (KFC): Conjecture has the epistemic function of generating knowledge of the speaker having warrant, but not sufficient warrant, to believe that p in hearers. To see the plausibility of this claim, recall that according to the etiological theory of functions, then, a token of type T has the e-function of type B of producing effect E in system S iff (1) tokens of T produced E in the past, (2) producing E resulted in benefit of type B in S/S’s ancestors, and (3) producing E’s having B-benefited S’s ancestors contributes to the explanation of why T exists in S.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

118      Note that KFC meets these conditions for conjecture with flying colours: plausibly, the practice of conjecture has generated knowledge of the speaker having warrant, but not sufficient warrant, to believe that p in hearers in the past (indeed, this phenomenon is plausibly ubiquitous in mainstream intra-science communication), which was epistemically beneficial for hearers, and which, in turn, plausibly keeps the practice alive. In turn, if KFC holds, KNC nicely drops out of it. Conjectures will reliably generate knowledge about the speaker’s warrant in hearers in normal conditions, i.e. conditions similar to those in which they acquired their function, when functioning normally, that is, when functioning in the way in which they were functioning when they were selected for their beneficial epistemic effects. When functioning normally, conjectures will meet the epistemic norm constitutively associated with their epistemic efunction of reliably generating testimonial knowledge of the speaker having warrant but not sufficient warrant to believe that p; they will work the way they are supposed to work, where the right way of working is partly constituted by reliably delivering these epistemic goods in normal conditions. Thus, an epistemically proper conjecture will be one that, in normal conditions, generates knowledge of the speaker having warrant but not sufficient warrant to believe that p in the hearer in the way it did back when it acquired its epistemic function. Since it is plausible that conjecture fulfilled said function via being knowledgeable— after all, the paradigmatic¹⁰ way in which knowledge is generated in hearers in testimonial exchanges is via transfer from speakers—the knowledge account drops right out of conjecture’s etiological function. As such, what we get is a derivation of KNC that, once again, does not rely on individuating speech acts by the attitudes they express.

8.4 The Knowledge Norm vs. the Credence Norm Before I close, I would like to go back to what I take to be the main competition for my view—albeit merely assumed rather than explicitly ¹⁰ For cases of knowledge generation on the hearer’s side in spite of lack of knowledge on the speaker’s side see e.g. Lackey (2008).

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

    

119

developed or defended in the literature—i.e. the warranted credence view discussed in section 8.2, or CNC. Recall the view: The Credence Norm of Conjecture (CNC): One’s conjecture that p is epistemically permissible only if one has a warranted credence that p that is higher than .5 but lower than full belief. In what follows, I will run through a number of points on which I take KNC to be superior to its competition.

8.4.1 CNC Is Too Strong One first advantage of KNC over the warranted credence view is that it is weaker, in that it does not require the speaker to entertain any degree of belief with p content. This seems right: plausibly, one of the main roles (although, of course, not the only one—we conjecture in everyday life all the time as well) of conjecture in our epistemic life is to make progress in scientific inquiry. It is plausible, however, that while inquiry is still ongoing, it is permissible for the subject to withhold on p altogether. KNC neatly accommodates this intuition. I contrast, CNC is too strong, in that it requires the conjecturer to have a credence with p-content.

8.4.2 CNC Is Too Weak In a different way, KNC is, of course, stronger than the warranted credence view, in that it requires epistemic access to one’s warrant for p. Again, this seems right: conjecturing, as opposed to asserting, is an epistemically highbrow affair, mostly employed in scientific practice. Small children and animals don’t engage in this practice: overintellectualization is a feature of the view, not a bug. To see this, note that if I just flat out assert e.g. that electrons are negatively charged, and you ask me how I know this, it’s fine for me to say something along the lines of: ‘I don’t remember, I must have learnt it in school.’ However, if I explicitly say ‘I conjecture that electrons are negatively charged’, and

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

120      you ask me ‘What makes you say that?’, a parallel reaction seems intuitively impermissible. This seems to suggest that access to one’s reasons for conjecturing, but not for asserting, is epistemically required for permissibly performing the corresponding speech act. Again, conjecture is a highbrow affair, and KNC makes good on this fact. In contrast, CNC is too weak, in that it predicts that one can conjecture based on mere warranted credence, without having access to the warrant makers, and thereby without being in a position to discursively justify one’s conjecture.¹¹

8.4.3 Prior Plausibility Another important thing to note is that neither of the arguments discussed in the beginning in support of CIC (SATA and ET) will serve to support a warranted credence norm over my knowledge norm; the knowledge account proposed here is perfectly compatible with—indeed, it can be read as an incarnation of—the view that conjecturing requires

¹¹ Relatedly, one other (slight) advantage of KNC over CNC is that it is not wedded to predicting that conjectures are permissible if based on mere high probabilities. Consider: (1) I conjecture that my ticket didn’t win the lottery. (2) I conjecture that this woman is good at housekeeping. (3) I conjecture that this person of colour is a member of the club’s staff. To see what is wrong with conjecturing (1), note a few things. First, if conjecturing on lucky events is permissible, it’s not clear what use we still have for the speech act of guessing. Second, note that if (1) is permissible, then the corresponding conjecture about each and every other ticket is also permissible. This is already fairly hard to swallow normatively. Furthermore, this raises the question as to why, then, does it seem clearly impermissible to conjecture that all tickets lost the lottery. Things become even more intuitively problematic once we move on to gender- and raceloaded conjectures on mere probabilistic grounds. Clearly, conjecturing that a woman makes for a good housekeeper and that a person of colour is a member of staff is impermissible no matter how high the probabilities built into the case might be. Furthermore, importantly, the impermissibility at stake here is distinctively epistemic, not just moral. To see this, note that moral blameworthiness features an epistemic condition, viz. that the subject is not epistemically blamelessly ignorant for the norm breach in question. Since it is intuitive that someone uttering the conjectures in question is morally blameworthy for doing so, it seems to follow that they are not epistemically in the clear either. On the assumption that (1) the probabilities involved in the cases above are above .5, and (2) that probabilities warrant credences, CNC will predict these conjectures are permissible. In contrast, KNC need not do so: it will all amount to the way in which the relevant notion of warrant is unpacked. If one is e.g. a Lockean about warrant, KNC will make the same mistaken predictions as CNC. However, there are several views of warrant out there (e.g. Simion 2019a, 2020a) that avoid the prediction that lottery beliefs are warranted. In conjunction with these views, KNC will avoid this problem. The fact that it has the flexibility to do so gives KNC a slight advantage over CNC on this front.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

    

121

less epistemic support than assertion, insofar as the claim is restricted to a quantitative claim about the warrant needed for the particular proposition that constitutes the content of the speech act in question (which is, I’m fairly convinced, what was meant by the people cited in the two sections in question to begin with). After all, on my view, while assertion that p requires knowledge that p, conjecturing that p merely requires knowledge of one having warrant but not sufficient warrant to believe that p. As such, the quantity of epistemic support required for the proposition at stake—p—is much weaker for proper conjecturing than for proper assertion. The view proposed here is also compatible with the claim that conjecture expresses a degree of belief that p that is lower than full belief; even if this is so, it can still be that what is epistemically required for permissibly conjecturing is what KNC requires, i.e. access to the grounds for said degree of belief: knowledge of one having warrant, but not sufficient warrant, to believe that p.¹²

8.4.4 Moorean Paradoxicality KNC also nicely explains several conjecture-specific varieties of Moorean paradoxicality. Consider, for one: Moore#1: I conjecture that this patient has the flu but I have no reason to believe that they have the flu. This sounds bad. KNC explains why: knowing that one has enough warrant for believing that p is required for one’s permissible conjecture. CNC struggles to do so: if all that is needed for permissible conjecture is warranted credence that p, it may well be¹³ that I don’t have access to the ¹² Although, importantly, in isolation, KNC together with this type of view will imply the denial of the claim that felicity conditions are normatively charged: after all, KNC does not require that any credence that p be present for conjecturing that p. Compatibly, though, there could be that there is an additional, moral sincerity norm, for instance, that delivers the normative charge. ¹³ This will depend on your view on warrant. On the vast majority of views in the literature warrant does not entail access to grounds for belief. Strong access externalism denies this claim,

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

122      grounds that warrant my credence, and, furthermore, that I take myself not to have any.¹⁴ Similarly, consider: Moore#2: I conjecture that this patient has the flu but I have no idea why one might think that. Again, KNC, but not CNC, nicely explains the paradoxical soundingness of the above: knowledge of one’s grounds for p is necessary for proper conjecture. Since, on most views on warrant, having a warranted credence is compatible with not having access to the grounds constituting the warrant in question, CNC delivers the result that there is nothing wrong with Moore#2.¹⁵

8.4.5 The Upper Warrant Threshold Consider: Moore#3: I conjecture that this patient has the flu, I have conclusive evidence that he does. Again, this sounds weird: why not just go ahead and assert it outright if you know you have conclusive evidence for the claim? Once again, KNC explains the weird-soundingness: it is necessary for permissible conjecture that the warrant possessed not be enough for outright belief. One

so on this construal the differences between KNC and CNC will shrink considerably. I take strong access internalism to be a relatively unpopular view, however. ¹⁴ People who think there is such a thing as merely psychological defeat might argue that, in taking myself not to have warrant for my belief, my belief is thereby defeated. For arguments why there is no such thing as merely psychological defeat see e.g. Graham and Lyons (Forthcoming) and Simion (2020a). ¹⁵ Couldn’t champions of CNC simply add an access requirement to avoid this problem? Or perhaps an additional norm that permits the audience to request sources that will leave the conjecture not in good standing if sources cannot be provided? They could. The question that remains to be answered by CNC champions is the epistemic condition on access. If it’s knowledge, CNC comes close to collapsing into the competitor KNC. If it’s not knowledge, why not? After all, discursive justification will likely involve a bunch of assertions, which, if KNA is correct, will have to be knowledgeable. These are questions that are for the CNC champion to answer though.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

    

123

might think that CNC does not have much trouble with Moore#3 either: after all, just like KNC, it imposes an upper threshold for permissible conjecture at knowledge level. Recall, though, that, as opposed to KNC, CNC asks for the warranted credence backing the conjecture to be lower than full belief. Short of this, CNC does not concern itself with the amount of propositional warrant present at the case. As such, there will be cases in which one has conclusive reasons to believe that p, although one does not form the relevant outright belief, but rather a lower credence, which will be diagnosed as cases in which it is permissible to conjecture that p.¹⁶ This not only gets Moore#3 wrong, but is a bad result independently of Moorean paradoxicality: we don’t want people to restrict themselves to merely conjecturing that p when they have conclusive reasons for p. One might, though, on second thought, wonder whether this is an advantage for KNC rather than a bug.¹⁷ After all, Moore#3 is easily explained by Grice’s maxim of quantity, whereby one should try to be as informative as one possibly can. Moore#3 flouts this maxim: given that I know that the patient has the flu, I can assert it outright rather than merely conjecture it. If Gricean maxims take care of the issue, the thought would go, the epistemic norm at stake need not. Furthermore, other norms for speech acts that are discussed in the literature do not impose an upper threshold for warrant: take KNA, for instance. It is perfectly fine, on KNA, to assert from certainty. Why would the norm of conjecture be any different? A few things about this. First, the relation between epistemic norms of speech acts and Gricean maxims is not as straightforward as one might think: importantly for our purposes here, it is not normatively exclusive: it is not the case that, just because a Gricean maxim takes care of regulating a particular conversational phenomenon, it need not be taken care of via the corresponding epistemic norm. Take Gricean Quality—asking for one’s assertion to be truthful and supported by

¹⁶ Note also that judging ‘I am not fully confident that p but I have conclusive reason that p’ already is intuitively problematic, which suggests that this problem also appears at doxastic level. ¹⁷ Thanks to Chris Kelp for pressing me on this.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

124      evidence—as the paradigmatic example:¹⁸ several people have, in fact, relied on it to argue for one account or another of the normativity of assertion. There is also a good reason why this—i.e. non-normative exclusivity—is so: Grice’s maxims are sourced in a prudential cooperative principle; as such, Grice’s maxims are prudential maxims. Crucially, though, prudential and epistemic norms can coincide: indeed, they do so, very roughly, in cases in which the prudential goods at stake are also epistemic goods, such as knowledge, or information. Since conversations are instances of epistemic cooperation—cooperation concerning information exchange—unsurprisingly, Gricean maxims and epistemic norms will overlap to a significant extent. Second, I want to stress that the maxim of quantity is a paradigmatic such example: after all, the worry with not being as informative as one possibly can is that (1) we lose epistemic goods in the testimonial exchange, and (2) a false belief will be generated in the audience—i.e. a false belief about how much the speaker knows about the issue. In our case, if I utter ‘I conjecture that the patient has the flu’, you’re likely to believe I don’t have enough warrant to outright assert it. Since losing epistemic goods and generating false beliefs are uncontroversially epistemically bad things, we shouldn’t be surprised that quantity-related considerations will feature in conjecture’s epistemic norm. Third, relatedly, note that there is an asymmetry between assertion and conjecture when it comes to quantity-related considerations. To see this, let us follow Turri (2010) and take the speech act that is higher than assertion on the credibility index to be the speech act of guaranteeing, and let’s stipulate, just for the sake of the argument, that it is governed by a certainty norm of the following form: The Certainty Norm of Guaranteeing: One’s guaranteeing that p is permissible only if one is epistemically certain¹⁹ that p. Now, here is the interesting disanalogy between assertion and conjecture: explicitly conjecturing that p when one knows that p is inappropriate in ¹⁸ See Benton (2016) for an excellent discussion. ¹⁹ As distinguished from mere unwarranted psychological certainty.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

    

125

that it intuitively generates the false implicature that one does not know that p. Relatedly, when a speaker does so, they are the proper subject of criticism: ‘Why didn’t you tell me you knew that p?’ In contrast, merely asserting rather than explicitly guaranteeing when one is certain that p seems fine, and does not open the speaker to similar criticisms. If I’m certain that there’s a table in front of me and I assert as much, it’s not common that you would criticize me along the lines of ‘Why didn’t you guarantee that there’s a table in front of you if you were sure there was one?’ I will not attempt to explain the source of this disanalogy here: likely, the rarity of occasions when we enjoy certainty, as well as the rarity of occasions when we need to guarantee that something is the case, will play a role. What is important to note for my purposes here is that the disanalogy exists: conjecturing, but not asserting, generates false quantity implicatures when employed in spite of having enough warrant for making stronger speech acts. Furthermore, it does so unexceptionally—i.e. this effect does not seem to suffer from much contextual variance. If that is so, this explains why there is a disanalogy between the norm of assertion and that of conjecture, in that the latter, but not the former, should feature an upper warrant threshold.

8.4.6 The Epistemic Function of Conjecture I want to propose that KNC, but not CNC, makes sense of the epistemic function²⁰ of conjecture in our practice of inquiry. Recall that we have seen that, on my view, a plausible epistemic function of conjecture is to generate knowledge in the hearer that the speaker has warrant but not sufficient warrant to believe that p. I take this to be uncontroversially a valuable piece of information in e.g. scientific contexts, but not only in those contexts. In turn, we have seen that this function of conjecture vindicates KNC: likely, the way in which conjecture generated this knowledge at the moment of function acquisition was via being

²⁰ Compatibly, conjecture may well serve several other practical functions in scientific practice and beyond, of course. Also compatibly, tabling hypotheses might have an epistemic function pertaining to productivity in inquiry.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

126      knowledgeable. Also, plausibly, this is the most reliable way for it to generate said knowledge: via transmission. Since a knowledge function straightforwardly delivers KNC, the defender of CNC will have to countenance a different function for conjecture to vindicate her view. One obvious candidate would be a corresponding warranted-credence-generating function: the function to generate a warranted credence that is higher than .5 but lower than full belief. The problem with this view, however, is that it is not clear how exactly CNC will deliver on this front either. To see this, say that I utter ‘I conjecture that this patient has he flu’ based on my .65 warranted credence that the patient has the flu. What are you supposed to do (doxastically) with this? What credence should you now form based on my conjecture? You have no idea what evidence I have, and how high a credence it warrants. At best, all you know is that, if my conjecture is epistemically permissible, it’s based on a credence that’s at least slightly higher than .5. As such, that’s the only credence my conjecture seems to warrant you to form: one slightly higher than .5. The trouble with this result is that (1) it does not fit the function we started with; we’re going to have to adjust the function of conjecture to something along the lines of ‘generating a credence that p that’s slightly above .5’; relatedly, on this model, we get very serious epistemic loss in scientific communication: with every conjecture that p, the epistemic support transmitted from speaker to hearer drops close to .5. Since .5 just is chance level, we are not very likely to make significant progress via scientific communication if this view is correct.

8.5 Practical Shiftiness and Scientific Revolutions Consider the Darwin/Wallace evolution theory,²¹ propounding the preposterous idea almost universally rejected at the time that complex structures could arise other than by design and construction by an even more complex mind. Or Einstein’s general relativity, abandoning

²¹ Thanks to Alan Weir for pressing me on this.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

    

127

the idea that Euclidean geometry is correct for spacetime, and denying that gravity exists as a force in the classical sense. One might worry that KNC is too strong to be able to accommodate scientific revolutions: did these people have stronger warrant for their theories than for their denials? It would seem not, given that they were going sharply against orthodoxy. Furthermore, it would seem that something along the lines of having more warrant against the new theory than in favour of it might be definitional when it comes to scientific revolutions. Furthermore, the thought would go, if this is plausible for scientific revolutions, it is plausible, mutatis mutandis, for general scientific practice. A few things about this: first, scientific revolutions will not serve to support KNC over CNC; after all, if we assume that Darwin and Einstein had more warrant against their theories than in favour of them at the time, they did not meet the warranted credence condition on permissible conjecture stipulated by CNC either. Since I take myself to have fairly solid textual evidence for the widespread support of (something in the vicinity of) CNC in the literature, I will not be extremely troubled by this issue. After all, if it is true that Darwin and Einstein had more warrant for not-p than for p, and if we want our norm of conjecture to be able to accommodate this result, it would seem that it should not feature a warrant condition to begin with. The norm would have to be something along the lines of: ‘Anything goes!’ Or maybe, somewhat more restrictedly, ‘Anything goes, insofar as it eventually leads to scientific progress!’ I don’t think this is the correct account of the normativity of conjecture, and many of the reasons in favour of KNC offered above will serve to explain my preference against it. Another, more plausible maybe, option for accommodating these cases, would be to impose a contextually variable credence threshold for conjecture. On this view, the epistemic support needed for permissible conjecture varies with context: in some scientific contexts—the thought would go—anything goes: not much warrant is needed for epistemically permissible conjecture. The natural question then becomes: what sets the contextual threshold for conjecture’s epistemic permissibility? For the reasons discussed throughout this book, practical stakes won’t do the trick: even if it

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

128      turns out that, intuitively, all-things-considered permissible conjecture will vary with practical stakes, epistemically permissible conjecture need not. If we find it intuitively permissible to conjecture the bank will be open on Saturday in the low-stakes case but not in the high-stakes case, our intuition will not plausibly track epistemic permissibility, but rather the plausibly overriding practical permissibility. A view that takes the permissibility at stake to be epistemic will encounter the same problems we have seen in Chapter 2 for its equivalent, context-variant norm of assertion. I do want to say a few things about how KNC can deal with scientific revolutions, however. First, if it is the case that they are properly described as cases of speech against extant epistemic support, I worry that they may not be conjectures to begin with, but rather cases of hypothesis tabling (recall the discussion in section 8.2), which, in turn, may be more plausibly subject to an ‘anything goes’ norm. Tabling hypotheses seems to be the paradigmatic tool for bringing something forth for consideration just to see what follows from it. If I am right, we can explain scientific revolutions as cases of hypothesis tabling that went really well. Furthermore, if I’m right, it would be mysterious why we would have two speech act theoretic tools (tabling hypotheses and conjecturing) for the same purpose, i.e. that of investigating the credentials of a scientific theory that is not supported by any evidence, and none for informing our scientist colleagues that we have some evidence (albeit not enough) for p.²² Second, it is not clear to me that the epistemic interpretation of scientific revolutions as cases of extant warrant against rather than for the conjecture is correct. To see this, note that it’s not true that these are cases where just ‘anything goes’. These people we are talking about did not come forth to conjecture that humans evolved from blueberry porridge, nor that an omnipotent pink elephant rather than gravity ²² That being said, if it turns out that this is indeed the case—that conjecturing and hypothesis tabling are often taken to be one and the same speech act—the reader should consider this chapter to be an exercise in conception engineering, motivated by the conflict between this usage and the data suggesting that conjecturing against the evidence is properly criticizable, as well as by our need for a speech act that communicates that the speaker has some warrant, but not sufficient warrant, to believe that p. Explicating conjecture as distinct from hypothesis tabling in terms of strength of warrant required will achieve this goal.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

    

129

makes stones fall to the ground. Conversely, it also doesn’t seem permissible to make these conjectures in everyday scientific practice. Some degree of plausibility of the theories in question, then, seems required for permissibly putting them forth. I’m hardly a specialist in the history of science, but here is an alternative way to look at the epistemic situation: these people were dissatisfied with the extant theories because they failed to explain a good amount of data. This, in turn, (partially) defeated their justification to believe that extant theories are correct, which, in turn, gave them incentive to look for alternative explanations of the data. It is easy to see that, on this picture, it is not clear that the old theories had much defeating power against the new theory, as the objection from scientific revolutions would have it. Last but not least: if it does turn out, though, that what these people were doing was pretty much tantamount to conjecturing that a pink elephant is making stones fall (epistemically speaking), then yes, I want to say their conjectures were epistemically impermissible, albeit luckily true and practically permissible: after all, not much hinged on the corresponding falsifications. On a different but related note, one could wonder whether both CNC and KNC are not too strong to account for conjecture’s function in inquiry after all. To see this, consider a case in which detective D is investigating a crime, and she has seven suspects: the butler and the six sons of the victim. Based on the evidence D has so far, she has a warranted .4 credence that the butler did it, and warranted .1 credences respectively that each of the sons did it. Surely, the thought would go, the detective is warranted to conjecture that the butler did it under these circumstances. If this is right, it seems to suggest that neither CNC nor KNC successfully account for conjecture’s function in inquiry after all.²³ I agree that detective D is warranted to conjecture that the butler did it in the case above. What I want to propose, however, is that the type of warrant at work here is practical rather than epistemic: D is under a contract to find the murderer, and she thereby is practically obliged (therefore permitted) to conjecture of at least one of the seven suspects ²³ Thanks to Chris Kelp for pressing me on this front.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

130      that they did it in order to move forward with the investigation. In other words, the detective cannot afford to just suspend judgement and go home in this case. To see that it is merely practical rather than epistemic warrant that is at work here, note that what having .4 warrant in favour of p—‘The butler did it’—amounts to is having .6 warrant for not-p. According to D’s evidence, then, it’s more likely that the butler did not do it than that he did it. It is highly implausible that this is an epistemic situation that warrants conjecturing that p—‘The butler did it.’ To see this, consider a case in which you know that a coin is rigged against landing heads such that it is now only .4 likely it will do so. Surely, if anything, in this case, it is permissible to conjecture it will not land heads.

Conclusion This chapter has defended a knowledge norm of conjecturing, according to which one’s conjecture that p is epistemically permissible only if one knows that one has warrant, but not sufficient warrant, to believe that p. I have argued for a knowledge account from three different angles: (1) the Bach and Harnish account of the nature of communicative speech acts, (2) the plausible type–species relation between assertion and other constatives, and (3) the normativity of belief in conjunction with constatives’ epistemic function.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

9 No Special Treatment for the Epistemic Normativity of Telling The previous chapter broke new ground in that it developed the first fully fledged account of the epistemic normativity of conjecture in the literature. This chapter discusses the epistemic normativity of the speech act of telling. Conversational patterns are usually taken to favour the thought that, just as assertion, the speech act of telling is governed by a knowledge norm (e.g. Buckwalter and Turri 2014). First, questions concerning your knowledge whether p is the case typically function as indirect requests to tell whether p is the case. Second, declaring ignorance is a legitimate way to answer questions. This, of course, suggests that knowledge was assumed to begin with. Third, if someone tells you something, it is typically appropriate to ask, ‘How do you know that?’ Fourth, Moorean tellings of the form ‘p but I don’t know that p’ come across as paradoxical. Last but not least, plausibly enough, knowing something seems good enough for (epistemically) appropriately telling it: if I know that there’s a desk in front of me, it looks as though, all else equal, I’m in a fine epistemic position to tell you as much. In Chapter 7 I have also argued from the nature of telling—as informative constative speech act—to its normativity. I have claimed that all constatives—telling included—are species of assertives, and thus afford knowledge norms. Several people in recent literature, however, follow a similar argumentative recipe and argue from the nature of telling to the claim that its normativity is importantly different from that of assertion, in that it is distinctively interpersonal:

Shifty Speech and Independent Thought: Epistemic Normativity in Context. Mona Simion, Oxford University Press (2021). © Mona Simion. DOI: 10.1093/oso/9780192895288.003.0009

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

132      Non-Commonality (NC): different epistemic norms.

Telling and assertion are governed by

Telling, it is argued, but not assertion, is directed towards an audience. This, in turn, impacts its normativity by rendering it audience-oriented as well. There are two broad arguments for NC in the literature: first, several people working on testimonial justification (e.g. Gibbard 1990, Hinchman 2005, Moran 2005, Ridge 2013, Tanesini 2020) argue that we should conceive of telling on the model of commissives, such as promising or offering, or directives, such as demanding, rather than constatives. This, in turn, is thought to render its normativity distinctively interpersonal. Second, Charlie Pelling (2014) ventures to show that, in virtue of the communicative intentions characteristic of the speech act of telling, the epistemic propriety of instances of telling partly depends on what’s at stake for the hearer. If these philosophers are right, of course, the account of the normativity of telling put forth in Part II is under threat. More importantly, though, the motivation behind the whole debate regarding the epistemic norm of assertion can be seriously called into question. After all, there are very few assertions that are not tellings; given their rarity, it is not clear why we should be concerned with the norms governing them to begin with. Since I take it that it would be rather unfortunate if it turned out that all the assertion-focused ink has been spilled in vain, this chapter gives the case for the special normative status of telling the deserved close scrutiny.

9.1 Telling as Directive or Commissive Several people in the literature on testimonial justification think, in line with G. E. M. Anscombe (1979) and J. L. Austin (1946), that you can insult someone by refusing their testimony, and that, furthermore, this insight is to be explained by looking into the nature of the speech act of telling. Although the details, as we are about to see, differ from one account to another, the general line of thought goes as follows: when I tell

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

    

133

you that p, I not only assert that p, I also perform a distinctively interpersonal act; I do something to you, as it were. Some of these philosophers believe that the act performed on the hearer by an instance of telling falls in the Bach and Harnish (1979, 41) category of directives. Allan Gibbard (1990, 172), for instance, thinks that what I do when I tell you that p is tantamount to demanding that you believe that p. In a similar, but somewhat more modest, vein, Mike Ridge (2013, 58) thinks that when I tell you that p I am advising you to believe that p. Others in this literature think that telling amounts to a commissive, rather than a directive. On Richard Moran’s (2005) view, for instance, telling someone that p amounts to promising them that p is the case. Alessandra Tanesini (2020) disagrees: according to her, while it is correct to think telling is a commissive, the right category thereof is ‘offering’. Telling you that p, on this view, is tantamount to offering you a gift: the gift of testimony. Similarly, Ted Hinchman (2005) thinks that telling amounts to offering one an invitation: the invitation to trust the speaker’s testimony. If these people are right, the worry would go, there seems to be trouble for the account defended in this book, on which there is nothing really special about telling, normatively speaking. The first source of trouble is that, if it turns out that telling is not a constative after all—but rather a directive, or a commissive—the type–species argument for normative inheritance from assertion will seem not to go through. As such, we are left without an account of the normativity of telling. Second, one might think, if the nature of telling is as distinctively second-personal as these accounts suggest, maybe a hearer-oriented normative account is more suited for telling. After all, it seems plausible to think that if I perform an action phi on you, the normativity of my phiing will be somewhat influenced by your interests: if I cook dinner for you, for instance, I’d better not cook something that I know you’re allergic to. There are a few things I would like to say in the way of response to these worries. First, I would like to resist this type of argument against my view on descriptive, speech act theoretic grounds. To begin with, note that what is being claimed by these accounts is not that when I tell you

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

134      that p I do not also thereby assert that p. That would be highly implausible. Rather, the thought goes, telling also has an interpersonal dimension on top of its constative dimension: in telling you that p, the thought goes, I not only assert that p, I also thereby promise you that p is the case, or demand that you believe me, etc. If so, crucially, what is actually being claimed about the nature of telling by these views is that telling is a performative: a speech act that performs a further act. The classical example of a performative is that of marrying: granted that I have the relevant authority, when I state ‘I hereby marry you’, I do two things: I describe the world (the world is such that I am marrying you as we speak) and I change it (you are now married). Similarly, on these views of telling, what is going on when I tell you that p is: I make an assertion that p (describing the world) and, thereby, issue a directive (e.g. I thereby demand that you believe me) or a commissive (e.g. I thereby promise you that p). If all of this is correct, though, the argument for my view of the normativity of telling remains unaffected. First, that is because telling remains a species of assertion—albeit an assertion that does further, commissive or directive work—and, as such, the type–species normative transfer holds. Recall, also, that I recognized from the get-go that all my derivation gives me is a knowledge necessity claim for the normativity of constatives other than assertion, in virtue of the fact that the type–species derivation does not have the normative import needed for the sufficiency claim. That being said, I also don’t think that this picture will do much to threaten a knowledge sufficiency claim for the epistemic norm for telling. I am not the first one to notice (for a very thorough treatment, see Goldberg (2020)) that the worry for these views is that, even if they are right about the nature of telling, it is not clear that the normativity thus generated will be epistemic normativity proper. To see this, consider a case in which I lie to you that p is the case. On this view, I will have thereby e.g. broken a promise to you, or made a demand without having the authority to do so, or offered you bad advice. While all of these things are undeniably bad things to do to another, it is not clear that the badness at stake is properly epistemic rather than, say, moral, or social, or prudential. Or at least more needs to be said as to why we should think

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

    

135

it is epistemic. Insofar as we lack such argument, it is unclear that these ways of thinking about telling are likely to deliver a genuine competitor to the biconditional knowledge account.

9.2 Pelling’s Synchronic Different Stakes According to Charlie Pelling too, while a knowledge norm might be governing assertion, things are not equally straightforward when it comes to the speech act of telling. Pelling starts by pointing out that, while all tellings are assertions, not every assertion is an instance of telling. In particular, assertions that are not directed at anybody will not count as tellings. Thus, telling is a particular variety of assertion that is directed towards an audience.¹ Pelling goes on to argue that assertion and telling are governed by different epistemic norms (Non-Commonality, NC). Pelling supports NC from two angles. First, he argues that assertion and telling are characteristically associated with different communicative intentions. Thus, while intentions that are characteristic to assertion are also characteristic to telling (taking responsibility for the truth of p), the reverse does not hold; the intention to address one’s speech act to a hearer/ reader only characterizes telling (2014, 338). Take, for instance, assertions in a secret diary; plausibly enough, these are not directed towards any audience whatsoever, and still count as assertions. Further on, Pelling argues that there is a significant connection between the characteristic intentions of an act and the norms governing it. Think about any pair of speech acts; take assertion and supposition, for instance. The norms regulating the two will not share the same content. Nor are the

¹ Peling also imposes an uptake condition on telling (uptaking as in comprehending rather than believing). One might be rather sceptical regarding the legitimacy of such an uptake condition, though; after all, it looks as though we often say things like: ‘I told you but you would not listen!’; in the light of this, it might be thought that the mere directedness towards an audience is enough for an assertion to count as a telling. On the other hand, it seems pretty clear that if I just whisper something in your direction from three blocks away, independently of what I may have intended, my assertion will hardly count as an instance of telling. Fortunately, nothing here hinges on very strictly individuating the speech act of telling, so I will not rest further on it.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

136      intentions characteristically associated with them of the same nature (2014, 337). Second, Pelling argues that KNA is indefensible as a norm governing telling. Consider: SYNCHRONIC DIFFERENT STAKES: Alf is engaged in two email exchanges about the bank’s opening hours, one with Bev and one with Cat. Alf knows the matter is crucial for Bev, since she needs to urgently deposit her pay cheque to cover an impending bill. Alf also knows that nothing hinges on it for Cat. Alf remembers that the bank was open on a Saturday two weeks ago, and purely on that basis he sends an email to both Bev and Cat in which he tells them flat out that the bank is open on Saturday. (adapted from Pelling 2014, 345)

We all feel that it is less appropriate for Alf to tell Bev the bank will be open than for him to tell Cat. How can the supporter of a uniform knowledge account for assertion and telling accommodate this result? Pelling argues that, in a Classical Invariantist framework,² she can’t. As such, he concludes, in a Classical Invariantist framework, the knowledge norm is implausible for telling. Notice that, since Alf tells Bev and Cat at the same time that the bank will be open, it seems that Alf knows the bank will be open when he tells Bev if and only if he knows it when he tells Cat. This, Pelling argues, leaves the supporter of a knowledge account of telling unable to explain the difference in propriety. In Pelling’s view, the knowledge account of assertion does not face a similar problem because DIFFERENT STAKES involves two tellings, but only one assertion. In the light of this, Pelling suggests that, as opposed to mere assertion, one’s entitlement to tell a person that p depends partly on how much it matters, in the context of one’s relationship to that person, that p is true (2014, 343). As such, he argues, the norm for telling, as opposed to the norm for assertion, needs to be hearer-oriented and stakes-sensitive. In sum, Pelling supports his case for NC by arguing that: ² Pelling dismisses contextualism as implausible on independent grounds, and does not discuss pragmatically encroached invariantisms.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

    

137

(1) there are different characteristic communicative intentions behind assertion and telling (which suggests that they are governed by different norms), and (2) the invariantist defender of a knowledge norm for both cannot explain SYNCHRONIC DIFFERENT STAKES. Below, I will argue that both Pelling’s claims above are problematic. Let us start with (1), and distinguish between two types of communicative intentions associated with speech acts: essential communicative intentions—accompanying every token of the type—and characteristic communicative intentions—often attached to tokens of that type. Now, it is true that telling, but not assertion, has directedness towards an audience as an essential communicative intention. However, contra Pelling, the picture looks quite different when it comes to characteristic communicative intentions: extremely plausibly, the vast majority of assertions are tellings. As such, generating true belief/knowledge in the hearer is a characteristic communicative intention for assertion and telling alike. Furthermore, plausibly enough, as I have argued at length, this is the main epistemic function of both assertion and telling to begin with. If that is the case, against Pelling, telling and assertion share their characteristic communicative intentions. Of course, this is plausibly so because the vast majority of assertions are instances of telling. If this is right, we can maybe reconstruct Pelling’s case as follows: the essential communicative intention involved in assertion is presenting p as true; the essential communicative intention involved in telling is presenting p as true to an audience; therefore, assertion and telling are characterized by different essential communicative intentions, which gives us reason to believe that the norms regulating the two are different too. This reformulation will not do the intended work, however. Notice that, as Pelling himself notes, for actions in general, norms governing them tend to be associated with the characteristic intentions they involve, rather than merely with their corresponding essential intentions. Consider any action governed by norms, say for instance driving; the essential intention involved in driving is, plausibly, something like: using the vehicle to locomote. The characteristic intention involved in driving is, plausibly, safely getting us to our destination. But safely getting to our

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

138      destination is not an essential intention associated with driving; I might just drive around town for the fun of it. Importantly, though, norms regulating driving will be associated with this characteristic intention: for instance, the norm regulating speed limit will be there to ensure that we get to our destination within a reasonable amount of time, while still being on the safe side. Sure, I might be driving around aimlessly, for the fun of it; that does not, however, imply that separate norms would apply to me.³ Assertion and telling are types of action. In line with other types of action, then, we should expect the norms regulating them to also be associated with the characteristic intentions they involve: generating true belief/knowledge in an audience. If that is the case, it is less than clear that telling and asserting should be governed by different norms, as Pelling would have it. We have seen that Pelling’s argument fails to establish claim (1). Let us now move on to his claim (2). Recall that Pelling’s SYNCHRONIC DIFFERENT STAKES was supposed to create a difficulty for the Classical Invariantist defender of the thesis that KNA governs both assertion and telling due to the fact that, according to Pelling, it featured one assertion but two tellings. Since one telling in this case is intuitively appropriate, while the other is not, and since the speaker either knows or he doesn’t at the moment of the utterance, it follows, according to Pelling, that telling does not afford a knowledge norm. In contrast, Pelling argues, KNA may still be true, since the case only features one assertion. The first problem with this way to describe the case is speech act theoretic: just because assertion and telling are not one and the same speech act, it does not follow that telling can be instantiated independently. After all, telling is plausibly, and by Pelling’s own lights, a species of assertion—i.e. an assertion that is directed at an audience. If that’s so, telling cannot be instantiated without assertion being instantiated. Similarly, for instance, waltzing is a species of dancing; as such, waltzing cannot be instantiated without dancing being instantiated.

³ Thanks to an anonymous referee for pressing me on making this point clear.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

    

139

From this it follows that Pelling’s speech act theoretic description of the case is mistaken: there are no possible cases in which two tellings but only one assertion are instantiated, just as there are no possible cases in which two instances of waltzing but only one instance of dancing occur. As such, we are left with two ways to describe SYNCHRONIC DIFFERENT STAKES: either it features two instances of telling (and the corresponding two assertions)—one to Bev and one to Cat—or one instance of telling (and the corresponding assertion)—to Bev and Cat. From this it follows that the case cannot offer support for a picture on which assertion and telling are governed by different norms: if the telling is appropriate, so is the corresponding assertion, and, conversely, if the telling is inappropriate, so is the corresponding assertion. Furthermore, the champion of knowledge norms for constatives and Classical Invariantism can easily explain the data on either of the two options: just as in the ASPIRIN cases, the shiftiness intuition here tracks all-things-considered permissibility. If there are two tellings instantiated in this case, epistemically—i.e. pertaining to the epistemic function of telling—both tellings are permissible. Prudentially—i.e. pertaining to the prudential function of telling—telling Cat that the bank will be open is permissible, while telling Bev is not. Since prudential normativity overrides epistemic normativity at the context, due to survival-stringency, telling Cat that the bank will be open is all-things-considered permissible, while telling Bev is not. Alternatively, one may think that the case only involves one instance of telling, directed towards an audience made out of two people, Cat and Bev. If so, Alf ’s telling—i.e. pressing the ‘send’ button—is all-things-considered inappropriate on prudential grounds, pertaining to the prudential function of his speech act, and in virtue of Bev’s prudential concerns. Of course, had he only told Cat, the stakes of that particular context would have been lower, which would have rendered his telling all-things-considered permissible.

Conclusion I have argued that the case for Non-Commonality between the normativity of telling and assertion fails. I have argued that views construing

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

140      telling as a commissive or a directive will not affect the argument for a knowledge norm for telling put forth here. Further on, I have shown that assertion and telling share the same characteristic communicative aim, which suggests that they are governed by a common norm. In a nutshell, then, I have argued that there’s nothing special about the epistemic normativity of telling.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

10 A Special Case Moral Assertion

In the previous two chapters of this third part of the book, I have developed non-shifty, purist knowledge accounts of the epistemic normativity of two constatives—conjecturing and telling—and resisted attempts by others to push for their practical context sensitivity. This chapter partially departs from the party line: I argue that context—albeit epistemic contextual factors rather than practical contextual factors— matters for the special case of the normativity of moral assertion. To be clear, the account defended here for moral assertion fits snugly with my general view, in that it remains a knowledge account in a Classical Invariantist framework. It does, however, incorporate a degree of sensitivity to context, pertaining to the epistemic needs for moral understanding of the audience involved. Also in the way of further clarification: the account developed here rests on arguments by others that understanding is the norm for belief with moral content. If this claim turns out to be too strong—e.g. if it turns out that knowledge is enough for permissible moral belief—moral assertion will cease to be a special case of assertion, and will thus afford a biconditional knowledge norm. In this sense, the claim defended here can be read as a conditional claim. Here it goes: In recent years, much attention has been given to the epistemic credentials of belief based on moral testimony. Some people think pure moral deference is wrong, others disagree. It comes as a surprise, however, that while the epistemic responsibilities of the receiver of moral testimony have been closely scrutinized, little to no discussion has focused on the epistemic duties of the speaker. As far as I can see, the reason why this might be is that people have generally assumed that whatever the norm for assertion in general turns out to be, it will also govern moral assertions. This, however, of course, need not be the case; Shifty Speech and Independent Thought: Epistemic Normativity in Context. Mona Simion, Oxford University Press (2021). © Mona Simion. DOI: 10.1093/oso/9780192895288.003.0010

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

142      remember: just because waltzing is a species of dancing, it need not follow that it is merely governed by norms pertaining to dancing in general. To the contrary, what makes waltzing into a particular species of dancing to begin with is the fact that is governed by extra-norms, on top of the ones governing general dancing. Similarly, the norm for moral assertions can be stronger than the norm for assertion in general: more can be required in terms of epistemic support for proper moral assertion. This chapter defends a functionalist account of the normativity of moral assertion. According to this view, in virtue of its function of reliably generating moral understanding in the audience, a moral assertion that p needs be knowledgeable and accompanied by a contextually appropriate explanation why p.

10.1 Testimony and Moral Understanding Consider the following case from Alison Hills (2009, 94): ELEANOR: Eleanor has always enjoyed eating meat, but has recently realized that it raises some moral issues. Rather than thinking further about these, however, she asks a friend who tells her that eating meat is wrong. Eleanor knows that her friend is normally trustworthy and reliable, so she believes her and accepts that eating meat is wrong.

Many people in the literature have the intuition that there is something wrong with the way in which Eleanor proceeds; that rather than taking her friend’s testimony for granted, she should reflect on the issue herself. Call these people ‘pessimists’¹ about moral testimony. Now, importantly, there are two ways in which one can be a pessimist: a weak and a strong way.² According to the strong pessimist, one should not form moral beliefs on the mere say-so of others, because the latter cannot generate knowledge. This philosopher takes it that knowledge of moral matters is ¹ I borrow the ‘pessimism/optimism’ terminology from Hopkins (2007). The further distinction between weak and strong pessimism is mine. ² For people who think moral testimony can generate knowledge, see e.g. Jones (1999), Driver (2006), Hopkins (2007). For strong pessimism, see e.g. Wolff (1990), Lackey (2013).

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

  :  

143

too highbrow an affair to be gained second hand: more work than that is needed. Furthermore, she generally takes it that the function of testimonial exchanges is to generate the corresponding knowledge in the audience. Since moral testimony cannot do that, according to the pessimist, one should not form beliefs about moral issues from mere testimonial sources. Now, a lot of ink has been spilled on disputing the claim that one cannot gain knowledge from moral testimony, and with some degree of success too. It is fair to say that, in recent years, the state of the art is such that most people would agree that moral testimony can generate knowledge (Driver 2006, Hills 2009, Jones 1999). As such, for the purposes of this chapter, I will set strong pessimism aside. At the same time, a new variety of pessimism is born: weak pessimism. What these philosophers want to say is that, even though one can get knowledge of moral facts from testimony, this is still not good enough; there is still something wrong about forming moral beliefs on mere say-so, even though they are knowledgeable beliefs. According to weak pessimists, first, when it comes to moral matters, our epistemic goals are more highbrow affairs than just merely gathering a fair number of knowledgeable beliefs: what we are after is understanding the relevant matters (e.g. Nickel 2001, Hills 2009, Hopkins 2007, Howell 2014, McGrath 2011). Alison Hills, for instance, argues that there are at least two reasons why this is so: moral understanding is an essential mean to morally good character and morally worthy action. The further assumption is that understanding is more than just knowledge of the relevant matter. Hills (2009) suggests that understanding is constituted by a set of abilities that allow one to employ the relevant piece of information beyond the issue at hand: To understand why p you must have an ability to draw conclusions about similar cases, and to work out when a different conclusion would hold if the reasons why p were no longer the case’ (194). For this, it is argued, mere knowledge is not enough. One possible reply that easily comes to mind, however, is that maybe what is actually needed is, well, more knowledge. To see this, think again about Eleanor: it looks as though, if on top of knowing that eating meat is wrong, she were to also know why this is so, her moral beliefs would be

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

144      beyond criticism. In reply to this worry, pessimists rely on views taking understanding to be a sui generis epistemic state; according to these views, there is more to understanding that p than merely having knowledge why p. In line with people like Stephen Grimm (2006),³ for instance, Sarah McGrath suggests that understanding is a matter of possessing a particular skill of grasping the relevant connections between your pieces of knowledge, and, in this, it works a lot like a priori insight. There seems to be a clear difference between knowing that a conclusion is entailed by a set of premises and being able to see the entailment relation. Similarly, according to McGrath, moral understanding requires not only knowing that an action is wrong, or even why it is wrong, nor simply the ability to make inferences about its moral status based on information about its morally relevant properties, but also seeing or grasping the connection between its having those features and its being wrong. As such, McGrath argues, merely relying on moral testimony cannot and will not get you far enough, epistemically; what needs be the case is that you come to grasp the relevant relations yourself. Now, several optimists about moral testimony, such as, notably, Julia Driver (2006) and Paulina Sliwa (2012), disagree about the limited capacities of testimony to generate permissible belief or build good moral character. According to them, this is a mere contingent feature associated with most cases of moral testimony, but not with all. In particular, according to Sliwa, when we (justifiably) take the speaker to be more reliable than us on the matters at hand—maybe because we are concerned that our judgement is compromised by bias or self-interest, or because the other person is generally better at making moral judgements—it looks as though not only is it perfectly permissible to reply on their testimony, it is the responsible thing to do. Similarly, it seems perfectly fine for one to seek and follow moral advice when not able to make up one’s own mind on a particular moral matter. As such, Sliwa argues, it is not the case that there is something essentially wrong about relying on moral testimony. Furthermore, Sarah McGrath brings cases featuring agents who, against their own beliefs on the matter, take good moral advice from ³ See also Kelp (2015) for an account of understanding as systematic knowledge.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

  :  

145

others, to show that there are instances where pure moral deference is not only permissible but a good idea. According to McGrath, this faces us with a puzzle: what is it that explains the fact that, in some cases (e.g. ELEANOR), pure moral deference seems criticizable, while in others it seems the right thing to do? Whether weak pessimists or optimists are right about these issues will fortunately be irrelevant for our purposes here. Note, however, that there are a few things that most actors in the debate—both friends and foes of moral testimony—seem to agree upon: in most cases, moral testimony that p is deficient in generating the right epistemic standing with regard to p in the hearer; the right epistemic standing is moral understanding, which, in turn, is more than just a bundle of pieces of knowledge, and reliably conducive to morally worthy actions and character traits. Now, recall the derivation of KNA in the previous chapters: the function of assertion is to generate testimonial knowledge in hearers; since in the vast majority of cases, knowledge on the side of the speaker is needed for generating knowledge on the side of the hearer, KNA holds. One question that arises at this point is: how about moral assertion? What is the function of moral assertion? After all, the phenomenon does seem to be fairly ubiquitous, so it would be surprising if it turned out to be a pointless endeavour. If the literature is right, and, in most cases, moral testimony cannot generate moral understanding in hearers, why are we engaging in it to begin with? To what aim? We have seen above that most people take it that the goal of moral inquiry is understanding. The next section will defend the norm for moral assertion that drops out of this picture. Again, for the reader who favours different accounts of the goal of moral inquiry, this chapter should be read as making a conditional claim: if understanding is the goal of moral inquiry, then the norm defended here governs moral assertion.

10.2 The Function and Norm of Moral Assertion Importantly, note that my functionalist KNA does not claim that the speaker’s asserting from knowledge is all there needs to happen for knowledge to be generated on the hearer’s side; on top of this, the speaker needs

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

146      be decently reliable, the hearer needs to accept the testimony and display minimal non-gullibility when doing so, and so on. What is of central concern to us, though, given that we are after the norm for assertion, is what epistemic features the assertion involved in the relevant testimonial exchange needs to display for the relevant function to be fulfilled, independently of whatever else needs be present on top of a proper assertion to this effect. As such, the question is, ceteris praesentibus, what should assertion (epistemically) bring to the table in order for the function to be reliably fulfilled? And, importantly, the answer to this question is: knowledge. We have also seen above, however, that many people think that in some domains knowledge is not enough: the domain-specific central epistemic goal is, rather, to understand the issue at hand. Morality is such a domain: according to many, to be able to engage in morally worthy action, build a morally virtuous character, and so on, one needs more than to know the relevant moral matters of fact: one needs moral insight into how they hang together. Now, importantly, just as in the case of general testimony, it is hardly surprising that, no matter how epistemically endowed one’s moral assertion might be, it might still fail to generate the relevant epistemic standing in the hearer in isolation: other features need be in play to this effect. According to pessimists, on top of the requirements for general testimony, when it comes to moral affairs, the hearer also needs to actively participate in the exchange and think through the relevant matters herself, in order to achieve moral understanding. However, this does not mean that the moral assertion at work does not have its role to play in this affair: we should expect that, just as in the case of testimony in general, moral assertion should meet the strongest epistemic standard needed for generating the target epistemic standing in the hearer, ceteris praesentibus. Since in the case of moral matters the widely agreed target epistemic standing is moral understanding, here is, on a first approximation, my proposal: The Explanation Proffering Norm of Moral Assertion* (EPNMA*): One’s moral assertion that p is epistemically permissible only if (1) one

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

  :  

147

knows that p and (2) one’s assertion is accompanied by an appropriate explanation why p. The intuitive thought behind EPNMA* is the following: say that pessimists are right, and no amount of knowledge will get the hearer to understand the moral matter at hand, unless she also ‘runs’ the relevant issue through her own rational capacities in order to ‘grasp’ how things hang together. Now, take a hearer who is perfectly disposed to do the relevant grasping work; what is the most reliable way to generate moral understanding in this particular hearer? It looks fairly intuitive that the answer is: by knowledgeably asserting to the relevant matter of fact (‘p’) and offering the appropriate explanations as to why p is the case. Recall Eleanor, and ask yourself: what would have to be the case on the speaker’s side in order to reliably generate understanding why eating meat is wrong on Eleanor’s side, granted that she would be well disposed to do the needed ‘grasping’ work? Presumably, her friend would need to say something along the lines of: ‘eating meat is wrong because animals are living beings just like you and they deserve to be treated with equal respect’. In this case, it looks as though, if Eleanor runs the entailment relation through her own rational mechanisms, she comes to understand why eating meat is wrong. Also, it does not look as though there is anything wrong with her corresponding belief. Now, this being said, one pressing question that arises is: what will count as an appropriate explanation? First, importantly, I take it that it will not be a merely quantitative matter: to the contrary, appropriateness will refer to both amount and kind of reasons brought in support of the target moral claim. Second, importantly, the running hypothesis of this chapter is that this will be a contextually variant matter: more/stronger reasons will need to be brought in support of some assertions, and less/ weaker for others. One motivation for this is that I take it, together with many people in the relevant literature,⁴ that the proper attribution of understanding simpliciter is a contextually sensitive matter: it looks as

⁴ See, e.g., Kelp (2015).

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

148      though it is one thing for my ten-year-old to count as understanding physics, and quite another in the case of myself: different standards seem to apply.⁵ Also, according to many, proper explanation itself is a highly context-sensitive matter.⁶ This is not all there is to the context-sensitivity of EPNMA*, however; even if it turns out that understanding is invariant, some contexts seem to require more than others in the way of explanation for understanding to be generated in hearers, depending on several contextual determiners such as: the previous knowledge and level of general intelligence of the hearer, the general assumptions about morality present at the context, the level of difficulty and controversy of the subject matter, how much (morally)⁷ hinges on the issue, the urgency of the situation, etc. As such, we should reformulate our norm as follows: The Explanation Proffering Norm of Moral Assertion (EPNMA): At a context C, one’s moral assertion that p is epistemically permissible only if (1) one knows that p and (2) one’s assertion is accompanied by a C-appropriate explanation why p. EPNMA nicely drops out of the widely accepted epistemic goal with regard to moral matters—i.e. moral understanding—and it is, as such, well motivated. Note, also, that this result should be fairly uncontroversial, no matter what side of the moral testimony debate one prefers. After all, note that even if the sui generis view about understanding turns out to be false, EPNMA is unaffected: it would only mean that there’s not much ‘grasping’ work lying on the shoulders of the hearers: all they need to do is listen, comprehend, and believe what they are being told.

⁵ If one is inclined towards a virtue-theoretic account of understanding, in terms of some skill to grasp the relevant relations, it makes sense to have a contextualist view: after all, standards for what counts as skilled action also seem to vary with context. Here is John Greco: ‘[W]hen I say that S has the ability to hit baseballs, the practical reasoning context helps to determine what I am claiming here. If I am a baseball executive in a discussion about whether to trade for Jeter, I will be claiming something very different than if I am a Little League coach trying to decide where to put the new seven year-old in the line-up’ (Greco 2009, 21). ⁶ See e.g. van Fraassen (1980). ⁷ Importantly, this account does not take practical considerations to encroach upon the epistemic normativity of moral assertion.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

  :  

149

Also, the view nicely explains what is intuitively wrong about the initial Eleanor case: in general, it is impermissible to believe based on epistemically impermissible sources. Unreliable testifiers, misleading evidence, and wishful thinking are just a few cases in point. Similarly, one should not believe based on epistemically impermissible assertions. Eleanor should not form a belief based on her friend’s assertion because the latter misses an appropriate explanation. Furthermore, the account offers a nice solution for Sarah McGrath’s puzzle about moral deference: what is it that explains the fact that, in some cases, pure moral deference seems criticizable, while in others it seems the right thing to do? According to the present account, in some cases, contextual determiners are such that little to no explanation is needed for proper assertion and, therefore, for generating the relevant stance in the hearer. This will, for instance, be the case in urgency cases: if I shout ‘Killing is wrong!’ and thereby stop a murder from happening, my assertion will be perfectly fine in the absence of much explanation.⁸ Similarly, the Sliwa (2012) cases featuring worries about one’s own reliability being affected by bias and moral advice are cases in point. To the contrary, in contexts where the hearer, like Eleanor, has never given the subject matter any thought, and there’s nothing keeping the speaker from going more into depth about the subject matter, more and better explanatory support is needed for proper assertion.

10.3 Response to Critics In the previous sections, I have been arguing that, if it is plausible to think that knowledge is the norm of assertion in general, in virtue of assertion’s function of generating knowledge in hearers, and if in the case of moral assertion, the epistemic function at stake is moral understanding, the norm governing assertion will be EPNMA. That is, epistemically permissible moral assertion will be knowledgeable and accompanied by a contextually appropriate explanation, since this will maximize its ⁸ Note that this case nicely maps on to Tim Williamson’s ‘TRAIN’ urgency case for general assertion (2000).

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

150      reliability in fulfilling its epistemic function: generating moral understanding on the hearer’s side. In this section, I will try to answer a few objections that have been raised—in print or conversation—against my account. Objection 1. How about non-sophisticated cognizers? Uncontroversially, small children do not have much in the way of reflective capacities, so they will likely not be able to deliver sophisticated explanations about moral matters of fact. However, many people believe that they are knowers, and therefore able to perform perfectly fine assertions. Furthermore, it does not look as though they cannot assert when it comes to moral matters: surely it is fine for a three-year-old to tell her mother that lying is wrong, for instance. Where does EPNMA leave them? Note, though, that treating non-sophisticated cognizers like a special case is hardly uncommon in the epistemology of moral testimony (or, indeed, general testimony, see e.g. Goldberg (2008)), and therefore not easily imputable to EPNMA; for instance, both pessimists and optimists will agree that pure moral deference is perfectly fine in the case of unsophisticated hearers, such as small children. This suggests that, when it comes to highbrow moral matters, unsophisticated cognizers just require special treatment both as receivers and as producers of testimony. Objection 2. Many people in the literature take it that there is an interesting analogy between moral and aesthetic testimony. That is, just as in the moral case, it feels equally wrong to form beliefs based on someone’s testimony about aesthetic matters in isolation. However, it does not look as though EPNMA will map equally nicely to the normativity of aesthetic assertion, in that explaining why an aesthetic claim holds does not seem to be a difference maker. Here is Robert Hopkins on the matter: ‘If you tell me that some film you have seen is excellent, it is far from clear that I can legitimately adopt your view. Your testimony might motivate me to watch the film, but it does not give me the right to the belief that the film is excellent. I won’t have that right until I’ve seen the film myself ’ (Hopkins 2011). I don’t know enough about aesthetics to take a clear stance on this issue. If Hopkins is right, though, it looks as though proper belief about aesthetic matters is more difficult to come across than proper

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

  :  

151

belief about moral matters. In the case of the former, something like direct acquaintance with the object of approval or disapproval seems to be needed (at least somewhere down the road);⁹ needless to say, one can have moral beliefs about abortion without ever having had one. Objection 3. Why not an understanding norm of assertion? We have seen that, plausibly, if generating testimonial knowledge is the main epistemic function of assertion, and if its norm follows from its function, knowledge is the norm of assertion. One question that arises at this point is: if we take the relevant function in the case of moral assertion to be generating understanding, why does this not generate an understanding norm of assertion (Kelp Forthcoming, Lewis 2019)? The answer is twofold: first, let us grant that whatever the theorist of understanding takes it that needs to be present on the hearer’s side is present; now let us ask ourselves: is it plausible that understanding on the speaker’s side is going to be the most reliable way to generate understanding on the hearer’s side? The answer seems to be ‘no’. Here is why: for one, it is hardly the case that understanding that p reliably leads one to being able to explain why p is the case. Most of us know our domains well, but are hardly good teachers. This, if not immediately intuitively plausible, is evidenced by the amount of resources we invest in training teachers in pedagogical techniques. Second, if the sui generis account of understanding is right, it looks as though understanding is too strong a requirement on the speaker’s side to begin with. That is because she can reliably generate understanding in her hearers while missing the relevant ‘grasping’ capacities herself. All that is needed for epistemic function fulfilment in this case is that the speaker presents the hearer with the relevant explanation and that the hearer does the grasping work herself.

Conclusion This chapter has proposed an account of the normativity of moral assertion. I have argued that, in view of reliably fulfilling its function of ⁹ One possible exception is coming to know that the film is good from testimony in cases where you know that the speaker has the exact same taste as you in cinematography.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

152      generating moral understanding, moral assertion should be knowledgeable and accompanied by an explanation, where the relevant notion of appropriateness is a contextually determined one. Crucially, though, the efficacious contextual parameters are epistemic, rather than practical, factors.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

Conclusion Many philosophers think that practical stakes affect epistemic standards for proper thought and speech. From the get-go, my gut feeling was that this must be incorrect: it can’t possibly be that whether you or I care more or less about a subject matter affects how much epistemic support we need in order to think or speak about it. After many years of research, I remain convinced that my initial gut reaction was correct. Indeed, the epistemic modifies the practical, not the other way around. This book has outlined the argumentative roadmap that got me from the initial reaction to my evidenced confidence in the independence of epistemically proper thought and speech from practical stakes. The first part of the book started off (Chapter 1) by noting that most actors in the debate assume that the intuitive variability of proper assertion with practical stakes motivates the following Shiftiness Dilemma: either (1) we embrace a knowledge norm of assertion, and are forced into a view that takes knowledge, or ‘knowledge’, to be sensitive to practical stakes (Knowledge Sensitivism), or (2) we stick to our Classical Invariantist guns, but then we lose the knowledge norm and we get practical sensitivity in the normativity of assertion (Assertion Sensitivism). I have argued here that this dilemma is more dangerous to our epistemic affairs than it might seem, since it threatens to generalize in three ways: to all context-invariant norms of proper assertoric speech, to all constative speech acts, and to all epistemic normative constraints. The remaining chapters of this part of the book were concerned with looking for a way out of this dilemma on behalf of the champion of epistemic independence from the practical. The first move I discussed was by authors taking the second horn of the dilemma: i.e. venturing to preserve Classical Invariantism by arguing for context-sensitivity of epistemically proper assertability. I argued that the view either collapses into

Shifty Speech and Independent Thought: Epistemic Normativity in Context. Mona Simion, Oxford University Press (2021). © Mona Simion. DOI: 10.1093/oso/9780192895288.003.0011

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

154      Knowledge Sensitivism, or is incompatible with highly uncontroversial value-theoretic claims (Chapter 2). The second attempt to escape the dilemma places the blame for the context-sensitivity data on the pragmatics of language. I have argued that the view has difficulties with explaining cases of shiftiness of assertions that do not involve knowledge attributions or explicitly tabled error possibilities (Chapter 3). Last, I have looked at a solution according to which the unassertability in the high-stakes contextualist cases can be explained in terms of the subjects’ lack of higher-order knowledge. I have argued that the absence of higher-order knowledge is not a difference maker (Chapter 4). The discussion in the first part of the book was mostly pessimistic: in a nutshell, I have argued that a widely embraced dilemma generalizes dangerously to most of our epistemic affairs, and found no good extant proposal to escape it. In contrast, the second part of the book is exceptionally optimistic. First, I have argued that the perceived and widely accepted Shiftiness Dilemma is a false dilemma, resting on normative ambiguation: the shiftiness of proper assertion with practical stakes concerns practical and all-things-considered propriety, not epistemic propriety. Insofar as the Knowledge Norm of Assertion is an epistemic norm, then, it is perfectly compatible with Classical Invariantism (Chapter 5). Indeed, I offered independent reason to believe that if the Knowledge Norm and Classical Invariantism are correct, variation in all-things-considered proper assertability with practical stakes is exactly what we may expect. My basic proposal is that we should combine invariantism with a functionalist account of assertion (Chapter 6): according to the account I defend, assertion is governed by a knowledge norm in virtue of its epistemic function of generating testimonial knowledge. Requirements generated by other, more survival-stringent functions of assertion, though, such as its prudential function, can override the constraints imposed by the epistemic function, and render the knowledge requirement either too strong or too weak for all-things-considered permissible assertion. All-things-considered permissible assertion can vary with practical stakes; epistemically permissible assertion does not.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi



155

Last but not least, the final chapter of Part II generalizes the functionalist knowledge account of the normativity of assertion to all constative speech, thereby also delivering the first integrated account of the epistemic normativity of constatives. I argue for a knowledge account for constative speech from three different angles: (1) the nature of communicative speech acts, (2) the relation between assertion and other constatives, and (3) the normativity of belief together with constatives’ epistemic function. The third and last part of the book looks more closely into the contextual behaviour of three constatives in particular—conjecture, telling, and moral assertion. Chapter 8 develops the first fully fledged account of the epistemic normativity of conjecture in the literature, and it goes sharply against orthodoxy, in arguing that the epistemic requirements placed on us by conjecture are stronger than those governing assertion. Chapter 9 argues, against recent work, that there is no reason to give the speech act of telling any special, hearer-oriented normative treatment. Finally, in Chapter 10, I turn to a special case of assertive—moral assertion—and I defend a view according to which, in virtue of its function of reliably generating moral understanding in the audience, a moral assertion that p needs be knowledgeable and accompanied by a contextually appropriate explanation why p. Some readers will not have agreed with me at times, I would expect. That’s why, on a final note, I find it important to stress once again that, while for me and my knowledge-centric, purist taste in epistemic normativity this book hangs together nicely as is, several results here hold independently of others: one need not be a Knowledge Norm champion, for instance, nor a Classical Invariantist, in order to accept that the argument for the incompatibility of the two doesn’t work, due to normative ambiguation; people liking other non-shifty norms, for instance (e.g. justification norms), can just enjoy the good news, since the result generalizes. Also, one need not like the knowledge norm in order to like and embrace my functionalist picture of the normativity of constatives: different views on the function in question may well deliver different views on the norms dropping out of it. What really matters is to be found at a higher level of abstraction: the norm-generating functionalist mechanisms, and their excellent non-shifty credentials/explanation of the

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

156      shiftiness data. Last but not least, the results in the third part of the book can survive even for readers who are not inclined to buy into anything that happened before, since the accounts there, while fitting snugly within my general normative framework for constatives, are defended on largely independent grounds.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

References Anscombe, G. E. M. (1979). What Is It to Believe Someone? In C. F. Delaney (ed.), Rationality and Religious Belief. South Bend: University of Notre Dame Press, 141–51. Anthony, G. (1996). Active Learning in a Constructivist Framework. Educational Studies in Mathematics 31 (4): 349–69. Austin, J. L. (1946). Other Minds. In J. L. Austin, Philosophical Papers, 3rd edition. Oxford: Oxford University Press (1979), 76–116. Bach, K. (2010). Knowledge in and out of Context. In J. Campbell, M. O’Rourke, and H. Silverstein (eds), Knowledge and Skepticism. Cambridge, MA: MIT Press. Bach, K. and Harnish, R. M. (1979). Linguistic Communication and Speech Acts. Cambridge, MA: MIT Press. Benton, M. (2014). Knowledge Norms. Internet Encyclopedia of Philosophy. http:// www.iep.utm.edu/kn-norms/#H2 [29.01.2015]. Benton, M. (2016). Gricean Quality. Noûs 50: 689–703. BonJour, L. (1985). The Structure of Empirical Knowledge. Cambridge, MA: Harvard University Press. Brown, J. (2005). Williamson on Luminosity and Contextualism. The Philosophical Quarterly 55 (219): 319–27. Brown, J. (2006). Contextualism and Warranted Assertibility Manoeuvres. Philosophical Studies 130: 407–35. Brown, J. (2010). Knowledge and Assertion. Philosophy and Phenomenological Research 81: 549–66. Buckwalter, W. and Turri, J. (2014). Telling, Showing and Knowing: A Unified Account of Pedagogical Norms. Analysis 74 (1): 16–20. Buller, D. J. (1998). Etiological Theories of Function: A Geographical Survey. Biology and Philosophy 13 (4): 505–27. Cañadas, M., Deulofeu, J., Figueiras, L., and Reid, D. (2007). The Conjecturing Process: Perspectives in Theory and Implications in Practice. Journal of Teaching and Learning 5 (1): 55–72. Capes, J. (2012). Blameworthiness without Wrongdoing. Pacific Philosophical Quarterly 93 (3): 417–37. Cappelen, H. (2011). Against Assertion. In Jessica Brown and Herman Cappelen (eds), Assertion: New Philosophical Essays. Oxford: Oxford University Press, 21–47. Cohen, S. (1999). Contextualism, Skepticism, and the Structure of Reasons. Philosophical Perspectives 13: Epistemology: 57–89. Cummins, R. (1975). The Journal of Philosophy 72 (20): 741–65.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

158  David, M. (2005). Truth as the Primary Epistemic Goal: A Working Hypothesis. In M. Steup and E. Sosa (eds), Contemporary Debates in Philosophy. Oxford: Blackwell, 296–312. Davidson, D. (1987). Knowing One’s Own Mind. Proceedings and Addresses of the American Philosophical Association 61: 441–58. DeRose, K. (1992). Contextualism and Knowledge Attributions. Philosophy and Phenomenological Research 52 (4): 913–29. DeRose, K. (2002). Assertion, Knowledge, and Context. Philosophical Review 111: 167–203. DeRose, K. (2009). The Case for Contextualism. Oxford: Oxford University Press. Douven, I. (2006). Assertion, Knowledge, and Rational Credibility. Philosophical Review 115: 449–85. Driver, J. (2006). Autonomy and the Asymmetry Problem for Moral Expertise. Philosophical Studies 128 (3): 619–44. Ewing, A. C. (1947). The Definition of Good. London: Macmillan. Fantl, J. and McGrath, M. (2002). Evidence, Pragmatics, and Justification. Philosophical Review 111 (1): 67–94. Fantl, J. and McGrath, M. (2009). Knowledge in an Uncertain World. Oxford: Oxford University Press. Fantl, J. and McGrath, M. (2012). Arguing for Shifty Epistemology. In J. Brown and M. Gerken (eds), Knowledge Ascriptions. Oxford: Oxford University Press, 55–74. García-Carpintero, M. (2004). Assertion and the Semantics of Force-Markers. In C. Bianchi (ed.), The Semantics/Pragmatics Distinction. CSLI Lecture Notes. Chicago: The University of Chicago Press, 133–66. Garcia-Carpintero, M. (2019). Conventions and Constitutive Norms. Journal of Social Ontology. Online First. Geach, P. (1956). Good and Evil. Analysis 17: 33–42. Gerken, M. (2012). Discursive Justification and Skepticism. Synthese 189: 373–94. Gibbard, A. (1990). Wise Choices, Apt Feelings. Oxford: Clarendon Press. Goldberg, S. (2008). Testimonial Knowledge in Early Childhood, Revisited. Philosophy and Phenomenological Research 76: 1–36. Goldberg, S. (2018). Assertion: The Philosophical Significance of Assertoric Speech. Oxford: Oxford University Press. Goldberg, S. (2020). Conversational Pressure. Oxford: Oxford University Press. Graham P. J. (2010). Testimonial Entitlement and the Function of Comprehension. In Duncan Pritchard, Alan Millar, and Adrian Haddock (eds), Social Epistemology. Oxford: Oxford University Press, 148–74. Graham, P. J. (2012). Epistemic Entitlement. Nous 46 (3): 449–82. Graham, P. J. (2014). Functions, Warrant, History. In Abrol Fairweather and Owen Flanagan (eds), Naturalizing Epistemic Virtue. Cambridge: Cambridge University Press, 15–35. Graham, P. and Lyons, J. (Forthcoming). The Structure of Defeat: Pollock’s Evidentialism, Lackey’s Framework, and Prospects for Reliabilism. In Simion, M. and Brown, J. (eds), Reasons, Justification, and Defeat. Oxford: Oxford University Press.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi



159

Greco, J. (2009). Knowledge and Success from Ability. Philosophical Studies 142 (1): 17–26. Greenough, P. (2011). Truth-Relativism, Norm-Relativism, and Assertion. In J. Brown and H. Cappelen (eds), Assertion: New Philosophical Essays. Oxford: Oxford University Press, 197–231. Grice, P. (1989). Studies in the Way of Words. Cambridge, MA: Harvard University Press. Grimm, S. (2006). Is Understanding a Species of Knowledge? British Journal for the Philosophy of Science 57: 515–35. Haji, I. (1998). Moral Appraisability. Oxford: Oxford University Press. Hawthorne, J. (2004). Knowledge and Lotteries. Oxford: Oxford University Press. Hazlett, A. (2009). Knowledge and Conversation. Philosophy and Phenomenological Research 78 (3): 591–620. Hills, A. (2009). Moral Testimony and Moral Epistemology. Ethics 120: 94–127. Hinchman, E. S. (2005). Telling as Inviting to Trust. Philosophy and Phenomenological Research 70 (3): 562–87. doi: 10.1111/j.1933–1592.2005.tb00415. Hopkins, R. (2007). What is Wrong with Moral Testimony. Philosophy and Phenomenological Research 74 (3): 611–34. Hopkins, R. (2011). How to Be a Pessimist About Aesthetic Testimony. The Journal of Philosophy 108 (3): 138–57. Howell, R. (2014). Google Morals, Virtue, and the Asymmetry of Deference. Nous 48 (3): 389–415. Jones, K. (1999). Second-Hand Moral Knowledge. The Journal of Philosophy 96 (2): 55–78. Kelp, C. (Forthcoming). Moral Assertion. Ethical Theory and Moral Practice. Kelp, C. (2014). Two for the Knowledge Goal of Inquiry. American Philosophical Quarterly 51: 227–32. Kelp, C. (2015). Understanding Phenomena. Synthese 192: 3799–816. Kelp, C. (2018). Assertion: A Function-First Account. Nous 52: 411–42. Kelp, C. and Simion, M. (2017). Commodious Knowledge. Synthese 194 (5): 1487–1502. Kelp, C. and Simion, M. (2020). The C Account of Assertion: A Negative Result. Synthese 197: 125–37. Kelp, C. and Simion, M. (Forthcoming). Sharing Knowledge: A Functionalist Account of Assertion. Cambridge: Cambridge University Press. Kluwe, R. H. (1982). Cognitive Knowledge and Executive Controll: Metacognition. In Griffin, D. R. (ed.), Animal Mind – Human Mind. Springer, 201–24. Lackey, J. (2007). Norms of Assertion. Noûs 41: 594–626. Lackey, J. (2008). Learning from Words: Testimony as a Source for Knowledge. Oxford: Oxford University Press. Lackey, J. (2011). Assertion and Isolated Second-Hand Knowledge. In Jessica Brown and Herman Cappelen (eds), Assertion: New Philosophical Essays. Oxford: Oxford University Press. Lackey, J. (2013). Deficient Testimonial Knowledge. In T. Henning and D. P. Schweikard (eds), Knowledge, Virtue and Action: Essays on Putting Epistemic Virtues to Work. New York: Routledge, 30–53.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

160  Lackey, J. (2019). Assertoric Quality. In Sanford Goldberg (ed.), Oxford Handbook of Assertion. Oxford: Oxford University Press, 625–43. Lehrer, K. and Cohen, S. (1983). Justification, Truth, and Coherence. Synthese 55 (2): 191–207. Levin, J. (2008). Assertion, Practical Reason, and Pragmatic Theories of Knowledge. Philosophy and Phenomenological Research 76: 359–84. Lewis, M. (2019). The Norm of Moral Assertion: A Reply to Simion. Ethical Theory and Moral Practice 22 (4): 1043–9. Lord, E. (2018). The Importance of Being Rational. Oxford: Oxford University Press. Maitra, I. (2011). Assertion, Norms and Games. In Jessica Brown and Herman Cappelen (eds), Assertion: New Philosophical Essays. Oxford: Oxford University Press. McGrath, S. (2009). The Puzzle of Pure Moral Deference. Philosophical Perspectives 23 (1): 321–44. McKinnon, Rachel. (2013). The Supportive Reasons Norm of Assertion. American Philosophical Quarterly 50: 121–35. McNamara, P. (2019). Deontic Logic. The Stanford Encyclopedia of Philosophy (Summer 2019 Edition), Edward N. Zalta (ed.). https://plato.stanford.edu/arch ives/sum2019/entries/logic-deontic/. Millikan, R. (1984). Language, Thought, and Other Biological Categories. Cambridge, MA: MIT Press. Millikan, R. (2005). Language: A Biological Model. Oxford: Oxford University Press. Miscione, J., Marvin, R., O’Brien, R., and Greenberg, M. (1978). A Development Study of Preschool Children’s Understanding of the Words ‘Know’ and ‘Guess’. Child Development 49: 1107–13. Moore, G. E. (1962). Commonplace Book: 1919–1953. London: Allen and Unwin. Moore, G. E. (1993). Principia Ethica, revised edn. Cambridge: Cambridge University Press. Moran, R. (2005). Getting Told and Being Believed. Philosopher’s Imprint 5 (5): 1–29. Neander, K. (1991). The Teleological Notion of ‘Function’. Australasian Journal of Philosophy 69 (4): 454–68. Nickel, P. (2001). Moral Testimony and Its Authority. Ethical Theory and Moral Practice 4 (3): 253–66. Pelling, C. (2014). Assertion, Telling, and Epistemic Norms. Australasian Journal of Philosophy 92 (2): 335–48. Popper, K. (1963). Conjectures and Refutations: The Growth of Scientific Knowledge. London: Routledge. Portmore, D. (2005). Combining Teleological Ethics with Evaluator Relativism: A Promising Result. Pacific Philosophical Quarterly 86: 95–113. Pritchard, D. (2005). Epistemic Luck. Oxford: Oxford University Press. Rabinowicz, W. and Rönnow-Rasmussen, T. (2004). The Strike of the Demon: On Fitting Pro-Attitudes and Value. Ethics 114: 391–423. Reed, B. (2012). Resisting Encroachment. Philosophy and Phenomenological Research 85: 465–72. Reed, B. (2010). A Defense of Stable Invariantism. Noûs 44: 224–44.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi



161

Rescorla, M. (2009). Assertion and Its Constitutive Norms. Philosophy and Phenomenological Research 79: 98–130. Reynolds, S. (2002). Testimony, Knowledge, and Epistemic Goals. Philosophical Studies 110: 139–61. Ridge, M. (2013). Impassioned Belief. Oxford: Oxford University Press. Rysiew, P. (2001). The Context-Sensitivity of Knowledge Attributions. Noûs 35 (4): 477–514. Scanlon, T. M. (1998). What We Owe to Each Other. Cambridge, Mass: Harvard University Press. Schroeder, M. (2012). Value Theory. The Stanford Encyclopedia of Philosophy (Summer 2012 Edition), Edward N. Zalta (ed.). http://plato.stanford.edu/arch ives/sum2012/entries/value-theory/. Schroeder, M. (2015). What Makes Reasons Sufficient? American Philosophical Quarterly 52 (2): 159–70. Searle, J. R. and Vanderveken, D. (1985). The Foundations of Illocutionary. Cambridge, England: Logic, Cambridge University Press. Sidgwick, H. (1907). The Methods of Ethics, 7th edition. Indianapolis: Hackett. Simion, M. (2016). Assertion: Knowledge Is Enough. Synthese 193 (10): 3041–56. Simion, M. (2018a). Assertion: The Context Shiftiness Dilemma. Mind & Language 34: 503–17. Simion, M. (2018b). No Epistemic Norm for Action. American Philosophical Quarterly 55 (3): 231–8. Simion, M. (2019a). Conceptual Engineering for Epistemic Norms. Inquiry. Online First. Simion, M. (2019b). Epistemic Norm Correspondence and the Belief-Assertion Parallel. Analysis 79 (2), 260–5. Simion, M. (2019c). Knowledge-First Functionalism. Philosophical Issues, A Supplement to Nous. Online First. Simion, M. (2019d). Saying and Believing: The Norm Commonality Assumption. Philosophical Studies A 176 (8): 1951–66. Simion, M. (2020a). Epistemic Norms and Epistemic Functions. Book manuscript. Simion, M. (2020b). Testimonial Contractarianism: A Knowledge-First Social Epistemology. Nous. Online First. Simion, M. (2020c). The Epistemic Normativity of Conjecture. Manuscript. Simion, M., Kelp, C., and Ghijsen, H. (2016). Norms of Belief. Philosophical Issues, A Supplement to Nous 26 (1): 375–92. Sliwa, P. (2012). In Defense of Moral Testimony. Philosophical Studies 158 (2): 175–95. Slote, M. (1979). Assertion and Belief. In J. Dancy (ed.), Papers on Language and Logic. Keele: Keele University Press, 177–190. Slote, M. (1989). Beyond Optimizing. Cambridge, Mass: Harvard University Press. Smithies, D. (2012). The Normative Role of Knowledge. Nous 46 (2): 265–88. Sosa, E. (1993). Proper Functionalism and Virtue Epistemology. Nous 27: 51–65. Sosa, E. (2007). A Virtue Epistemology: Apt Belief and Reflective Knowledge, Vol. 1. Oxford: Oxford University Press.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 18/3/2021, SPi

162  Sperber, D. (2001). An Evolutionary Perspective on Testimony and Argumentation. Philosophical Topics 29: 401–13. Stanley, J. (2005). Knowledge and Practical Interests. Oxford: Oxford University Press. Stone, J. (2007). Contextualism and Warranted Assertion. Pacific Philosophical Quarterly 88: 92–113. Tanesini, A. (2020). The Gift of Testimony. Episteme. Online First. Turri, J. (2010). Epistemic Invariantism and Speech Act Contextualism. Philosophical Review 119: 77–95. Turri, J. (2014). Knowledge Guaranteed. Nous 47 (3): 602–12. Unger, P. (1975). Ignorance: A Case for Scepticism. Oxford: Clarendon Press. van Fraassen, B. C. (1980). The Scientific Image. Oxford: Clarendon Press. Weiner, M. (2005). Must We Know What We Say? Philosophical Review 114: 227–51. Whiting, D. (2013). Stick to the Facts: On the Norms of Assertion. Erkenntnis 78: 847–67. Williams, B. (1973). A Critique of Utilitarianism. In J. J. C. Smart and B. Williams, Utilitarianism: For and Against. Cambridge: Cambridge University Press, 77–150. Williams, B. (1981). Persons, Character, and Morality. In B. Williams, Moral Luck. Cambridge: Cambridge University Press, 1–19. Williamson, T. (2000). Knowledge and its Limits. Oxford: Oxford University Press. Williamson, T. (2005). Contextualism, Subject-Sensitive Invariantism and Knowledge of Knowledge. The Philosophical Quarterly 55 (219): 213–35. Williamson, T. (Forthcoming). Justification, Excuses, and Sceptical Scenarios. In Julien Dutant and Fabian Dorsch (eds), The New Evil Demon: New Essays on Knowledge, Rationality, and Justification. Oxford: Oxford University Press. Wolff, R. P. (1990). The Conflict between Authority and Autonomy. In Joseph Raz (ed.), Authority. Oxford: Blackwell. Zimmerman, M. (1997). Moral Responsibility and Ignorance. Ethics 107 (3): 410–26.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

Index For the benefit of digital users, indexed terms that span two pages (e.g., 52–53) may, on occasion, appear on only one of those pages. All-things-considered 59, 61–6, 79–80, 82–3, 86–7, 99–100, 127–8, 139, 154 Assertability 4, 8, 13, 18–20, 30, 32, 36–7, 39–41, 50, 66–7, 153–4 Assertion, epistemically permissible 7–8, 25, 60, 68, 86–7, 154 Assertion, epistemic norm of 12, 21, 29–30, 59, 70, 132 Assertion Sensitivism 4, 12, 18–24, 31, 53–4, 153 Association Claim 19–21, 23–4, 31, 69 Axiological 68–9

Deontic 19, 21, 24, 68–9, 109–10 DeRose 3–4, 8–12, 33–4, 37–8, 48–9, 58–61, 66, 68–9, 79–80, 82–6

Bach and Harnish 14, 88–96, 99–100, 113–17, 130, 133 Bank case 8–9, 82–3, 85 Belief, justified 3, 113 Belief, true 6, 20, 22, 28–9, 65, 77–8, 137–8 Blame 40–8, 52–3, 154 Brown vii, 12, 25, 29–30, 33–4, 38, 49–51, 81

Fantl and McGrath 5, 12, 16 Function, etiological 70, 74, 93, 98, 118 Function, prudential 68–9, 80–2, 87, 139, 154

Cappelen vii Conjecture, epistemically permissible 127–8 Conjecture, knowledge norm of 113–18 Constative speech 4, 6, 14–15, 18, 57, 87–8, 95, 103, 106, 131, 153, 155 Content Individuation 57, 61–3, 65–6, 68–9

Encroached 16 Epistemic aim 12, 30–1 Epistemic content 4, 13, 16–17, 61–6, 68–9 Epistemic good 21–4 Epistemic independence 57, 153–4 Epistemic position 3–4, 7–8, 33, 38, 50, 65, 82, 131 Error possibility 33–4

Gerken vii, 12, 28 Goldberg vii, 12, 73, 134–5, 150 Graham vii, 20, 70–2, 77 Greenough vii, 12, 30 Hazlett 33–4, 37 Hawthorne 12 Hills 142–3 Implicature 26–7, 33–4, 37–8, 124–5 Independence 5, 16–18, 32, 53–4, 57, 153–4 Inquiry 5–7, 116, 119, 125–6, 129, 145

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 19/3/2021, SPi

164  Justification, discursive 27–31 Justification, knowledge level 13, 25–6 Kelp 73–4, 76, 79, 151 KK 6, 32, 40–1, 48–51, 53–4 Knowledge, higher order 40–2, 44–5, 48, 53, 154 Knowledge norm of conjecture 113–18 Knowledge of knowledge 28–30, 40–2, 48–53 Luminosity 42, 44–5, 49, 53 Maxim 26–7, 34, 38, 123–4 Millikan 72, 74, 77, 97 Moore 7–8, 120–3, 131 Moran 132–3

Reasoning 16, 41, 50–1 Relevance 26–7, 33, 38–9, 44 Rysiew 33–4, 38 Sosa vii, 70–1 Speech, assertoric 14, 18, 57 Speech, constative 4, 6, 14–15, 18, 57, 87–8, 95, 103, 106, 131, 153, 155 Teleology 21–2 Truth norm 77 Turri 40–1, 48–9, 73–4, 106–7, 111–12, 115, 124, 131 Unassertability 8, 36–7, 39–41, 50, 154

Norm, prudential 29–30, 53 Norm, moral 63–4 Norm with epistemic content 65–6

Value 16, 19–32, 68–9, 153–4 Variability 7–8, 12, 18–20, 27–9, 32–3, 40–1, 50–1, 58, 153

Permissible assertion 8, 25–7, 58, 60, 68, 86–7, 154 Practice viii, 5–7, 50–1, 73–4, 77–9, 97–8, 104, 117–20, 125–9 Pragmatic encroachment 12, 16, 18

WAM viii, 19–20 Warranted assertability 19–20, 32, 40 Williamson vii–viii, 7–8, 21, 40–53, 82, 87, 95–6, 106–7