Democratic Speech in Divided Times [1 ed.] 2020951209, 9780198869757

Citation preview

Democratic Speech in Divided Times

Democratic Speech in Divided Times M A X I M E L E P OU T R E

1

1 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 © Maxime Lepoutre 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: 2020951209 ISBN 978–0–19–886975–7 DOI: 10.1093/oso/9780198869757.001.0001 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.

Acknowledgements In the course of writing this book, I have benefited from the advice and support of wonderful mentors. Thanks first of all to Clare Chambers and Rae Langton, who oversaw and vastly improved the first stage of this project as my doctoral advisers in Cambridge. They were (and continue to be) incredible role models, whose work and advice never fails to remind me of why I care about political philosophy. I am also deeply indebted to Cécile Laborde and Michael  P.  Lynch, who, in addition to providing support and sound counsel throughout, created the ideal intellectual environments—at Nuffield and UConn, respectively—for me to pursue this research. Sharon Krause and David Runciman, my doctoral examiners, gave thoughtful and generous feedback on my dissertation, and were the first to encourage me to turn it into a book. Cécile Fabre and Lynne Tirrell were always there to give invaluable advice, and talked me patiently through the process of writing a proposal. No one has had more impact on the substance of the book than Rob Simpson, whose even-­handed and open-­minded approach to philosophy first introduced me, and later guided me, through many of the issues that now constitute its foundation. For innumerable discussions and/or help on written drafts, thanks to Gabriele Badano, Sam Bagg, Heather Battaly, Teresa Bejan, Rachel Bernhard, Sam Berstler, Udit Bhatia, Paul Billingham, Chris Blake-­ Turner, Matteo Bonotti, Mariana Borges Martins da Silva, Matthias Brinkmann, Etienne Brown, Fabien Cante, Karamvir Chadha, Laura Caponetto, Charlie Crerar, Jennifer Daigle, Luke Davies, Fiona Doherty, Matt Dougherty, Cécile Fabre, John Filling, Rachel Fraser, Corrado Fumagalli, Alexander Greenberg, Steven Gubka, Michael Hannon, Reier Helle, Jeff Howard, Samuel Hughes, James Hutton, Stephen John, Jess Kaplan, Matt Kramer, Nikhil Krishnan, Hélène Landemore, Ruairí Maguire, Cathy Mason, David Miller, Alfred Moore, Zeynep Pamuk, Avia Pasternak, Allison and Juan Piñeros Glasscock, Alexander Prescott-­ Couch, Louise Richardson-­ Self, Katie Robertson, Elise Rouméas, Jay Ruckelshaus, Bernhard Salow, Paul Sagar, Ian Shapiro, Yves Sintomer, Shyane Siriwardena, Jason Stanley, Tamás Szigeti, Anthony Taylor, Jens Van ’t Klooster, Lani Watson, Ralph Weir, Andreas Wiedemann, Wes Wrigley, and Yuan Yuan. Among these, I owe special thanks to Cathy Mason, Matt Dougherty, Karamvir Chadha, Jess Kaplan, and Tamás Szigeti, who have provided steadfast intellectual and emotional support for the past five years. I am also very grateful to Dominic Byatt, my editor, for taking an interest in

vi Acknowledgements this project, for his encouragement, and for his responsiveness as I was writing it up. The thoughtful comments provided by three anonymous reviewers for Oxford University Press also greatly improved the manuscript. Thanks also to all of those who attended the Democratic Public Discourse graduate seminar I taught in 2020, as well as to audiences at the UCL Early Career Workshop in Applied Philosophy of Language, the Cambridge Political Philosophy Workshop, the Cambridge Forum for Legal and Political Philosophy, the Yale Moral Philosophy Working Group, the Cardiff Workshop on Religion, Hate, and Offence in a Changing World, the Princeton Workshop in Social Philosophy, the UConn Social Epistemology Working Group, the UConn Language and Power Workshop, the UConn Humanities Institute, the Oxford Ethics of Online Interaction Workshop, the Oxford Political Imaginaries and Models Conference, the UCL Early Career Workshop in Applied Philosophy of Language, and the Society for Applied Philosophy Conference. Some of the material in this book is adapted from, and expands upon, work that has previously appeared in a number of journal articles, and I am grateful to the publishers for allowing me to do so: • ‘Hate Speech in Public Discourse: A Pessimistic Defense of Counterspeech’. Social Theory and Practice 43, no. 4 (2017): 851–83. • ‘Rage Inside the Machine: Defending the Place of Anger in Democratic Speech’. Politics, Philosophy & Economics 17, no. 4 (2018): 398–426. • ‘Can “More Speech” Counter Ignorant Speech?’. Journal of Ethics and Social Philosophy 16, no. 3 (2019): 155–91. • ‘Democratic Group Cognition’. Philosophy & Public Affairs 48, no. 1 (2020): 40–78. I am also thankful for the various sources of financial support that made it possible for me to carry this project to fruition. Thanks to the Arts and Humanities Research Council, who provided a full scholarship for doctoral study and a Research Training Support Grant, as well as to Alistair Swiffen for helping me organise a study visit at Yale. Thanks to Nuffield College, whose warm and stimulating community I will sorely miss. Finally, parts of this project were generously supported by a grant from the John Templeton Foundation. All opinions expressed here are my own, and do not necessarily reflect those of these institutions. Finally, and most importantly, I couldn’t have completed this project without the perceptive criticism, sound advice, and unwavering encouragement of Joanna Demaree-­Cotton, and the endless support of my brothers (Adrien and Greg) and parents (Etienne and Marie-­Caroline).

Contents Introduction 1 PA RT I .   N O R M S O F D E M O C R AT IC P U B L IC D I S C OU R SE 1. Offering Shared Reasons

15

2. Voicing Anger

49

3. Countering Public Hate Speech

83

4. Countering Public Misinformation

106

PA RT I I .   P R E SU P P O SI T IO N S O F D E M O C R AT IC P U B L IC D I S C OU R SE 5. The Problem of Goodwill

131

6. The Problem of Political Ignorance

156

7. The Problem of Fragmentation

188

8. Resisting Political Minimalism

214

Conclusion

239

Bibliography Index

245 261

Introduction 1.  Democratic Discourse in Divided Settings Open and public speech is central to democracy. When we think of democracy, and why we care about it, among the first images that come to mind are images of such speech. We think, for example, of representatives passionately arguing for same-­sex couples’ right to marry. We think of whistle-­blowers exposing the corruption of political officials. And we think, similarly, of people taking to the streets to denounce, loudly and resolutely, wanton police violence against racial minorities. More generally, then, a well-­ functioning democratic system intuitively involves, at its core, inclusive public speech. It involves people from all walks of life bringing forward their concerns, revealing their experiences, denouncing unjust conventions, and demanding public justifications of others. Philosophers have long acknowledged this intuition in their normative theorizing. For over three decades, the dominant tradition of democratic theory—deliberative democracy—has placed inclusive processes of public speech at the heart of the democratic ideal.1 The universal and equal right to vote is of course an essential feature of democratic decision-­ making. But deliberative theories insist that the universal and equal right to demand justifications and to give reasons in the public sphere is no less important. All should have a meaningful opportunity to participate, directly or via their representatives, in public discourse aimed at influencing practical decision-­making. There are at least two reasons why inclusive public speech seems, to so many philosophers, to be so important. The first is epistemic: inclusive public communication, it is often argued, is instrumental to identifying good policies.2 Because different social groups regularly experience different constraints and opportunities, the knowledge needed to identify pressing social problems (such as racial injustices or public health crises), and to effectively counteract those problems, tends to be dispersed across different groups. As a result, getting people from different walks of life to engage with one another seems an effective way of pooling the knowledge that is required for good policymaking. 1  See, e.g., Manin (1987), Habermas (1996), Benhabib (1996), Elster (1997), Christiano (1997), Young (2000), Krause (2008), Talisse (2009), Anderson (2009), Dryzek (2010), Mansbridge et al. (2010), Forst (2012), Landemore (2013). 2  On the epistemic argument, see, e.g., Young (2000), Anderson (2006), Landemore (2013). Democratic Speech in Divided Times. Maxime Lepoutre, Oxford University Press (2021). © Maxime Lepoutre. DOI: 10.1093/oso/9780198869757.003.0001

2  Democratic Speech in Divided Times But inclusive public speech is not merely an epistemic tool: it also constitutes a bulwark against domination.3 To be dominated is to be subjected to arbitrary power—to power that is not forced to consider and track the concerns of those subjected to it. Now, public debate from which none are excluded and from which none can insulate themselves can make political power less arbitrary. This is because vocally challenging decision makers, or publicly demanding justifications of them, is a key mechanism for holding them accountable to one’s concerns. Hence, democratic speech does more than simply identify pressing social problems and solutions to those problems. In addition, it helps motivate powerful actors to take action against those problems. Yet, to many sceptics, this ideal of inclusive public speech seems tragically out of touch with the reality of political discourse.4 Perhaps public discourse can indeed help address injustices and other social problems when it is governed, as deliberative democratic theorists have often recommended, by norms of orderly argument,5 of mutual respect,6 and of sincerity.7 But this picture bears little resemblance to the public speech of contemporary democracies. In contemporary democracies, where different social groups are divided by mutually antagonistic attitudes, by patterns of spatial and social fragmentation, and of course by deep and unjust inequalities, political speech unsurprisingly takes a very different form. In place of orderly argument, we find a public sphere saturated with emotional appeals, including intensely negative emotions such as rage and resentment. Instead of mutual respect, public speakers routinely use their airtime to ridicule, demean, or vilify others. And where sincerity should reign, campaigns of misinformation instead proliferate unimpeded. Some might be unfazed by this mismatch between the ideal and the reality of democratic speech. The task of political philosophy, they might say, is to describe the way things ought to be, not the way they are. It is no surprise that the two do not coincide. But this response misses the point. Even for an ideal, there is such a thing as being too distant from reality. For one thing, a very distant ideal may lack the ability to guide action. If the ideal of democratic speech is too far removed from actual democratic speech, then it may be unclear how we are supposed to move from current discursive practices to better ones. Even more worryingly, distant ideals of democratic speech tell us very little about the value of democratic speech in the foreseeable future. From the fact that

3 On the domination argument, see, e.g., Young (2000), Forst (2012), Anderson (2010, ch. 5), Pettit (2012). 4  E.g., Posner (2003), Shapiro (2003), Green (2010), Parvin (2015), Somin (2016), Brennan (2016), Achen and Bartels (2016), and Bagg (2018a). 5  See, e.g., Cohen (1989), Elster (1998), and Habermas (1996). 6  See, e.g., Gutmann and Thompson (1996), Young (2000), Krause (2008), Mansbridge et al. (2010). 7  See, e.g., Habermas (1981), Schwartzman (2011).

Introduction  3 a very remote form of democratic speech would play valuable epistemic and ­anti-­domination functions, it does not follow that the kind of public speech we are likely to have any time soon can do so too. Indeed, the opposite might conceivably be the case: a public sphere saturated with expressions of rage, with vilifying or hateful stereotypes, and with campaigns of disinformation might seem more likely to bolster ignorance and existing patterns of domination than to elim­in­ ate them. Thus, even if inclusive public speech could in principle be a good thing, it may be hard to see why we should value democratic public discourse in divided contexts such as our own. In a searing indictment that captures this sceptical outlook, Richard Posner dismisses deliberative democracy as ‘a pipe dream hardly worth the attention of a serious person’.8 Many, like Posner, are tempted to jettison the deliberative ideal of democracy, in favour of political ideals that de-­emphasize inclusive public speech.9 But I believe that this is a mistake. History is replete with examples of inclusive public speech being used, with great effect, to tackle pressing social problems such as gender oppression and racial injustice. The suffragettes, the abolitionist movement, the civil rights movement, among many others, remind us of the power vested in the public speech of ordinary people, even in conditions marked by deep and unjust social divisions. We must not lose sight of this. Accordingly, what we must do instead is adjust our political ideals of inclusive public discourse, to make them more relevant to settings marked by grave social divisions. This is the project that this book will carry out. I will investigate what democratic public speech would have to be like, for it to be as close as possible to non-­ideal conditions—with their characteristic divisions and injustices—while remaining an effective tool for addressing pressing social problems. More specifically, I will develop an integrated account of the norms that should govern democratic speech in settings marked by deep social divisions. And my aim, in articulating these discursive norms, will be to show that we can navigate between idealism and pessimism about democratic public speech: when governed by the right norms, public discourse can play a key role in identifying injustices and other social problems, and in making political power responsive to those problems, even under conditions marked by grave and often unjust social divisions. What exactly is a discursive norm? A norm, as I am understanding this term, is a conventional rule. And by ‘discursive’, I simply mean ‘relating to speech or discourse’. So, a discursive norm is a conventional rule that regulates speech or discourse.10 To make this clearer, consider some of the norm-­related questions that I will be examining. When speakers offer justifications for their preferred 8  Posner (2003, 163). 9  See note 4. 10  I am using ‘speech’ and ‘discourse’ interchangeably.

4  Democratic Speech in Divided Times policies, must they appeal to reasons that are shared by their interlocutors? Do expressions of immoral or degrading views have any place in democratic discourse? And even supposing that a speaker’s public utterance expresses considerations that are morally appropriate and shared, should she nevertheless refrain from voicing those considerations with negative emotions, such as anger or resentment? These examples highlight two things. First, discursive norms can be more or less formal, depending on how strong the moral reasons underpinning them are. For instance, because degrading speech is commonly taken to engender deeply problematic harms, the salient question is whether it should be restricted by formal legal rules, which impose coercive sanctions on offenders. By contrast, hardly anyone suggests legally restricting the expression of profound anger per se, or legally prohibiting appeals to non-­shared reasons. Instead, the relevant norms in such debates tend to be informal rules, which are either unenforced or enforced less intrusively, for example through verbal expressions of disapproval. Second, the notion of a discursive norm is a broad one. Deliberation, understood as the give and take of reasons relating to practical matters, is an important part of public discourse. But it is only one of many things, good and bad, that public discourse can be used to do. Among many other things, public discourse can also be used to express emotions; it can be used to arouse emotions; it can be used to degrade or otherwise harm others; and it can be used, as we will see, to block the harms of degrading or otherwise harmful speech. In line with this broad view of public discourse, the norms I will consider in this book do not only consider what sorts of reasons can permissibly be offered in deliberation. They also examine the place of emotion in public discourse, and how we should respond to harmful public speech. Though the question of what norms should regulate democratic public discourse in non-­ideal contexts is pressing, democratic theorists have yet to address this question in a sufficiently sustained and integrated way. A natural place to look might be the vast literature on public reason. Public reason theorists have developed extremely detailed analyses of the kinds of reasons that are capable of legitimating political power. However, their primary focus in fact tends to differ from my own. In the first place, many public reason theorists simply supply standards for the public justifiability of political power, while maintaining that these standards need not constrain actual discursive processes of justification.11 Furthermore, public reason theorists who do supply standards for actual discursive processes—for instance, Rawlsian public reason theorists—tend to be primarily interested in ideal theory. That is, they investigate what norms should regulate public deliberation in contexts that are substantially better than

11  E.g., Vallier (2015, 155–6), Gaus and Vallier (2009, 65–6), Lister (2008, 281–2).

Introduction  5 actual.12 To be clear, this is not to say that public reason theory is irrelevant to my inquiry, or that its prescriptions are necessarily inappropriate in actual circumstances. Rather, the point is that it remains an open question—which I will explore—to what extent the detailed prescriptions public reason theorists make for ideal theory are also valuable in non-­ideal conditions. What about philosophical work on public discourse that concerns itself with non-­ideal conditions? On the one hand, much of it is driven by specific practical problems. Consequently, it tends not to be integrated with a broader account of democratic discourse. Consider, for example, examinations of whether legal restrictions should curtail degrading hate speech,13 and discussions of anger’s political role.14 As we will see, these examinations have developed crucial conceptual resources for understanding the moral concerns associated with hate speech and angry speech. However, they are commonly carried out in isolation from broader theories of democratic public discourse. What is lacking here is a systematic account of how the particular norms being defended should fit together with other norms regulating democratic discourse. This is a problem, because the value of any given discursive norm depends partly on what other discursive norms are operative. Whether it makes sense to require people to appeal to shared considerations depends partly, I will suggest, on the role allotted to emotionally charged forms of speech. Likewise, the value of welcoming expressions of anger depends partly on what mechanisms are in place to counter hateful abuses of angry speech. This complexity is lost when we focus exclusively on particular discursive norms. On the other hand, many democratic theorists concerned with non-­ideal conditions face the opposite problem: they defend the value of democratic discourse broadly construed in non-­ideal circumstances, but without closely examining specific discursive constraints. Most prominently, ‘agonistic democrats’ such as Iris Marion Young emphasize the key role played by inclusive speech in opposing unjust power structures. As such, agonistic theories of democracy are acutely sensitive to the grave power asymmetries and social divisions that characterize non-­ideal societies. But agonistic democrats seldom explore at great length what particular norms should govern the content of public discourse at great length. This is partly for principled reasons—namely, that they are often hostile to constraints on the content of democratic speech.15 Though my investigation is in many respects deeply indebted to Young’s democratic theory, I am not in principle hostile to such constraints, and therefore wish to examine them more closely.

12  E.g., Rawls (1997), Quong (2011, esp. 139, 143; 2017, section 8), Cohen (1989), Estlund (2008). 13  E.g., Waldron (2012). 14  E.g., Nussbaum (2016), Srinivasan (2018). 15  Young (2000). See also Mouffe (1999), Tully (2005), Schaap (2006), and Dryzek (2010). On Dryzek’s nuanced position, see note 61, Chapter 1.

6  Democratic Speech in Divided Times We should not overstate the point. Some democratic theorists have aimed to develop systematic accounts of democratic public discourse that offer guidance in non-­ ideal societies. This is particularly true of the increasingly influential ‘systemic’ approach to deliberative democracy. The systemic approach emphasizes how different real-­world arenas of public discussion, which are governed by different discursive norms, work collectively to produce democratic functions.16 Nevertheless, its proponents typically only offer cursory discussions of many of the key challenges that arise in the public discourse of divided democracies. They rarely discuss, for example, what should be done about public occurrences of hateful speech, how we should counteract public campaigns of misinformation, or what the distinctive place of anger in public discourse should be. So, while my argument is consistent with the systemic approach—and indeed, as we will see, positively works within it—it aims to speak more directly to the characteristic challenges of divided democracies. In light of this brief overview, we can better appreciate the desiderata that my account of democratic public discourse will seek to satisfy. I will articulate and defend a normative account (1) that explores what particular norms should regulate public discourse, (2) that focuses on non-­ideal political contexts, with their deep and often unjust social divisions, and (3) that gives a sense of how these particular norms should be integrated with one another. In the process of doing so, I will bring into closer dialogue the valuable strands of research cited above. My investigation will be divided into two main parts. The first half of the book (‘Norms of Democratic Public Discourse’) examines what kinds of democratic speech are desirable in divided settings. This part shows, first, how some features of public discourse that are symptomatic of social divisions can nevertheless play a positive and important democratic function. In particular, it explores the distinctive role of emotionally charged speech—most notably expressions of deep anger—in overcoming entrenched epistemic divisions. In virtue of performing this epistemic function, these forms of speech are needed to facilitate the mutual exchange of shared reasons, the contestation of political power, and the struggle against injustice (Chapters  1 and  2). Second, this part also examines how, in contrast, other characteristic features of the public discourse of divided societies endanger democratic life. Here, I will focus on the threats posed by the proliferation of public hate speech and misinformation. On the basis of this diagnosis, I articulate what forms of democratic speech should be used to counter these dangers (Chapters 3 and 4). Having thus outlined a unified picture of the norms that should govern non-­ ideal democratic public discourse, the second half of the book (‘Presuppositions 16  E.g., Krause (2008), Chambers (2009), Dryzek (2010), Mansbridge et al. (2012), Bächtiger and Parkinson (2019).

Introduction  7 of Democratic Public Discourse’) scrutinizes its realism. In other words, it examines how far the desirability of the recommended norms presupposes background conditions that are better than those that obtain in divided democracies. Specifically, Chapters 5 to 7 assess the extent to which my normative account of democratic speech is complicated by, respectively, different groups’ mutual dislike (or ‘affective polarization’); the fragmentation of the public sphere into insulated communities; and the fact that people often know very little about politics. The normative picture of democratic public discourse that this book defends can, I argue, largely withstand these problems, and thus remain action-­ guiding in divided political settings. What is more, while these social conditions do qualify the value of democratic speech in some respects, they are at least as problematic for normative political ideals that give up on inclusive democratic speech altogether (Chapter 8). Accordingly, while realizing the ideal of democratic speech I offer is challenging, we should not lose patience with this task.

2. Overview 2.1  Norms of Democratic Public Discourse As advertised, Chapters 1 to 4 investigate what norms should govern the content and form of democratic public speech in divided settings. Chapter  1 considers one of the most widely supported norms of public discourse—the ‘shared reasons’ constraint—according to which deliberators must appeal to considerations that are suitably shared when justifying their political demands. There are strong reasons to support this norm: requiring that deliberation be framed in terms of shared reasons facilitates the contestation of political power, and thereby keeps such power from being exercised arbitrarily. Nevertheless, this constraint tends to be overly exclusive. In divided democracies, different social groups have radically different experiences of the social context. As a result, even the weakest formulation of the shared reasons constraint ends up excluding important forms of political contestation from public discourse. What should we do to offset this exclusiveness? In the first place, we should restrict the shared reasons constraint to more formal arenas of public discussion, while waiving it in other arenas. But this popular recommendation is far from enough. More controversially, I argue that we must also welcome the presence of emotionally charged and non-­argumentative forms of speech in public discourse. These forms of speech have traditionally been viewed as antithetical to the shared reasons constraint. But the two should in fact operate hand-­in-­hand. Emotionally charged narrative, I suggest, constitutes a distinctively useful tool for communicating experiences between divided groups. In so doing, it enriches

8  Democratic Speech in Divided Times the pool of shared resources from which the shared reasons constraint proceeds, and thereby mitigates its exclusionary implications. Chapter  2 substantiates this recommendation by examining the role of anger and angry narratives, specifically, in democratic public discourse. It is commonly held, both in political philosophy and in actual political discourse, that people should refrain from publicly expressing intense anger, as doing so is counterproductive: that is, expressing intense anger at societal injustices tends to impede rather than advance efforts to eliminate those injustices. To resist this challenge, this chapter articulates a novel account of why publicly expressing anger is epistemically valuable. Specifically, by drawing attention to the ‘salience’ function of anger, and to its non-­conceptual content, I contend that expressing anger can play an indispensable role in alerting others to previously overlooked injustices, and in enhancing their understanding of these injustices. This specific epistemic role is vital in divided societies. Because divided societies typically involve significant social segregation and epistemically detrimental ideologies, the injustices endured by some groups are often invisible to, or misunderstood by, other groups. Chapter  3 turns to an important problem raised by the norms defended in Chapters 1 and 2: that inviting emotionally charged personal narratives, including intensely angry narratives, is too inclusive. In particular, doing so may open the door for public assertions of deeply disrespectful or hateful views. Such hate speech, which is rife in divided societies, constitutes a serious danger for democratic public discourse. Unless it is countered appropriately, it risks eroding the standing or dignity of its targets, and thereby preventing them from participating effectively in public discourse. It is often said that the dignitarian harm of hate speech cannot adequately be countered with more speech. As a result, many recommend introducing legal norms that coercively suppress hateful or degrading public speech. But I argue that this view in fact relies on an overly unsophisticated understanding of how we might counter hate speech with more speech. First, it overlooks the role that the state can play in endowing ‘counterspeech’ with authority. And, second, it overlooks the distinction between ‘negative’ counterspeech (which focuses on rejecting hateful perspectives) and ‘positive’ counterspeech (which instead affirms a countervailing perspective). Counterspeech that is both state-­sponsored and positively framed constitutes a prima facie preferable tool than legal norms for upholding the standing of targets of hate speech. Chapter  4 complements this discussion of hate speech by considering a different species of harmful speech: misinformation. Misinformation, such as fake news and conspiracy theories, threatens to undermine democratic public discourse by corrupting its epistemic and anti-­domination functions. When public discourse is saturated with misinformation, it risks generating ignorance

Introduction  9 about politics; and this ignorance, in turn, makes it more difficult for citizens to hold political power accountable. How should we respond to misinformation? Legal norms suppressing speech are, if anything, an even less promising response here than in the case of hate speech. Yet speech-­ based responses also face grave difficulties. The damage wrought by misinformation is often ‘sticky’, or difficult to reverse. In fact, drawing on social psychology and philosophy of language, I argue that countering misinformation with more speech runs the risk of amplifying the damaging effects of misinformation. The rest of the chapter examines what forms of verbal responses are best placed to overcome this ‘stickiness’. Here, the distinction between positive and negative counterspeech developed in Chapter 3 once more proves helpful: unlike negative counterspeech, positive counterspeech can respond to misinformation without triggering the properties that render it ‘sticky’. This insight in turn has two noteworthy upshots. First, it helps explain why (as social psychologists have noted) attempts at correcting misinformation sometimes work, and sometimes do not. And, second, it casts doubt on the increasingly popular practice of online fact-­checking. Still, engaging in positive counterspeech cannot be the full solution. This is, among other things, because positive counterspeech is unavailable for some forms of misinformation (most notably, conspiracy theories). To alleviate this and other limitations of positive counterspeech, I recommend adopting a diachronic conception of counterspeech. The task of verbally responding to misinformation should be a continuous process, which not only follows misinformation, but pre-­ empts it as well. Only if counterspeech takes such a diachronic form can we keep the most resilient forms of misinformation, such as conspiracy theories, from taking root in the first place.

2.2  Presuppositions of Democratic Public Discourse The discursive norms articulated in Chapters  1 to  4 aim to be relevant to, and capable of guiding action in, deeply divided societies. Yet one might worry that my account remains too idealistic, and presupposes social conditions that are significantly better than those that characterize contemporary divided democracies. Chapters  5 to  8 turn to this problem, and examine how the normative account of democratic speech I offer is complicated by the lack of goodwill between groups; the persistence of political ignorance; and the fragmentation of the public sphere. The first problem, which I raise in Chapter 5, is this: for inclusive democratic speech successfully to generate knowledge or hold policy makers accountable,

10  Democratic Speech in Divided Times people need to be willing to listen to, and engage with, one another. After all, the epistemic value of angry narratives and positive counterspeech will be lost unless their intended targets hear them and consider them seriously. But the contemporary trend towards affective polarization indicates that this is unlikely, because members of opposing political factions deeply dislike and distrust one another. This is a real problem. Yet I believe that we can alleviate it in two important ways. First, the lack of mutual goodwill appears less troubling once we appreciate the systemic character of public discourse—the fact that discourse takes place in a network composed of many arenas, with different constituencies, that are connected more or less closely to one another. The systemic form of public discourse means that not everyone needs to like and trust each other. One reason for this is that, if two social groups strongly dislike one another, their respective insights and concerns can be relayed between them via the very many intermediate arenas in the discursive system. Moreover, even to the extent that mutual dislike is troubling, features of public discourse that are widespread in divided conditions—namely hypocrisy, anger, and even the occurrence of hateful speech—have properties that can be harnessed to rebuild goodwill in divided societies. This might seem strongly counter-­ intuitive: prima facie, these features of public speech seem wholly at odds with trust and goodwill. But I argue that this commonly held intuition should be challenged. Importantly, moreover, the properties of non-­ ideal democratic discourse that help regenerate goodwill are consonant with the discursive norms that Chapters 1 to 4 recommend. Accordingly, this chapter lends further support to the normative picture of democratic public speech that I am offering. There is, however, a further problem. An influential narrative suggests that, even when people are willing to engage with one another in good faith, they may be insufficiently knowledgeable to do so competently. Survey after survey suggests that people are ignorant about basic political facts. What is worse, political scientists insist that this ignorance is robust: people seem dogmatically attached to the political commitments that they derive from their social group identities. So, not only do people seem insufficiently knowledgeable to distinguish sound public interventions from unsound ones, but they also appear resistant to the information and advice that would make them more capable of doing so. In Chapter 6, I argue that this influential concern misunderstands the epistemic function of social group identities. The reason it misunderstands this function is that it overlooks crucial insights from two areas of philosophy of science: feminist social epistemology, on the one hand; and philosophical investigations of value judgments in scientific testimony, on the other. The first body of research reveals that the experiences of constraint and enablement involved in being a member of a particular social group are epistemically relevant to deciding what politicians or policies one should support. As for the second, it indicates that—contrary to what

Introduction  11 the recent literature on ‘motivated reasoning’ contends—these group-­ based experiences are even epistemically relevant to assessments of scientific testimony. Hence, the influence of group identities on political judgments is not in and of itself a sign that people are robustly ignorant about politics. Still, one might worry that people give too much weight to the information associated with their social group identities, relative to other sources of evidence. From this perspective, the problem is not that people derive political commitments from their group identities, but that they are dogmatically attached to these commitments. However, this lingering concern overlooks another important lesson from the philosophy of science: that, while dogmatism may be irrational on an individual level, it can play an important role in the broader epistemic system. The analogy between political communities and scientific communities reveals that even if people are dogmatically attached to commitments they derive from their group identities, this needn’t be a problem for democratic public discourse. On the contrary, this dogmatism can play a fruitful epistemic role in the system of public discourse, provided information circulates widely between different discursive arenas, which are populated by different social groups. One implication of this analysis is that we should worry less about the influence of group identity on people’s political commitments, and more about the social and spatial fragmentation that prevents information from circulating between different arenas. Chapter 7 therefore tackles the problem of fragmentation head-­on. Opposing political groups live in different areas, occupy different professions, tune in to different news sources, and inhabit different social media ‘bubbles’. This is an enormous problem for the value of democratic speech. Indeed, the strategies I outlined above for defusing the problem of mutual dislike (Chapter  5) and the problem of ignorance (Chapter  6) both depend importantly on the wide circulation of insights and information across the system of public discourse. Yet social fragmentation threatens this circulation. To address this problem, Chapter 7 recommends ‘integrative’ policies aimed at desegregating opposing social and political factions. Integrative policies, however, remain deeply controversial. It is often held that making social groups live, work, or interact alongside one another is normatively inappropriate. For one thing, integrative policies may appear to violate people’s freedom of association. And, in particular, such policies seemingly overlook the important interest that members of historically oppressed social groups have in residential self-­segregation. Integrative measures, I suggest, in fact need not jeopardize these interests. To begin, integrative policies create meaningful associative opportunities even as they restrict others. Furthermore, keeping the systemic nature of public discourse in mind is again helpful: it reminds us, first, that integration can take place in many types of arenas, some of which are not residential; and it also reveals that residentially integrated arenas can serve as intermediaries between segregated

12  Democratic Speech in Divided Times residential arenas. Both insights highlight that the task of de-­fragmenting the public sphere, though costly, can leave space for historically oppressed groups to engage in some measure of residential self-­segregation. Chapters 5 to 7 suggest that combating the interrelated problems of political ignorance, mutual dislike, and fragmentation is central to the project of rehabilitating democratic speech in divided times. But, for all that I have said, doing so remains a deeply challenging and arduous task. Some philosophers and theorists might lose patience with this task, and insist that we should after all abandon the ideal of inclusive democratic public discourse. On this line of thought, we would do better, for the time being, to pursue a more minimalistic political ideal, democratic or non-­democratic, that does away with inclusive processes of public discussion. The book’s final substantive chapter (Chapter  8) demonstrates that this is a mistake. Even if widespread political ignorance (together with the mutual dislike and fragmentation that sustain it) is problematic for inclusive public discourse, it is just as problematic for the political models that are touted as alternatives. To begin, I argue that more minimalistic models of democracy—which eschew inclusive political reason-­giving or restrict it to small-­scale settings—neither avoid nor alleviate the problem. Because these proposals remain minimally democratic, they retain a bottleneck of inclusive decision-­making where political ignorance, ill will, and fragmentation generate difficulties. As for non-­democratic ideals such as epistocracy, they simply postpone these difficulties. Indeed, the difficulties at hand resurface when we consider how we are to transition to and maintain these non-­democratic systems. In this light, departing from the ideal of inclusive democratic discourse on the basis of the problems canvassed in Chapters 5 to 7 is misguided. Far from being the source of the problem, inclusive democratic speech is best understood as an attempted solution to the difficulties stemming from ignorance, ill will, and fragmentation. Given that these difficulties are ubiquitous, and given the promise of speech-­based innovations, we must try to make this solution work.

1

Offering Shared Reasons 1. Introduction Inclusive public speech is widely regarded as a normatively central component of democracy. There are, as we have seen in the Introduction, at least two reasons why this practice seems so valuable. The first is epistemic: having citizens exchange reasons for and against practical proposals is, many argue, an effective way to pool socially dispersed information that is relevant to good policymaking. The second concerns accountability: processes of inclusive reason-­giving enable citizens to challenge exercises of political power, so as to make that power more responsive to their concerns. In this manner, inclusive public discourse protects citizens from the condition of domination, or subjection to arbitrary rule. But not every kind of public discourse can realize these aims in equal measure. And, for any given kind of public discourse, its effectiveness at achieving these aims will vary based on the political context. The norms of public discourse that are appropriate in ideally just settings may be inappropriate in deeply divided and unequal societies. Accordingly, if we want a normative ideal of public discourse that can guide action in real-­world democracies, we need to address the following question: what rules should govern the content and form of public discourse in settings marked by deep and often hierarchical social divisions? This chapter, and the following three, are devoted to answering this question. The present chapter kicks off this investigation by examining one of the most widely defended discursive norms: the ‘shared reasons’ constraint, according to which participants should appeal only to reasons that are suitably shared.1 I will argue that the shared reasons constraint poses a dilemma in divided circumstances. On the one hand, this constraint plays a key role in facilitating the contestation of political power. So, if we wish to protect citizens from domination in divided settings, we have strong reasons to uphold the shared reasons constraint. At the same time, however, the shared reasons constraint risks regulating the content of public discourse in an overly restrictive way. Because of the epistemic divisions that plague contemporary democracies, requiring that debates be framed in terms of shared reasons risks excluding too many considerations (Section 2).

1  Following Scanlon (1998, 18), I understand reasons as considerations that purport to count in favour of something. Democratic Speech in Divided Times. Maxime Lepoutre, Oxford University Press (2021). © Maxime Lepoutre. DOI: 10.1093/oso/9780198869757.003.0002

16  Democratic Speech in Divided Times How can we dissolve this dilemma? In other words, how can we maintain the shared reasons constraint, while making it less objectionably exclusive? The solution is not, as is often thought, to abstract away from misinformed or misguided citizens when applying the shared reasons constraint (Section 3). Nor is it sufficient to restrict this constraint to select arenas of public discourse. While doing so is an important part of the solution, it is nevertheless far from enough (Section 4). We must therefore go further, and pair the shared reasons constraint with another discursive norm. Specifically, I propose that we can mitigate this constraint’s exclusionary implications by giving a greater role to emotionally charged and non-­argumentative forms of speech, such as narrative, in democratic public discourse. Thus, we can rehabilitate the shared reasons constraint in divided conditions, by harnessing the epistemic potential of a form of speech— emotionally charged and non-­argumentative speech—that is widespread in such conditions (Section 5). These conclusions, I will suggest, mark a rupture with existing accounts of what public discourse should look like under non-­ideal conditions. The shared reasons constraint has been a staple of theories that articulate norms for public deliberation occurring under more or less idealized conditions. But democratic theorists concerned with non-­ideal conditions have traditionally been suspicious of this constraint.2 Part of what is distinctive in my approach, then, is the suggestion that there are powerful reasons to try to maintain a version of the shared reasons constraint in non-­ideal conditions. The challenge then becomes how we can do so while minimizing the extent to which this constraint is overly exclusive. That is what this chapter—by considering how we should interpret the shared reasons constraint, and what norms should complement it—investigates.

2.  Sharing Reasons: A Dilemma 2.1  The Shared Reasons Constraint Most political philosophers agree that some norms should regulate the content of publicly admissible considerations. The shared reasons constraint constitutes one of the most widely defended such norms. Consider an initial statement:

2  E.g., Mouffe (1999, 755), Young (2000, 49), Talisse (2009, ch. 2), Schaap (2006, 263–4), Waldron (1999, ch. 7), and Dryzek (2010, 86–93). Now, some non-­ideal deliberative theorists do explicitly defend the shared reasons constraint. E.g., Williams (2000, 130), Gutmann and Thompson (1996, ch. 2), and Krause (2008, ch. 5). However, Williams’ discussion of non-­ideal public deliberation is largely programmatic. As for other non-­ideal theorists who embrace the shared reasons constraint, such as Krause, Sections 3–5 will engage with their more specific arguments and positions.

Offering Shared Reasons  17 The shared reasons constraint: In public deliberation, one must appeal to reasons that are shared by all participants when justifying one’s claims.

There is something immediately appealing about this norm. Consider, for instance, the case of judicial courts. The shared reasons constraint makes sense of the widely held intuition that judges should ideally try to frame their decisions in terms of a society’s constitutional values (such as, in the United States, equality, liberty, or the pursuit of happiness). Relatedly, the shared reasons constraint seems to capture the common view that, in religiously divided societies, courts should refrain as far as possible from appealing to contested religious doctrines in their deliberations. Yet, despite this intuitive appeal, how exactly we should understand the above formula is vigorously disputed. One crucial source of controversy concerns the constituency of the shared reasons constraint—that is, who exactly is included under ‘participants’. For now, I am interpreting ‘participants’ in a commonsensical way, such that it refers to the actual adult citizens who populate the democratic system. But many philosophers (most famously Rawlsian public reason theorists) disagree with this interpretation. I will return to this key issue in Section 3. Another important ambiguity concerns the precise site of the shared reasons constraint—that is, which specific forums of public deliberation this norm is meant to govern. Should all arenas comply with the shared reasons constraint? Or only some of them? And, if only some, which ones? Section 4 will be devoted to considering this question. In the meantime, let us focus on the terms ‘must’ and ‘shared’ in the above formula. The first thing to notice is that the shared reasons constraint should not be understood as a coercively enforced legal norm.3 Deliberators ‘must’ appeal to shared reasons in the sense that they are morally required to appeal to such considerations. But the moral considerations supporting this obligation do not warrant legally suppressing public utterances of non-­shared reasons. Nonetheless, lighter and more informal types of conventional enforcement may be warranted. In a parliamentary setting, for instance, this might manifest itself in having the presiding officer verbally rebuke speakers who appeal to non-­ shared considerations.4 The second observation has to do with what it means for a consideration to be ‘shared’ by participants. According to the most influential interpretation, which I am also inclined to favour, a reason is shared when it is acceptable to all—that is, when all participants can accept that reason or consideration as normatively forceful. In other words, on this view, as Sharon Krause expresses it, a reason is

3  Rawls (1997, 769).

4  Rawls (1997, 769).

18  Democratic Speech in Divided Times shared when it is a consideration that all participants ‘care about’, such that they see it as counting in favour of some actions or policies.5 To make matters more concrete, consider again the case of religion. The fact that the Ten Commandments denounce killing is not a widely accepted reason for outlawing murder. Indeed, it is not something that all care about, such that they see it as counting against killing. Accordingly, on the acceptability interpretation, the Ten Commandments should not publicly be used to justify a political demand.6 Notice what the acceptability interpretation does not say. First, this in­ter­ pret­ation does not entail that the claim one is defending based on acceptable ­reasons must itself be deemed acceptable by others. If I defend tax cuts for the wealthy by appealing to their well-­being, what matters is that you accept the consideration that well-­being is valuable. But you may nonetheless deny that tax cuts would best promote it. Second, the need to appeal to acceptable considerations does not necessarily prohibit appealing to one’s self-­interest rather than to the common good.7 I could, for instance, publicly oppose tax cuts by declaring that well-­being is valuable, and that they hinder my well-­being. Finally, the obligation to appeal to acceptable reasons does not altogether rule out publicly referring to considerations that others do not find acceptable. What it rules out is making the justification for one’s political demand depend on such a consideration.8 So, I could permissibly affirm that the Ten Commandments forbid murder, so long as this claim plays no substantial role in the justification I offer for outlawing murder. This acceptability interpretation is popular across different traditions of democratic theory. In their prominent account of deliberative democracy, for example, Amy Gutmann and Dennis Thompson enjoin deliberators to ‘appeal to reasons that are [. . .] mutually acceptable in content’.9 Similarly, in the critical theory tradition, Seyla Benhabib recommends using ‘discursive language that

5  Krause (2008, 151, 156). 6  Note that because I prefer (and will later defend in Section  3) a non-­idealized account of the constituency of the shared reasons constraint, religious reasons are only contingently excluded by the shared reasons constraint on my account. That is, if all actual adults came to agree on a particular religious worldview, then it would no longer be impermissible for (say) the Supreme Court to appeal to religious considerations when justifying its decisions. This result strikes me as intuitively correct. Having said that, this consequence is not relevant to my investigation here, since I am focusing on societies marked by deep social divisions, where members of society are overwhelmingly unlikely to agree on a particular religious worldview. 7  Mansbridge et al. (2010, 73–80). 8  See, for a similar observation, Rawls’ (1997, 784–7) famous ‘proviso’. 9  Gutmann and Thompson (1996, 57). See also Cohen (1996, 100), Williams (2000, 130), Krause (2008, 151, 156), Mansbridge et al. (2010, 66), Bächtiger and Parkinson (2019, 12–13), and Curato et al. (2019, 34–5). Bohman and Richardson (2009, 274) defend a related view, according to which shared reasons are mutually accepted.

Offering Shared Reasons  19 appeals to commonly shared and accepted public reasons’.10 Finally, this interpretation of ‘shared’ also features in Rawlsian public reason theory. Jonathan Quong affirms that, to comply with the Rawlsian version of the shared reasons constraint (the so-­called ‘duty of civility’) one must ‘offer a reason [. . .] that will be acceptable to everyone in their capacity as free and equal citizens’.11 I too am sympathetic to the acceptability interpretation, for reasons that will emerge in the following section. Yet there are other possible interpretations besides this one. In particular, some theorists choose to interpret ‘shared’ in a weaker fashion. On the weakest interpretation, a reason is shared when it is intelligible to all—that is, when everyone understands its meaning.12 Intelligibility is a precondition of acceptability: to genuinely accept a consideration, one must understand it. But the converse is clearly false. I can be acquainted with the Bible and understand appeals to particular divine commands, but nevertheless reject the view that these have normative force. Intelligibility is not the only alternative to acceptability as an interpretation of ‘shared’. Deliberative democrats have developed a wide range of competing interpretations, many of which occupy an intermediate position between acceptability and intelligibility. Gabriele Badano and Matteo Bonotti, for example, defend accessibility, which they take to be less permissive than intelligibility and more permissive than acceptability.13 This diversity of interpretations will nevertheless prove unproblematic for my present argument. In what follows, after arguing that the shared reasons constraint has considerable prima facie appeal (Section 2.2), I aim to diagnose a problem for this norm: even on the weakest understanding of what it means for a reason to be shared—intelligibility—the shared reasons constraint seems to exclude too much (Section 2.3). If this criticism holds for the weakest interpretation, then it will also hold for the stronger alternatives. The dilemma I articulate in what follows should therefore apply irrespective of how one chooses to interpret ‘shared’.

2.2  The Appeal of the Shared Reasons Constraint There are strong reasons to uphold the shared reasons constraint in divided settings, where different groups often have different normative outlooks and care about different things. This is because, in such conditions, this norm plays an important part in protecting subjects from domination. 10  Benhabib (1996, 83). See also Habermas (1996, 119). 11  Quong (2011, 261). See, similarly, Rawls (1997, 770–1). 12  Weinstock (2017, 647); Dowding (2018, 244); Vallier (2016); Gaus (2010, 279–83). 13  Badano and Bonotti (2020). Furthermore, there is diversity within interpretations of intelligibility: Vallier’s (2016) definition of intelligibility is somewhat stronger than the folk definition I give above.

20  Democratic Speech in Divided Times To be dominated, recall, is to be subjected to arbitrary or alien power. One is subjected to another’s arbitrary power when that other has ‘the capacity, not subject to [one’s] direct or indirect check, to interfere in [one’s] choice’.14 By contrast, one is protected from domination when the power to which one is subjected is checked, so that it must track one’s concerns and interests. How does the shared reasons constraint protect individuals from domination? One suggestion is that unless A (who is subjected to political power) appeals to reasons that are shared by B (who exercises that power), B will remain unmoved by her demands. Thus, to make political power accountable to her concerns and interests, and hence non-­arbitrary, A must appeal to considerations that are shared by those wielding power. In this vein, Krause observes that Martin Luther King Jr.’s demands for racial equality ‘became convincing to white Americans [. . .] because they appealed (in a logical fashion) to things most Americans really cared about: security, opportunity, and dignity, [. . .] liberty and equality’.15 Krause is right that appealing to shared considerations contributes to effective political contestation (and therefore, to non-­ domination). But the above argument falls short of establishing that conclusion. As stated, it simply shows that A should appeal to values that B cares about. But this strictly speaking, does not require that A herself share those reasons. To see why A and B should appeal to reasons that they both share, we must therefore say more. Public deliberation involves the exchange of justifications over time. Accordingly, when subjects make claims relating to an exercise of power, they typically do not do so in a vacuum. Rather, they are often responding to whatever justifications have been given for that power in an ongoing debate. Indeed, citizens often hold political agents accountable by poking holes in the justifications they have offered. Now, one’s ability to respond to a justification is substantially enhanced when that justification was made using a reason one shares. This is most obvious with intelligibility. If I do not understand the reason that is given to me, it is unclear how I could even begin to respond to it. If a theocratic government condemns me on the basis of its esoteric religion, and I do not understand many of its concepts, my ability to contest and check that condemnation is severely impaired. To the extent that this is the case, I am subjected to arbitrary power. But we should arguably go further than mere intelligibility. A similar observation holds with acceptability. Even when I understand the terms of the debate, I may be unable adequately to respond to certain charges because I feel that the debate has been severely misframed—framed using reasons I reject and consider to be deeply misguided. Indeed, if I feel alienated from the reasons that have been advanced, it may be impossible for me to articulate the points I feel are 14  Lovett and Pettit (2009, 14).

15  Krause (2008, 164).

Offering Shared Reasons  21 genuinely important in terms of those considerations. And if I cannot articulate the points I care most about, my ability to contest political power is thereby diminished. To appreciate this, imagine a scenario where a political decision maker justifies a policy in terms that are intelligible, but wholeheartedly rejected by a political subject. Suppose, for example, that the government rejects same-­sex marriage by alleging that same-­sex unions are unnatural. A member of the opposition may well think that it is simply irrelevant whether or not same-­sex unions are ‘natural’. Naturalness, she might think, is not a normatively significant category. But this leaves her in an awkward position. Either she attempts to respond to the charge of unnaturalness, which is likely to distort the considerations she finds genuinely significant; or she does not, but then she leaves unanswered the justification that those in power find most compelling. Hence, because the power she is subjected to is justified in terms she profoundly rejects, she is unable to force it to track the considerations she deems fundamentally important. Or consider, similarly, a theocratic government that implements racially discriminatory policies based on the claim that racial equality is unholy. I may understand perfectly well what it means for something to be unholy. Yet, given that I attach no normative force to holiness, my options for contesting this exercise of power are seriously impoverished. If I do not respond to the charge of unholiness on its own terms, then I again leave unaddressed the considerations that those in power care most about. On the other hand, it may be impossible for me to express what I take to be wrong about racial discrimination within the constraints of the state’s religion. After all, it might be true that, by the lights of this religion, racial equality is unholy. In sum, the argument for the shared reasons constraint is not simply that giving reasons that one’s interlocutors understand and accept contributes to making them more responsive to one’s concerns. It is also that when one is given reasons one both understands and accepts as a justification for political power, one is better able to hold that power accountable and thereby make it less dominating. And to ensure that, in a debate where socially and ethically divided participants exchange reasons over time, all participants are given reasons they understand and accept, the debate must be framed in terms of reasons acceptable and intelligible to all. That is what the shared reasons constraint aims to guarantee. Thus, there are powerful pro tanto reasons to uphold the shared reasons constraint—if we do not, we risk exposing subjects to significant forms of domination. This justification is nevertheless defeasible: though powerful, the reasons for upholding the shared reasons constraint could in principle be outweighed by countervailing considerations. Having provided a defeasible case for this norm, then, I will now introduce what I take to be the strongest consideration against it.

22  Democratic Speech in Divided Times

2.3  The Over-­exclusiveness of the Shared Reasons Constraint The most daunting problem with the shared reasons constraint is that it risks being overly exclusive, or ruling out too many considerations. This is especially evident on the strong interpretation of ‘shared’ that I have been emphasizing, according to which shared reasons are acceptable to all participants. Given non-­ideal societies’ deep divisions, it is likely that for most moral considerations, someone rejects them.16 Homophobes may find same-­sex unions worthless and unnatural. White supremacists might judge that racial minorities are innately inferior. And some ideological doctrines, religious and non-­religious, assign women to a subordinate role. It seems a deeply unwelcome result that, according to the shared reasons constraint, we should therefore refrain from appealing to the values of loving same-­sex relationships, of racial equality, and of gender parity as reasons for or against policies. One might simply take this as grounds for reverting to the weaker shared reasons constraint, which only requires that reasons be intelligible to all. But there are two problems with this suggestion. We have already encountered the first. In the previous section, I suggested that when a debate is framed in terms that one wholeheartedly rejects, this can seriously diminish one’s ability to contest political power—even if one understands those terms. So reverting to a mere ‘intelligibility’ requirement to avoid excluding too many considerations comes at a cost in terms of domination. But there is a second and more fundamental problem. The problem is that, contrary to what is often assumed, ‘over-­exclusiveness’ in fact also plagues the weak shared reasons constraint. Contemporary work in social epistemology brings this out particularly clearly. Social epistemologists have extensively explored and theorized the obstacles that prevent relatively disempowered groups from making important group-­specific concerns intelligible to others.17 According to these theorists, different social groups, who occupy different social positions based on their race, gender, class, and so on, systematically experience different constraints and enablements. Consequently, they may be radically ig­nor­ ant of one another’s experiences and perspectives. This is particularly true, social epistemologists suggest, with respect to what privileged groups know about less privileged groups. For example, Charles Mills and Elizabeth Anderson both argue that because of wide-­ranging segregation between white and black Americans, white Americans typically encounter very different facets of American society than

16  For a similar point, see Williams (2000, 133–4), Talisse (2009, 52–6), Young (2000, 40–1), Enoch (2015, 118). 17  Fricker (2007); Dotson (2011); Mills (2007); Anderson (2010).

Offering Shared Reasons  23 black Americans do. In consequence, significant portions of black Americans’ daily lives remain invisible to whites.18 These deep experiential divides, Miranda Fricker suggests, generate a dis­tinct­ ive epistemic problem for disadvantaged social groups. In highly unequal so­ci­ eties, disempowered groups typically have less influence over the creation and diffusion of social meanings. Consequently, the pool of conceptual resources a society has for conferring meaning on activities and experiences ends up being especially adapted to dominant groups’ experiences. The upshot is that disadvantaged groups struggle to make sense of and to articulate their group-­ specific experiences: ‘some significant area of [their] social experience [is] obscured from collective understanding’. Fricker calls this unjust deficit of intelligibility a ‘hermeneutical injustice’.19 In extreme cases of hermeneutical injustice, marginalized groups themselves lack the conceptual resources needed to make sense of their own experiences and concerns. Here, Fricker cites the case of sexual harassment. Before the concept ‘sexual harassment’ was developed in the 1970s, those experiencing it lacked the categories needed to fully understand the wrong they were being subjected to. Their experiences had to be conceptualized as mere instances of ‘flirting’ or ‘joking’.20 By contrast, in the more common case of hermeneutical injustice, although victims of hermeneutical injustice possess the concepts needed to make sense of their own experiences, some of their interlocutors do not.21 When this is the case, requiring that deliberators appeal to mutually intelligible considerations—as the weak shared reasons constraint does—keeps victims of hermeneutical injustice from appealing to concepts that are needed adequately to express important considerations or concerns. Consider a few examples. Micro-­aggressions, Kristie Dotson explains, are brief and commonplace slights and indignities directed at members of stigmatized groups, often automatically and unintentionally.22 They include, for instance, a white person instinctively checking that she still has her wallet when a black person walks by. When these behaviours are systematic, and when they cumulatively tend to express a denigrating stereotype, they are reasonably experienced with indignation.23 Now, to someone who never experiences such behaviours, it might be puzzling how very slight actions could ever amount to more than a trivial offence. In other words, someone in this position could fail to grasp the very concept of a micro-­ aggression. Just as many pre-­ Leibnizian mathematicians were unable to conceive of how a sum of infinitesimals could amount to a finite number, such a person might be unable to conceive of how 18  Mills (2007); Anderson (2010, 44–50). 19  Fricker (2007, 158). 20  Fricker (2007, 152–3). 21  Fricker (2013, 1319–20). 22  Dotson (2011, 246). 23  See McTernan (2018) for discussion of the injustice enacted by micro-­aggressions.

24  Democratic Speech in Divided Times many involuntary slights could cumulatively compose something as morally problematic as an aggression. Accordingly, members of privileged groups might (and sometimes do) think that denouncing behaviours as micro-­aggressions is misguided, and that those who do so are simply overly sensitive.24 In such instances, a public justification invoking the concept ‘micro-­aggression’ is ruled out by even a weak shared reasons constraint, because the concept is unintelligible to some participants. Similarly, Fricker cites marital rape as an instance of hermeneutical injustice.25 Many states did not recognize marital rape as a crime until very recently. Some still do not. Importantly, what made criminalizing marital rape difficult was partly the concept’s perceived unintelligibility. To some, marital rape seemed a contradiction in terms, because rape presupposes a lack of consent, whereas marriage (as they conceived it) produces the relevant consent by definition. And indeed, even where marital rape is criminalized, public pronouncements denying the coherence of this concept remain common.26 Here again, certain extremely significant considerations are not even weakly shared because they invoke concepts that are unintelligible to some. Finally, and returning to same-­sex marriage, the issue is not merely that same-­ sex relationships are not recognized as valuable. It is also that what they are can be fundamentally misunderstood. Homosexuality was once classified by hermeneutical authorities as an illness or disorder. In the US, for instance, it remained in the Diagnostic and Statistical Manual of Mental Disorders (DSM) until 1973. And witness, even today, the tragically common amalgamation of homosexuality and paedophilia.27 In these cases, in addition to the value of same-­ sex relationships being denied, what same-­ sex relationships are is itself fundamentally not understood by some. Consequently, both strong and weak interpretations of the shared reasons constraint seem over-­exclusive. Social epistemologists have shown that when members of privileged groups are profoundly ignorant of vulnerable groups’ daily experiences, they are liable not to understand important considerations raised by members of such groups. Often, as the phenomenon of hermeneutical injustice suggests, this is because they do not grasp crucial concepts that figure in these considerations. When this happens, even the weak shared reasons constraint prohibits appealing to such considerations. But some of these excluded reasons, including those discussed above, intuitively should have a place in public discourse. Excluding them silences marginalized groups, and thereby prevents them from contesting an unjust status quo. More precisely: excluding the important reasons in question undercuts public contestation’s efficacy with

24  E.g., Lukianoff and Haidt (2015), Campbell and Manning (2015). 26  E.g., Rothman (2015). 27  E.g., Walker (2014).

25  Fricker (2007, 155).

Offering Shared Reasons  25 respect to the policy issue on which the reasons were taken to bear. What this shows us is that there are powerful reasons to resist the shared reasons constraint grounded in concerns about domination. The result is a dilemma. Pro tanto reasons grounded in non-­domination support the shared reasons constraint in some respects, and pull against it in others. On the one hand, receiving justifications in terms of non-­shared reasons reduces individuals’ ability to contest exercises of power (Section 2.2). On the other, requiring that individuals appeal to shared reasons holds the political agency of vulnerable participants hostage to the ignorance or ill-­will of others (Section 2.3). Either way, it seems we are bound to fail to satisfy important domination-­related reasons. The rest of this chapter is devoted to assessing how we might resolve this dilemma. Specifically, I will examine how we might maintain the shared reasons constraint, while offsetting its exclusiveness. In evaluating different strategies for doing so, the question to bear in mind is therefore the following. How can we avoid making public deliberation overly exclusive, while responding as far as possible to the initial reasons for adopting the shared reasons constraint? How, put differently, can we best weaken the trade-­off between competing reasons grounded in the value of non-­domination?

3.  The Constituency of the Shared Reasons Constraint To make the shared reasons constraint less exclusive, one common strategy— which is typically associated with public reason theory—proposes that we idealize its constituency. The shared reasons constraint, recall, says that deliberators must appeal to reasons that are shared by all participants when publicly justifying their political demands. So far, I have been assuming that ‘participants’ refers to all actual adult citizens in the democratic system. But public reason theorists characteristically adopt a more restricted understanding of the shared reasons constraint’s constituency, by taking ‘participants’ to refer to idealized persons. Participants can be idealized in two ways: epistemically, when they are better at reasoning and responding to evidence than actual participants; and morally, when they have better values and motivations than actual participants. Note that although the parties are idealized, some subset of actual participants may satisfy the idealizing conditions, while others may not. Often the former are labelled ‘reasonable’, and the latter ‘unreasonable’.28

28  For overviews, see Quong (2017, section 3) and Billingham and Taylor (2020).

26  Democratic Speech in Divided Times This strategy is most clearly exemplified by public reason theorists working in the Rawlsian tradition. Rawls, for instance, claims that the reasons given when justifying coercive power must be ones that ‘free and equal citizen[s] might reasonably accept’.29 To be reasonable involves, first, a moral idealization. Reasonable participants view others as free and equal citizens and are therefore prepared to offer them fair terms of cooperation. Moreover, because reasonable citizens recognize the ‘burdens of judgment’—the many hazards that inevitably confront human reasoning—they do not expect or demand that others hold ethical views identical to their own.30 There is also some evidence that Rawls’s account of reasonableness involves an epistemic idealization: reasonableness, Rawls sometimes seems to suggest, requires upholding views that are basically coherent and responsive to new evidence.31 That being said, many political theorists working outside of the Rawlsian tradition adopt a similar strategy. Krause, for example, suggests that the test for whether a consideration is admissible in public deliberation is not whether it is endorsed by all actual citizens, but rather whether it could ‘be endorsed from within a suitably structured moral standpoint’. What makes concerns ‘endorsable in this context is that they can be affirmed [. . .] from within a deliberative perspective that manifests equal respect’.32 Thus, though Rawls’s account is the most influential, different theorists have offered many competing accounts of what the right level of idealization is (and by extension, of what exactly constitutes reasonableness).33 Consequently, the arguments that follow will not depend on the specific moral and epistemic conditions Rawls builds into his idealization. Instead, the moral costs that I will argue are associated with idealizing the participants who constitute the shared reasons constraint’s constituency, should apply to the extent that we idealize them. Restricting the shared reasons constraint’s constituency to idealized parties seems prima facie helpful when it comes to addressing the over-­exclusiveness problem. By bracketing ‘unreasonable’ participants who are insensitive to important descriptive or normative considerations, idealization prevents such actual persons from ruling out appeals to these considerations. For example, because this idealized shared reasons constraint abstracts away from those who fail to understand or accept the value of same-­sex relationships, it allows the value of same-­sex relationships to play a role in publicly justifying same-­sex marriage. So, this strategy prevents the content of public deliberation from being held hostage by the perspectives of unreasonable citizens.

29  Rawls (1997, 770), emphasis added. 30  Rawls (1997). 31  Rawls (1993, 58–9). See Nussbaum’s (2011, 24–5) discussion. 32 Krause (2008, 163). Additionally, see Estlund (2008, 4) and Gutmann and Thompson (1996, 55–6). 33  For discussion of this diversity, see Vallier (2011, 371–2) and Quong (2017, section 3).

Offering Shared Reasons  27 However, restricting the constituency to reasonable persons is problematic in two respects. The first worry concerns the fate of unreasonable people. Some theorists express the concern that excluding unreasonable persons from the shared reasons constraint’s constituency means stripping them of their basic political status: unreasonable persons, Marilyn Friedman suggests, ‘will be treated like the bearers of a pestilence [. . .] They will be denied the full protection of the system’s basic rights and liberties, particularly freedom of expression.’34 Friedman’s articulation of this initial worry rests on a misinterpretation. Excluding unreasonable persons from the shared reasons constraint’s constituency does not mean excluding them from the political community of citizens altogether. Actual unreasonable citizens are supposed to retain their basic legal and political rights, including the rights to vote, to stand for office, and even to publicly express unreasonable views.35 Indeed, recall that the shared reasons constraint is a moral norm, not a legal one. So, contra Friedman, it does not necessarily require that anyone be legally prevented from expressing their views. But there is nevertheless a worry for unreasonable citizens. Although they are not formally excluded from the democratic political community, they are discounted in the theoretical process of determining what qualifies as a shared reason. And this, in turn, has repercussions for actual deliberative politics. Because they are discounted, some reasons that are not actually shared by unreasonable citizens will nevertheless count as shared. Citizens who embrace these not-­actually-­shared reasons will therefore be welcome, morally speaking, to appeal to them when justifying their political demands. Hence, to the extent that the shared reasons constraint is realized, public deliberation will partly be framed in terms that unreasonable citizens do not accept and eventually do not understand. Moreover, although the shared reasons constraint is not a legal norm, we saw in Section 2.1 that the moral reasons underpinning it may warrant lighter, non-­ coercive ways of upholding it. This may involve, say, ignoring the public complaints of unreasonable citizens, should they object that they cannot accept or understand some reasons that are being publicly appealed to. Indeed, such complaints are irrelevant if unreasonable citizens are not part of the constituency for whom considerations must be shared. What this shows is that, even if restricting the constituency does not strip unreasonable citizens of formal citizenship rights, doing so nonetheless imposes a moral cost on them. Section 2.2 argued that the shared reasons constraint is pro tanto desirable because it advances the non-­domination of individuals who are subjected to power. Requiring that coercive power be justified using shared reasons enhances subjects’ capacity to contest that power. But consider now the 34  Friedman (2000, 23).

35  Quong (2011, ch. 10).

28  Democratic Speech in Divided Times position of unreasonable parties. Some, perhaps many, of the reasons raised in deliberative justifications are now unacceptable and eventually unintelligible to them. Consequently, they are relatively disempowered with respect to these justifications. Moreover, their objection to the presence of these reasons may be ignored as irrelevant. Therefore, it seems that the shared reasons constraint no longer protects them from domination. Insofar as this is so, idealizing the shared reasons constraint’s constituency undercuts the original rationale for that constraint.36 Admittedly, this initial problem does not entirely preclude idealizing the shared reasons constraint’s constituency. As Section 2.3 discussed, the over-­exclusiveness resulting from the non-­idealized shared reasons constraint in conditions where some citizens are ignorant or immoral (the unreasonable) exposes other citizens to domination (those whose legitimate concerns are misunderstood or rejected). It may well be that sometimes, protecting the latter from domination warrants, all things considered, subjecting the former to domination. This first concern, then, does not imply that restricting the constituency necessarily generates insurmountable moral costs.37 Nevertheless, it does show that this proposal is problematic for my purposes. My aim, remember, is to avoid or at least soften the dilemma raised by the shared reasons constraint. This, I ­suggested when concluding Section 2, means weakening the trade-­off between dominating those whose legitimate concerns are not publicly recognized, and dominating those who do not recognize those concerns. But restricting the constituency does nothing to weaken this trade-­off. Instead, it suggests how, given this trade-­off, we can minimize domination. Accordingly, it should be a last resort, introduced only once we have exhausted alternative solutions aimed at weakening the trade-­off. Idealizing the shared reasons constraint’s constituency generates a second, more serious, problem relating to non-­ideal conditions. Existing democracies are non-­ideal not just in that some individuals have unreasonable views, but also in that there is often a mismatch between who is right and who has power. Often, the unreasonable are also powerful. This, to reiterate, is a recurrent theme of contemporary social epistemology: as the examples of marital rape, homophobia, sexual harassment, and micro-­aggressions indicated, it is often the experiences and concerns of vulnerable groups that are rejected and misunderstood, and it is dominant groups who reject and misunderstand them. Public reason theorists sometimes obscure this point by focusing on cases where the unreasonable are

36  For a similar point, see Enoch (2015). 37  One might think that domination is not morally costly at all when it affects unreasonable citizens. I follow Enoch (2015, 123) and Raz (1998, 32–4) in resisting this view. Other things being equal, a world where we can protect vulnerable groups without dominating the unreasonable—say, by convincing them instead—seems morally better.

Offering Shared Reasons  29 marginal—for instance, religious extremists such as the Westboro Baptist Church.38 If the unreasonable are also powerful, however, idealizing the shared reasons constraint’s constituency is problematic. Such a norm welcomes debates that are framed in terms that unreasonable citizens may neither accept nor understand, and recommends ignoring their complaints concerning this framing. But this seems ill-­advised if they are dominant. If we fail to engage with the objections or concerns of powerful agents, we are unlikely to achieve political change. As mentioned earlier, King’s discourse was effective partly because, rather than ignore the concerns of the (often bigoted) white majority, he tried to speak a language they were responsive to. Similarly, Abraham Lincoln often appealed to legal grounds for ending slavery precisely because many of his Congressional interlocutors were unresponsive to the moral case. Hence, restricting the shared reasons constraint’s constituency jeopardizes one of the prime reasons for advocating inclusive discourse. This process helps hold the powerful accountable. By demanding justifications for their decisions, and publicly scrutinizing those justifications, weaker parties can publicly shame the powerful and thereby exercise some control over them. But if weaker agents frame their claims in terms of considerations that powerful agents are unresponsive to, or simply ignore powerful agents, public deliberation cannot serve this purpose. Hence, it is not just the non-­domination of unreasonable citizens that is problematic for the present proposal. It is also the non-­domination of vulnerable groups. Some public reason theorists might reply that not addressing this second worry is not an oversight. Quite simply, they are interested in theorizing public reason for ideal rather than non-­ideal conditions. Quong explicitly suggests as much when defending Rawls’s version of the shared reasons constraint, the duty of civility. He remarks that the idea of public reason and the duty of civility belong to an ideal of how things ought to be in a democratic society, assuming citizens as just and society as well-­ ordered [. . .] Pointing out that adhering to the duty of civility may be counterproductive under current conditions is thus no objection to the ideal, since current democratic societies are far from ideal, or well-­ordered.39

While this response might shield Rawlsian public reason theorists and other ideal theorists from my criticism, it does so only by making their topic of inquiry

38  E.g., Blake (2005, 233), Gutmann and Thompson (1996, 63–9). 39  Quong (2013, 270; 2017, section 8.1). See also Estlund (2008, 19) and Rawls (1997, 766n3). As Enoch (2015, 125n37) observes, however, Quong (2011, 158–9) occasionally speaks as if public reason norms applied in non-­ideal conditions.

30  Democratic Speech in Divided Times distinct from my own. So, it does not undermine my point that, in non-­ideal theory, restricting the shared reasons constraint’s constituency generates significant problems. Even if those authors were not concerned with non-­ideal theory, it remained an open question (which Quong acknowledges to be ‘hugely important’ and underexplored40) whether their recommendation could also be prescriptive there. I have reached the conclusion that in important respects, it is not. In summary, the proposal that we restrict the shared reasons constraint’s constituency to idealized or ‘reasonable’ participants faces two serious limitations when it is applied to public deliberation taking place in non-­ideal conditions. First, the proposal is insensitive to non-­ ideal contexts of power: when unreasonable parties are also powerful, the idealized shared reasons constraint jeopardizes the non-­domination of vulnerable groups. Where the unreasonable are not powerful, the proposal might seem more viable, as they can safely be ignored. However, even in such conditions restricting the shared reasons constraint’s constituency seems problematic: it risks subjecting unreasonable citizens to domination. While this might sometimes be a necessary cost to protect vulnerable groups from domination, we should only accept this solution as a last resort. This is because it does not weaken the trade-­ off between the non-­ domination of the unreasonable and of the reasonable. Rather, it simply suggests what the best decision is given that the trade-­off exists. Prior to idealizing the shared reasons constraint’s constituency, we must therefore investigate whether there are ways of weakening or dissolving this trade-­ off, and the resulting dilemma.

4.  The Site of the Shared Reasons Constraint The first step to weakening the dilemma at hand is to distinguish between different sites in the ‘system’ of public discourse, and to stipulate that only some of them should be governed by the shared reasons constraint.

4.1  The System of Public Discourse This solution starts from what has come to be known as the ‘systemic’ understanding of public discourse.41 Because this conception of public discourse 40  Quong (2013, 270). 41  A terminological note is in order. Advocates of the systemic approach usually refer to it as the systemic view of public deliberation. I prefer referring to the system of public discourse. This is because, as I explained in the Introduction, this book is concerned with the norms that should govern public discourse, and public discourse is a broader category than public deliberation. But nothing of

Offering Shared Reasons  31 will play a substantial role throughout my inquiry, it is worth briefly pausing to elaborate on what it involves. According to the systemic understanding of public discourse, we should not imagine democratic public discourse as involving a single forum, which aspires to include all citizens. Instead, we should see public discourse as a whole system composed of many smaller parts—that is, composed of many smaller arenas or sites of public discussion.42 These different sites are hugely varied.43 They include, of course, formal public arenas of deliberation, such as legislatures and judicial courts. But they also include an immense array of other, often less formal, sites, such as party conferences, newspaper opinion sections, radio talk shows, classrooms, church gatherings, barbershops, consciousness-­raising groups, social media feeds, town hall meetings, citizen juries, committees of scientific experts, as well as protests. In a well-­functioning system, these different sites complement one another by performing different functions.44 On the one hand, formal arenas such as legislatures and courts typically deliberate with a view to producing coercively enforced decisions. By contrast, more informal arenas typically do not issue binding decisions. Instead, they characteristically help voters form more considered opinions regarding policies, politicians, and parties. And they do so in various ways: some give voters the opportunity to express their concerns and to listen to those of their peers; others monitor formal decision-­making arenas and report their observations to voters; others still disseminate policy-­ related expertise to laypeople; and so on. The fact that different arenas perform different and complementary functions has several important implications. The first is that it is not necessary for each arena to satisfy all the desiderata of democratic public discourse. What matters is rather that, taken collectively, the different arenas that constitute the system produce outcomes that are epistemically sound and accountable to the concerns of the people. Thus, it is not necessarily a problem that many arenas—such as feminist consciousness-­raising groups or partisan media outlets—exclude some groups and their perspectives.45 These exclusions are acceptable so long as the excluded groups have meaningful opportunities to articulate their concerns in substance should hinge on this. As we will see below, the systemic view of deliberation in fact relaxes the requirements of deliberation in many arenas of public speech, as a result of which it welcomes non-­deliberative forms of public discourse in many sites. This, however, is another reason why it seems to me more natural to refer to the system of public discourse, than to the system of public deliberation. 42  See, e.g., Mansbridge (1999a), Habermas (1996), Goodin (2005), Krause (2008), Dryzek (2010), Mansbridge et al. (2012), Chambers (2017), Curato et al. (2019, ch. 4), Hertzberg (2018), and Bächtiger and Parkinson (2019, chs. 5–6). 43 On the variety of different sites, see Mansbridge et al. (2012) and Bächtiger and Parkinson (2019, ch. 5). 44  Mansbridge et al. (2012). 45  See, e.g., Mansbridge et al. (2012) and Curato et al. (2019, ch. 4).

32  Democratic Speech in Divided Times other parts of the system. And, in fact, exclusions such as these may actually play a positive role, by facilitating the articulation of distinctive perspectives within different arenas. This will prove important in Chapter 2, when we assess the place of anger in political discourse. The second implication is related. For different arenas, which perform different functions, to complement one another effectively, they need to be sufficiently closely connected to one another.46 For instance, for formal decision-­making arenas to be responsive to the concerns of the people, the concerns articulated in informal arenas must be able to reach them. Now, what exactly counts as a sufficiently close connection depends on context, and is a question I will return to in subsequent chapters.47 In the meantime, the essential point is simply that insights and arguments generated in different arenas must circulate between them over time. The final implication—and the one that will matter most for the purposes of this chapter—is that, because different discursive arenas play different functions, they need not be governed by the same discursive norms.48 The discursive norms that are appropriate for the purposes of producing high-­stakes decisions may differ from the norms that best facilitate public opinion formation. This is something that we already recognize in everyday practice: criminal proceedings, which lead to highly consequential verdicts, typically require higher standards of proof and argument than newspaper opinion sections, which do not. Thus construed, the systemic view of democratic public discourse is highly valuable for two immediate reasons. First, it addresses questions surrounding the scale of democratic public discourse. In mass democracies, it is clearly impossible to include everyone in a giant forum of public debate. Interpreting public reason-­ giving as a process that occurs within a system composed of many interconnected sites gives us a more realistic picture of what inclusive public discourse could look like in a large-­scale society.49 Second, as we have just seen, the systemic view naturally accommodates the intuition that consciousness-­ raising groups and partisan media outlets can play a fruitful role in realizing inclusive public discourse, even though individually, such arenas are not fully welcoming to all groups and to all perspectives. These two reasons are already sufficient to recommend embracing the systemic view of public discourse. But, in our context, the systemic approach is valuable for a more pressing reason: it suggests a promising solution to the dilemma encountered by the shared reasons constraint—though, as we will also see, this solution remains importantly incomplete.

46  Mansbridge et al. (2012). 47  Chapter 2, Section 5; Chapter 5, Section 3; Chapter 6, Section 5; Chapter 7, Section 4. 48  Bächtiger and Parkinson (2019, ch. 2). 49  Mansbridge et al. (2012, 2).

Offering Shared Reasons  33

4.2  Restricting the Site of the Shared Reasons Constraint The systemic view highlights the fact that different discursive sites, which perform different functions, can be governed by different discursive norms. This points to a way of maintaining the shared reasons constraint while weakening its exclusionary implications: we can confine this constraint to the most formal deliberative forums, while waiving it in other settings. In other words, while the opinions of Supreme Court justices, and the final stages of legislative justification, must appeal to shared considerations, this does not apply to the public speech of countless other arenas. In churches, barbershops, newspaper opinion pieces, and even the opening stages of legislative debates, for example, people may permissibly appeal to non-­shared considerations. This proposal is by no means uncommon in democratic theory. Rawls, for instance, affirms that the duty of civility only applies to what he calls the ‘public political forum’, which includes the discourse of judges, government officials, legislators, and candidates for public office.50 Similarly, Krause suggests that restricting the shared reasons constraint to formal decision-­making sites, while waiving it in informal sites of discussion, can mitigate the problem of over-­exclusiveness.51 Yet we need a clearer picture of exactly how this solution is supposed to deal with the dilemma introduced in Section  2. The rough idea is this. On the one hand, the non-­domination of vulnerable and misunderstood groups is preserved by waiving the shared reasons constraint in more informal settings, so that their perspectives are included there. At the same time, the domination-­related benefits of having the shared reasons constraint are maintained by applying it to formal decision-­making arenas. In this way, restricting the site of the shared reasons constraint introduces a division of labour between formal and informal sites of public discourse—and this division of labour weakens the trade-­off articulated at the end of Section 2. But the devil is in the details. The success of this solution at weakening the trade-­off at hand depends on exactly how this division of labour is supposed to work. An initial proposal is that simply because the system has a part—informal public sites—that includes many viewpoints and reasons, including non-­shared ones, we can consider the system as a whole to be sufficiently inclusive. This is unsatisfying because those who press the complaint of over-­ exclusiveness do not merely want to express their important concerns in informal public fora. They want the expression of those concerns to influence formal decision-­ making that produces coercive decisions. Only if this condition is

50  Rawls (1997, 768).

51  Krause (2008, 121, 156).

34  Democratic Speech in Divided Times satisfied will the inclusion of their concerns allow them to contest policies and protect themselves from domination. So, a process must exist whereby the non-­shared reasons articulated in the informal sphere are then transmitted to formal decision-­making arenas. But decision-­making arenas are supposed to be governed by a shared reasons constraint. Accordingly, for novel considerations to make their way into formal decision-­making sites, they would have to somehow become shared. The question that will exercise us in the rest of this chapter is how this might happen. Krause considers this question head-­on, and envisions the process of making non-­ shared considerations shared—and so, admissible in formal arenas—as follows: Public reason [. . .] always operates in medias res. [. . .] Sometimes new principles are introduced into public dialogue, principles that previously were external to public reason. Yet in order for a new principle [to be introduced] it must connect up with some principles that are already subject to public agreement.52

The idea, it seems, is that we introduce new reasons into the stock of shared ­reasons by showing that currently shared reasons support or make sense of these new reasons. Thus described, the process has a broadly inferential structure: new considerations are derived from, or translated in terms of, reasons that are already shared. So, through logical argument, public debate produces a more adequately inclusive set of shared reasons. This derivation process notably echoes Rawls’s famous proviso in his ‘wide’ understanding of public political culture. According to Rawls’s proviso, citizens may appeal to reasons that are not yet shared or public provided that, in due course, they give properly public reasons to support those appeals.53 A legislator could permissibly appeal to the Gospel story of the Good Samaritan, Rawls suggests, so long as they ‘do not stop there, but go on to give a public justification for this parable’s conclusions in terms of political values’.54 The recurring thought is that justifications in terms of non-­shared reasons can be brought into formal 52  Krause (2008, 160–1). See also Rawls (1997, 784–6) and Habermas (2008, 130–1). 53  Rawls (1997, 784–5). 54  Rawls (1997, 786). Note that Rawls’ proviso is slightly ambiguous between two interpretations. The shared reasons introduced in due course could be introduced either to support the non-­shared considerations (as I have been suggesting here) or directly to support the policy that the non-­shared reasons aimed to justify (as I suggest in note 8). However, the latter option seems unsatisfactory for the present purposes: it allows deliberators to appeal to non-­shared considerations that are dear to them only when these are redundant from a justificatory perspective. That is, it suggests that non-­ shared reasons can be given to support P only when some other consideration already sufficiently supports P. But this once more generates the concern mentioned above: citizens who are concerned that morally important considerations are excluded from public discourse do not simply want to express those considerations in public; they want to use those considerations to influence policymaking.

Offering Shared Reasons  35 decision-­making sites on the condition that they are shown to link up adequately to considerations that are currently shared. While this proposal—to derive new reasons from the current stock of shared reasons—is prima facie appealing, it is unclear how fruitful it would be in settings marked by deep divisions and injustices. If the initial set of shared considerations is sufficiently small and biased against certain participants—perhaps because, as we saw in Section 2.3, others are ignorant of their experiences—it may not be possible to derive new considerations from the existing pool of shared considerations. And, by extension, it may not be possible for the novel reasons articulated in informal sites to make their way into formal decision-­making arenas. Am I being too pessimistic here? Even if in principle the derivation strategy can be inadequate when starting from highly exclusionary conditions, one might reply that actual conditions are not like this. For instance, Krause observes that even in pluralistic polities like the United States, many evaluative concerns are shared. Indeed, equality, liberty, and the pursuit of happiness are enshrined in the US Constitution.55 This, it might seem, makes it possible logically to derive many new reasons through public deliberation. ‘If we truly care about liberty and equality,’ proponents of same-­sex marriage can argue, ‘then we are logically committed to respecting different lifestyles and recognizing an equal right to marriage for same-­sex couples.’ However, this reply overlooks another obstacle that stands in the way of deriving novel reasons from currently shared reasons. Despite invoking liberty and equality, the US Constitution was long interpreted as legitimating slavery. Likewise, in Plessy v. Ferguson, the US Supreme Court upheld the constitutionality of racial segregation by propounding the ‘separate but equal’ doctrine. These cases remind us that even when a consideration x is widely recognized, this may well conceal deep disagreement over its correct interpretation. So, deriving a novel reason in a way that is likely to make it become shared requires not just connecting it with x, but more specifically connecting it to the audience’s interpretation of x. What is worse, Plessy v. Ferguson highlights that widely held prejudices—say, beliefs in a natural hierarchy—may infect ‘promising’ values, like equality, so that they come to be interpreted in a manner consistent with existing injustices. As a result, even when a reason seems shared, disagreement over its interpretation and exclusionary interpretations of its meaning may combine to obstruct attempts at analysing legitimate concerns in its terms. This last difficulty is a common feature of non-­ideal, divided societies. Mills, for example, suggests as much when commenting on contemporary understandings of equality and racism:

55  Krause (2008, 157, 160).

36  Democratic Speech in Divided Times If previously whites were color demarcated as biologically and/or culturally unequal and superior, now through a strategic ‘color blindness’ they are assimilated as putative equals to the status and situation of nonwhites on terms that negate the need for measures to repair inequities.56

Mills is saying that although equality is now widely affirmed, many understand it in colour-­blind terms. Roughly, on this conception, racial equality is realized when no one is treated differently because of their race. This interpretation, Mills suggests, is incompatible with many concerns that black activists have defended on grounds of equality. To those who endorse a colour-­blind conception of equality, the consideration that black people are owed differential treatment (including reparations for historical injustices) cannot intelligibly be derived from considerations of equality. Indeed, as Mills notes, to them the real opponents of equality and ‘the real racists are the blacks who continue to insist on the importance of race’.57 Such individuals will therefore reject attempts at translating demands for reparations in terms of equality. Or consider again same-­ sex marriage. Suppose an advocate of same-­ sex marriage wanted to introduce the non-­ shared consideration that same-­sex relationships are valuable into formal deliberative sites that are governed by the shared reasons constraint. And suppose she did so by connecting same-­sex relationships to the idea of a loving relationship, where her opponents recognized that loving relationships are valuable. In response to similar arguments, opponents of same-­sex marriage often respond that true romantic love is essentially a relation between a woman and a man. Some religious believers maintain this restrictive concept of love because, citing Corinthians, they claim that true love is essentially godly, while same-­sex relationships are sinful.58 Alternatively, if one believes that loving relationships are inherently healthy and valuable, but (in line with the DSM’s pre-­1973 classification) that homosexuality is a disorder, then one is bound to think that same-­sex relationships cannot be subsumed under the concept of a loving relationship. Still others might object, as some philosophers notoriously have, that same-­sex relationships, in virtue of instantiating a form of narcissistic desire, are perversions of genuine love.59 More generally, the problem is the following: when participants hold such contrasting concepts of love, there is no shared understanding of ‘loving relationship’ from which one might derive the consideration that same-­sex relationships are valuable. These considerations suggest that, in divided settings, deriving non-­shared reasons from the existing pool of shared considerations is very challenging. Although there may seem to be agreement on some shared reasons, this 56  Mills (2007, 27–8). For similar concerns, see Fricker (2013), Stanley (2015, ch. 3), and Talisse (2009, 63–4). 57  Mills (2007, 28). 58  Halley (2017). 59  Scruton (2001, 307–10).

Offering Shared Reasons  37 agreement is often more apparent than real: it conceals profound disagreement on how these considerations should be interpreted; and it obscures the fact that, in unjust circumstances, even seemingly promising considerations may be interpreted in ways that accommodate injustices. When the initial pool of genuinely shared reasons is highly limited, so too are the prospects for deriving new considerations on their basis. The broader conclusion, for our purposes, is that simply restricting the site of the shared reasons constraint to the most formal arenas of public speech is not enough to offset its domination-­related costs. Such a restriction is of course an important first step: by waiving the shared reasons constraint in other arenas, it opens up space for groups to appeal to important considerations that others currently misunderstand or reject. But these appeals will not help voters contest political power (and thereby protect themselves from domination) unless they eventually penetrate formal decision-­making arenas. We therefore need something more: namely, a plaus­ ible account of how appealing to non-­shared considerations in more informal sites can in turn contribute to making those considerations shared—and so, admissible in formal arenas that are governed by the shared reasons constraint. We are still missing such an account. As I have just suggested, the idea that we can simply logically connect these non-­shared reasons to existing shared reasons has limited prospects for success in non-­ideal circumstances. The rest of this chapter is devoted to alleviating this problem.

5.  The Form of Public Reason-­giving How can appealing to non-­shared considerations contribute to making them shared, so that the shared reasons constraint no longer excludes them from formal decision-­making arenas? In focusing on the shared reasons constraint, we have so far been considering a norm that governs the content of public speech. But to appreciate how the articulation of non-­shared considerations can help to make these considerations shared, we need to turn to another category of norms: norms governing the form of public discourse. These are norms that regulate, not the content of what is said, but rather the manner in which it is said. Specifically, I want to suggest that, to mitigate the exclusiveness of the shared reasons constraint, we should adopt an expanded view of what forms of public discourse are appropriate. In particular, we should welcome non-­argumentative and emotionally charged forms of speech within our normative account of democratic public discourse. Now, the public discourse of actual divided democracies is in fact rife with intensely emotional utterances that are not argumentatively structured. Accordingly, a key upshot of my recommendation is that, while divided political

38  Democratic Speech in Divided Times settings generate challenges for the shared reasons constraint, they also offer distinctive resources for overcoming these challenges. Note that my primary aim here is to sketch how such a discursive norm, which expands the forms of speech that can be used publicly to express propositional contents, might help enlarge the pool of shared considerations—and so, might help make the shared reasons constraint less exclusive. Having said that, I will not develop this solution at great length here. Instead, I will develop it in Chapter 2, where I specifically discuss the epistemic merits of angry public narratives.

5.1  Beyond Argument Some forms of public discourse, most notably argument, are inferential. They involve deriving conclusions from premises, via logical steps. The account we examined in the last section, of how expressing non-­shared reasons can lead them to become shared, embodies this inferential picture of public discourse. On this account, recall, public speakers are supposed to establish that the non-­shared reasons expressed in informal sites are logically supported by, or logically derivable from, reasons that are already shared. In other words, public speakers are supposed to offer arguments from shared premises to non-­shared conclusions. Yet this picture of public speech is incomplete. In addition to argument, democratic public discourse should also welcome forms of speech, such as narrative, that are non-­ argumentative and typically charged with emotion. Narrative is discourse that aims to transmit one’s experience or perspective on the world. This can consist of personal testimony—that is, telling others about one’s experiences of an event or practice. But it also includes more general perspective-­ sharing—that is, telling a story to get people to see an event or practice as one does.60 As I define it, emotionally charged narrative can involve not just positive emotions such as hope or joy, but also more contentious emotions, such as anger and despair. Importantly, these forms of discourse are not inferential: they do not involve logically connecting what is said to premises that are currently shared. This is promising in light of the difficulties encountered earlier. What conclusion one can reach via logical inference depends fundamentally on what premises one starts from. This is particularly clear in the case of deductive argument, which is non-­ampliative: there can be nothing in the conclusion that is not in some sense already in the premises. That is why, as we saw in the last section, the strategy of deriving new reasons from reasons that are currently shared seems unpromising 60  Sanders (1997, 369–73); Young (2000, 70–7).

Offering Shared Reasons  39 when it starts from an extremely restricted set of shared considerations. In contrast, since emotionally charged narrative is not inferential, it is not hostage to the lack of shared premises. It aims directly to introduce or make visible certain experiences or perspectives, and thereby to enrich the pool of publicly recognized considerations. The next section will further outline how the epistemic value of emotionally charged narratives helps offset the shared reasons constraint’s exclusiveness. Before proceeding, however, we should consider how the present recommendation departs from, and improves upon, the way public deliberation has traditionally been theorized. The idea that we should give a greater role to emotionally charged and non-­ argumentative forms of speech, such as narrative, is by no means new. It has long been championed by democratic theorists concerned with non-­ideal theory, most notably Iris Marion Young. But these theorists, as mentioned in the introduction, are often sceptical of discursive constraints such as the shared reasons constraint.61 By contrast, I am arguing that expansive norms relating to the form of democratic speech are needed to complement the shared reasons constraint, by alleviating its exclusiveness. On the other hand, deliberative theorists who do support the shared reasons constraint have traditionally been less accommodating towards emotionally charged and non-­argumentative forms of speech. In fact, Young’s defence of such discursive forms is a critical response to classical theories of deliberative democracy. Traditional deliberative democrats, Young suggests, aim to stave off the Platonic objection that democracy hands power over to the mob’s unruly passions. To do so, they adopt ‘rationalist’ norms of deliberation: they affirm that proper public communication consists not of emotional outbursts or intoxicating rhetoric, but of calm, logical arguments from common premises.62 Unlike the shared reasons constraint, which governs the content of public reason-­giving, Young emphasizes that these exclusionary norms govern its form. They exclude people’s contributions ‘from serious consideration not because of what is said, but how it is said’.63 Defenders of the shared reasons constraint sometimes explicitly illustrate this tendency. Rawls, for instance, stresses that the ideal of public reason is concerned

61  Young (1996, 131–2). See also Sanders (1997, 372), Mouffe (1999, 755), and Dryzek (2000, 73–4). Dryzek is more open to discursive constraints than some of these other theorists. But he nonetheless claims that ‘anything goes [. . .] provided that it is (a) capable of inducing reflection, (b) noncoercive, and (c) capable of connecting the particular experience of an individual, group, or category to some more general principle’ (2010, 34). This is weaker than the requirement that deliberators appeal to shared considerations, as some general principles may not be shared (Dryzek 2010, 34). 62  Young (2000, chs. 1–2). 63  Young (2000, 56).

40  Democratic Speech in Divided Times with ‘reasoning’—which is governed by ‘principles of inference’—rather than ‘rhetoric or means of persuasion’—which are not.64 Similarly, Habermas famously defends deliberative democracy as a way of subjecting power to the ‘unforced force of better argument’.65 And he interprets emotions as essentially non-­ cognitive states, whose political expression seems lacking in epistemic value.66 In such passages, it might seem that Rawls and Habermas are explicitly excluding emotionally charged and non-­argumentative discourse from their deliberative ideals.67 For the most part, however, the problem is not so much that defenders of the shared reasons constraint explicitly exclude emotionally charged narrative from their deliberative ideals. Rather, it is that they are silent about it. On the whole, Rawls and Habermas have very little to say about this category of discourse. This is also true of more recent defenders of the shared reasons constraint, particularly Rawlsian public reason theorists. Quong, for example, almost never mentions emotional or narrative forms of discourse. The main exception to this is even more telling. While listing objections to the idea of public reason, he cites Young’s concern that public reason ‘marginalizes [. . .] emotive, passionate, or rhetorical forms of discourse’. But he promptly brackets this objection, and never returns to it.68 The point, then, is that the democratic theorists who have most closely theorized the shared reasons constraint commonly do not explicitly assign a central role to non-­argumentative and emotionally charged speech within their normative models of public discourse. This, I am suggesting, is a mistake: a commitment to the shared reasons constraint should motivate such an expanded conception of discursive forms. As the next section will discuss, emotionally charged narrative performs a crucial epistemic function that helps make the shared reasons constraint less exclusive. However, note that there are some exceptions to this tendency among defenders of the shared reasons constraint. Some defenders of this constraint do recognize the importance of emotionally charged narrative. Krause, in particular, explicitly combines the shared reasons constraint with a firm endorsement of emotionally charged narrative. And, unlike some other deliberative democrats who combine these two features,69 she does this partly out of recognition of its 64  Rawls (1993, 220). 65  Habermas (1996, 306). 66  Habermas (1996, 294). 67  For discussion of this tendency in Rawls and Habermas, see O’Neill (2002) and Abizadeh (2007). For other classical deliberative democrats who occasionally seem to explicitly endorse rationalist norms, see Cohen (2009, 248–51), Elster (1998, 109), and Spragens (1990, 128). 68  Quong (2011, 260). 69  Gutmann and Thompson (2004, 50–1), Mansbridge (1999a, 225), and Chambers (2004, 402) combine the shared reasons constraint with acknowledgement of emotional speech’s importance, but stress emotional speech’s motivational rather than its epistemic value. Recent deliberative systems theorists encounter a different problem: they often seem open to the possibility of combining the two in different arenas. But, partly because they seldom discuss the shared reasons constraint in any detail,

Offering Shared Reasons  41 epistemic value. So, although Krause sometimes emphasizes what I consider to be more ‘inferential’ ways of expanding the pool of shared resources—as we discussed in Section  4—her broader account of democratic public discourse does make space for emotionally charged narrative. Hence, to clarify the picture of public discourse I am recommending, I will later return to the important residual differences between her account and my own.

5.2  Emotionally Charged Narrative and the Over-­exclusiveness Problem How does emotionally charged narrative help make non-­shared considerations shared—such that the shared reasons constraint no longer excludes them from formal decision-­making arenas? By contrast with, say, deductive inferences from shared resources, narrative is an ampliative form of communication. By telling a story that gets people to experience the world in a certain way, it strives directly or non-­inferentially to introduce new experiences and concepts into the public sphere.70 For instance, in his Invisible Man, Ralph Ellison vividly depicts the perspective of a black American man who is persistently overlooked, misunderstood, and slighted. In so doing, he effectively reproduces those experiences, such that readers from other social groups may imaginatively perceive how daily slights cumulatively produce far more than a trivial offense. In this way, readers may come to see (among many other things) the irreducible feature of the world that the concept ‘micro-­ aggression’ aims to capture. More generally, then, emotionally laden narrative helps show something when, because of a lack of shared experiences or concerns, it cannot yet convincingly be logically argued for. What role does emotion play in this? Narrative could conceivably be expressed dispassionately. So why am I defending emotionally charged narrative? Narrative seeks to increase the pool of shared experiences. But short of making people experience the actual relevant events themselves, this can only be done by getting them to imagine those experiences. However, even if someone describes an event accurately, imaginatively reconstructing it may be difficult. This is where emotion becomes epistemically significant. There is a widespread and involuntary tendency among human beings to resonate with or ‘catch’ the emotions of others. As Hume famously puts it: ‘as in strings equally wound up, they typically do not offer a detailed account of how the two are meant to combine. See, e.g., Curato et al. (2019) and Bächtiger and Parkinson (2019). 70  For evidence corroborating narrative’s ability to publicize new experiences, see Polletta and Lee (2006, 718–19), Wesolowska (2007, 674, 676), Huijer (2009, 237). Hannon (2020, 603–4) also con­ siders narrative discourse to be a useful tool for sharing experiences (albeit not in the context of assessing the shared reasons constraint).

42  Democratic Speech in Divided Times the motion of one communicates itself to the rest, so all the affections readily pass from one person to another’.71 This, in turn, suggests that weaving emotion into an enumeration of facts can facilitate the communication of experiences: by transferring emotion to one’s hearers, it helps them imagine those facts more vividly, in a way coloured by emotion. The following chapter will elaborate at length on the precise and distinctive way in which the ‘colouring’ offered by emotions is epistemically valuable. For now, I leave things at the intuitive idea that emotions have a distinctive felt experience that highlights or renders more salient significant properties of our perceptions and imaginings. Experiencing fear, grief, or anger, for example, makes us more alert and sensitive to potential sources of (respectively) danger, loss, or injustice. As a result, emotionally laden narrative seems an effective means of getting people to share one’s experience of an event or practice, and, relatedly, of enriching the conceptual categories they have for interpreting events and practices. Consider two examples relating to same-­ sex marriage, both drawn from French public debates in 2013. The first illustrates how emotionally laden narrative in civil society can communicate experiences and thereby enrich publicly shared conceptual resources. Recall the deliberative impasse discussed in Section 4.2. In this example, defenders of same-­sex marriage want to publicize the consideration that same-­ sex relationships are valuable by connecting such relationships to the idea of a loving relationship, but the concept of a loving relationship is itself disputed. Consider now La Vie d’Adèle, a film which gained wide attention against the background of public debates concerning same-­sex marriage. The film portrays the relationship of two women, Adèle and Emma. Given our interest in how practices and claims are understood in public, observing the language critics used to characterize the film is instructive. La Vie d’Adèle was notably described by critics as ‘evok[ing] love in its purest and most passionate form’,72 as ‘brilliantly demonstrating how love feels’,73 and finally as a ‘devastating portrayal of the lifespan of a relationship, one that [is] [. . .] convincing and intimate’.74 La Vie d’Adèle, then, was widely received as a strikingly accurate and moving representation of love. By representing the two protagonists’ relationship in such an intensely emotional way, it seems to show what could not be logically argued for, given the previously mentioned deliberative impasse: that if anything constitutes a loving relationship, Adèle and Emma’s relationship surely does. Thus, it helps its audience see same-­sex relationships as loving relationships, and

71  Hume (2009[1738], 3.3.1.7). See also Coplan (2011) and Chapter 2, Section 4.1. 72  Bradshaw (2013). 73  Quirke (2013). 74  Clift (2014).

Offering Shared Reasons  43 thereby contributes to generating new, shared understandings—here, a more inclusive concept of ‘loving relationship’.75 A second example indicates how emotionally laden narrative can help to publicize new perspectives and conceptual resources in the early stages of legislative debates. In January 2013, Christiane Taubira, then French Minister for Justice, prefaced parliamentary debates on same-­sex marriage with a speech describing marriage’s historical evolution. As she reports with a mixture of indignation, resolution, and pride, though marriage in France was initially reserved for Catholics, it then became available to Jews and Muslims. Later, the secularization of marriage made it available to couples of all faiths, of no faith, and to some hitherto-­excluded professions. In parallel, the relationship marriage defines gradually became more egalitarian. Divorce became legal, wives no longer needed their husband’s legal authorization to open a bank account, and, eventually, marital rape was recognized. By telling this story with passionate rhetoric, Taubira communicated a distinctive perspective on marriage, and this perspective encouraged her audience to reconceptualize marriage as follows. First, according to the perspective she offers, marriage has no fixed meaning. Second, its meaning has always shifted towards greater inclusiveness and equality, and these shifts were widely regarded as moral progress. Finally, with this new conception of marriage in place, extending the right to marry to same-­sex couples could appear as the natural next step in marriage’s march towards greater inclusiveness. By publicizing this conception of marriage, Taubira’s emotionally charged narrative arguably made it easier subsequently to argue for same-­sex marriage by appeal to shared premises. What these cases illustrate is that emotionally charged narrative performs an epistemic function that is distinctively suited to mitigating the over-­exclusiveness of the shared reasons constraint. The shared reasons constraint’s over-­ exclusiveness, remember, results partly from the deep epistemic obstacles diagnosed by social epistemologists. Where different social groups have radically different experiences of the world, and where the conceptual resources needed to make sense of those experiences are often lacking, some considerations that are crucial to publicly contesting power will be unacceptable and eventually unintelligible to others. The resulting paucity of shared considerations makes even weaker interpretations of the shared reasons constraint overly exclusive (Section 2.3), and limits the extent to which we can introduce new considerations by logically connecting or deriving them from currently shared ones (Section 4.2). 75  La Vie d’Adèle’s accuracy as a representation of lesbian relationships is not undisputed. E.g., Maroh (2013). Accordingly, I present this as an example of epistemically effective narrative only in a qualified sense: against a political background where many citizens denied the standing of same-­sex relationships as loving relationships, the film supplied (1) a less inaccurate representation of a same-­sex relationship, which (2) vividly showed how unsustainable these prejudices were, and (3) did so in a way that widely impacted public perceptions. This is compatible with retaining doubts regarding its overall accuracy.

44  Democratic Speech in Divided Times Emotionally charged narrative tackles the problem at its source: it facilitates the communication of non-­ shared experiences and enriches the conceptual resources needed to interpret these. Crucially, moreover, since it is not inferential in form, its ability to publicize new experiences does not—contrary to the ‘derivation’ strategy canvassed in Section 4.2—depend on what considerations are already shared. In light of these promising qualities, we should adopt more permissive discursive norms regarding what forms of public discourse are desirable in order to redress the over-­exclusiveness of public discourse’s content. Discursive ideals that exclude or ignore emotionally charged narratives deprive themselves of a key epistemic instrument for publicizing new considerations, and thereby help to sustain the shared reasons constraint’s exclusiveness. Notice a further advantage of the present proposal. Expanding the forms of public discourse helps to soften the dilemma we started with. This is because it weakens the trade-­off between protecting the unreasonable from domination and protecting vulnerable participants from domination. On the one hand, it maintains the shared reasons constraint without necessarily excluding the unreasonable, and so preserves its generic domination-­related benefits. On the other, by making the introduction of new shared reasons less dependent on what reasons are already shared, it facilitates the introduction by vulnerable groups of previously overlooked concerns. Hence, it promotes the non-­ domination of vulnerable parties, and thereby blunts the force of the over-­ exclusiveness objection. This stands in contrast to the strategy of idealizing the shared reasons constraint’s constituency. As Section 3 explained, this strategy proposes to reduce the shared reasons constraint’s exclusiveness and thus to protect vulnerable groups from domination, but only by exposing unreasonable citizens to greater risks of domination. While defenders of the shared reasons constraint typically do not give a central role to emotionally charged and non-­argumentative forms of speech in their discursive ideals, I noted earlier that this is not always true. The rich account of public deliberation Krause offers in Civil Passions also combines the shared reasons constraint with an endorsement of emotionally charged speech in deliberative politics. To further clarify the ideal of public discourse I am recommending, it is therefore helpful to highlight the important ways in which her account nevertheless differs from my own. Though promising, Krause’s account ultimately falls short in two respects. The first and more substantive issue has to do with the constituency of the shared reasons constraint. As we saw in Section 3, Krause recommends that we idealize or restrict the perspectives that are relevant to determining which concerns are publicly admissible. I have argued that doing so is deeply problematic. Not only does it expose unreasonable citizens to domination, but it also may be counterproductive in non-­ideal contexts of power. As far as possible, we should therefore be reluctant to engage in such idealization.

Offering Shared Reasons  45 The second issue is more a matter of emphasis. When examining how non-­ shared reasons might become shared or public—and so, might come to be admissible in the final stages of formal decision-­making—Krause underplays important features of her own account of public discourse. She insists, as we discussed in Section 4.2, on the need logically to connect the novel considerations introduced in informal sites to existing shared considerations. But this is not strictly necessary. Emotionally charged narrative gets listeners to imaginatively perceive important experiences and events. In so doing, it can make listeners appreciate important considerations (e.g., that same-­sex relationships can be deeply valuable, or that micro-­ aggressions can be deeply harmful) without establishing that those considerations are logically implied by, or derivable from, some prior shared premise. This, to reiterate, is why emotionally charged narrative plays a fundamental role in making a relatively non-­idealized shared reasons constraint more palatable. It helps enlarge the pool of shared reasons, and thereby reduces the shared reasons constraint’s exclusionary character, when doing so via inferential forms of speech would be extremely difficult.

5.3  Preliminary Objections Chapter 2 will further elaborate on the function of emotionally charged narrative in public discourse, particularly by expanding significantly on the sense in which communicating emotions is epistemically valuable. In the meantime, let us forestall some preliminary worries. To begin, one might object, as Benhabib does, that inviting emotionally charged narrative into public debate undermines the impartiality of political decision-­making.76 Emotionally laden narrative involves conveying one’s partial perspective on the world, trying to make one’s particular concerns publicly recognized. Admitting such discourse might therefore seem contrary to the aim of impartial rule. But the introduction of partial emotionally charged perspectives is not genuinely opposed to impartiality. For one thing, the discursive norm I am advancing is compatible with insisting that, in the most formal stages of public debate, coercive decisions should be justified impartially, by appealing to shared considerations. I am inviting emotionally charged narratives as an epistemic instrument for making others see a new consideration as a reason for political action. Only if this attempt at making the consideration shared succeeds may that reason be used to justify one’s cause in the most formal sites of decision-­making. Thus, while Taubira may deliver an impassioned and partial narrative at the outset

76  Benhabib (1996, 83).

46  Democratic Speech in Divided Times of legislative debates to get her interlocutors to see marriage in a certain light, she must resort to shared reasons when those debates reach their final stages. In fact, and secondly, allowing partial narratives is not simply consistent with impartiality: it is also instrumentally necessary for truly impartial public deliberation, which gives fair consideration to the interests of all. Because we occupy particular social positions, no single person reflecting on societal problems can fairly consider the concerns of all. Indeed, I unavoidably focus on some considerations to the exclusion of others simply because they are more salient to me, from my particular standpoint. To correct these socially caused biases, we must engage with occupants of different social positions. In this light, inviting partial narratives helps, rather than hinders, impartiality: by pooling different social perspectives, it helps us register and appreciate the concerns of different social groups. A more daunting concern is that complementing the shared reasons constraint with a discursive norm assigning a central role to emotionally charged narrative might seem to overachieve—it is, one might think, over-­inclusive. Alongside admirable emotional narratives like those mentioned in the last section, this norm might—absent further qualification—open the door for extreme or deeply misguided emotional narratives. After all, for some public speakers, revealing their personal worldview might mean expressing deep-­ seated rage at racial minorities and immigrants. This is of course not just a hypothetical concern. In conditions of widespread ignorance, demagogues routinely use impassioned appeals to arouse misguided sentiments and thereby steer their listeners to unjust ends. Accordingly, Plato was extremely concerned that intoxicating rhetoric might induce the demos to embrace tyrants and tyrannical policies.77 But we hardly need to go all the way back to Plato. This concern is highly salient in contemporary divided democracies, where demagogues such as Marine Le Pen or Donald Trump express narratives of fear and anger that they unfittingly direct at vulnerable minorities. The first thing to note in response to this concern is that it is not specific to emotionally laden narrative. Dispassionate argument too can be abused. One of the most effective forms of sophistry consists in endowing fallacious inferences with the trappings of rational argument or scientific authority. We are all too familiar with how, as the DSM’s classification of homosexuality manifested, purported experts sometimes weigh in misleadingly to public debates by showering audiences with esoteric formulas, technical jargon, or deceptive statistics. So appeals to rational argument and expert knowledge too can be misused. And yet, this is not usually considered a sufficient reason to purge the deliberative ideal of argument or expert testimony.

77  Plato (2007[c. 380 BC], 565e–70c).

Offering Shared Reasons  47 But given how deeply problematic this ‘over-­inclusiveness’ concern is, and how pressing it is in actual divided democracies, we should not leave things at this first observation. Instead, we must investigate more closely what the appropriate normative response to extreme public narratives is. How should we qualify or complement the norm, advanced in this chapter, that democratic public discourse should invite emotionally charged personal narratives? What, in other words, should we do about public narratives expressing, say, a racist’s rage? There are actually two over-­inclusiveness concerns here. The worry might be about the content expressed by the narrative, the deeply disrespectful racist views that are publicly communicated. Alternatively, it might concern the emotion that is being expressed, independently of what that emotion is about. For example, one might think that publicly expressing rage is morally undesirable, irrespective of what that rage is about. More generally, this second worry targets negative emotions such as anger or resentment. The next two chapters will be devoted to exploring these two concerns. I investigate the issue of negative emotions in Chapter  2, where I defend the place of anger in democratic discourse. As for Chapter  3, it examines deeply disrespectful or vilifying speech, and how we should counter such forms of speech.

6. Conclusion The shared reasons constraint has strong prima facie appeal: by facilitating processes of political contestation, it contributes to protecting citizens from domination. As social epistemologists have argued, however, contemporary democracies are characterized by deep experiential divides and hermeneutical injustice, as a result of which some citizens fail to accept and even understand crucial concerns of other citizens. In such settings, even the weakest variant of the shared reasons constraint defines the content of public discourse too restrictively. In doing so, it politically disempowers those whose concerns are misunderstood or unappreciated. So, we have a dilemma relating to non-­domination: pro tanto reasons grounded in non-­domination pull both for and against the shared reasons constraint. The solution to this dilemma is not to restrict the constituency of the shared reasons constraint to idealized ‘reasonable’ participants. Although doing so expands the pool of shared reasons, it is ill-­advised in contexts where unreasonable citizens are also empowered. And even outside of such contexts, it exposes unreasonable citizens to domination. This strategy therefore does not attempt to soften, let alone dissolve, the above dilemma. It simply settles it in favour of reasonable citizens. Hence, it should only be considered as a last resort. Instead, then, I have offered a two-­step solution. In the first place, we should restrict the shared reasons constraint to the most formal sites of public discourse,

48  Democratic Speech in Divided Times which produce coercive decisions. Doing so creates space for people to appeal non-­shared considerations in a wide range of more informal contexts, such as social media, party conferences, or even the early stages of legislative debates. But this is not nearly enough. We also need an account of how expressing non-­ shared reasons in informal sites can lead them to become shared, so that they may eventually influence formal decision-­making. And this account must be sensitive to the following problem: in divided settings where shared considerations are scarce, it is often impossible to logically argue for these non-­shared reasons using the existing stock of shared reasons. To meet this challenge, we should adopt expansive norms concerning the form of public discourse. In addition to arguments from shared premises, public discourse should welcome emotionally charged narratives. Because of their non-­ inferential character and of their emotional tenor, such narratives constitute an effective epistemic tool for making new considerations shared. As a result, by giving such discursive forms a greater place in democratic public discourse, we can maintain the shared reasons constraint while attenuating its disempowering effect on vulnerable groups. Nevertheless, absent qualification, this proposal may seem over-­ inclusive. First, it might open the door for narratives whose content is deeply disrespectful. Second, it might invite narratives charged with highly negative emotions, such as anger and rage. Even if one thinks that some emotions have an important part to play in political discourse, one may have reservations about profoundly negative emotions. The following two chapters address these two concerns in reverse order.

2

Voicing Anger 1. Introduction In the classic novel Invisible Man, Ralph Ellison depicts the perspective of an unnamed black American whose skin colour renders him ‘invisible’: he is persistently ignored, misunderstood, or mistaken for another. In a final moment of introspection, the invisible narrator turns to the reader: So why do I write, torturing myself to put it down? Because in spite of myself I’ve learned some things [. . .] Why should I be the one to dream this nightmare? Why should I be dedicated and set aside—yes, if not to at least tell a few people about it? There seems to be no escape. Here I’ve set out to throw my anger in the world’s face [. . .] ‘Ah,’ I can hear you say [. . .] ‘He only wanted us to listen to him rave!’ But only partially true: Being invisible and without substance, a disembodied voice, as it were, what else could I do? What else but try to tell you what was really happening when your eyes were looking through? And it is this which frightens me: Who knows but that, on the lower frequencies, I speak for you?1

The passage is fascinating, not least for what it suggests about the communication of anger. First, the narrator emphasizes the importance of ‘throwing [his] anger’ at his audience. Further, he seems to do so for epistemic reasons, to make visible what had previously been invisible, by foregrounding his nightmarish experiences. Relatedly, perhaps, he claims that in expressing his anger, he speaks for the audience. Interestingly, the invisible narrator is not alone in voicing such thoughts. His suggestions resonate strongly with the poet Audre Lorde’s assertion that ‘my anger and your attendant fears, perhaps, are spotlights that can be used for your growth, [. . .] for corrective surgery’.2 Or consider, similarly, the novelist and civil rights advocate James Baldwin, who once declared that ‘to be a Negro in [America] and to be relatively conscious, is to be in a rage almost all of the time’. Baldwin urges his listeners ‘to control that rage so that it won’t destroy you’, and to instead turn it into a productive force. For him, as a novelist, doing so manifests itself in the process of ‘creat[ing] a person and mak[ing] other persons feel what this person feels’.3 A pattern emerges.

1  Ellison (1952, 579–81).

2  Lorde (1997, 278).

3  Baldwin et al. (1961, 205).

Democratic Speech in Divided Times. Maxime Lepoutre, Oxford University Press (2021). © Maxime Lepoutre. DOI: 10.1093/oso/9780198869757.003.0003

50  Democratic Speech in Divided Times In this chapter, I examine the place of anger in democratic public discourse. In doing so, I will try to make sense of the recurrent themes above. How is publicly expressing anger ‘corrective surgery’ for one’s audience? What is it capable of teaching us, or making us feel? And why does it matter that, as the invisible narrator suggests, the angry speaker be ‘dedicated and set aside’ for this purpose? These recurrent observations, I will argue, help us better understand the function that anger plays in public discourse—and, by implication, the place that it should occupy in the democratic system. Consider how this fits within my broader inquiry. In Chapter 1, to counter the objection that the shared reasons constraint excludes too many things from public debate, I recommended giving a greater role to emotionally charged narrative in democratic public speech. But a powerful objection to this recommendation is that it risks being too inclusive. It risks, for example, inviting in a racist’s enraged narrative. There are two worries here: one concerns the content of the racist views being expressed; the other bears on the rage, or intense anger, the speaker expresses, independently of what that rage is about. The present chapter considers this second worry. Chapter 3 will then turn to the first worry, by investigating how we should counter public expressions of deeply disrespectful views. To keep the two worries apart, the chapter at hand will focus on cases where the contents of the angry utterance are not immoral or disrespectful. Imagine, for example, an American civil rights activist in the 1960s, who affirms the principle of racial equality and denounces as unjust the fact that black Americans are denied equal rights. And suppose she expresses these views in a way that is charged with deep anger. Is it morally desirable that she publicly expresses this anger in conditions marked by deep divisions and pervasive injustices? Or ought she, morally, to refrain from expressing her views in a way that is suffused with anger? Democratic theorists who explicitly invite emotions into their normative conceptions of public deliberation often intimate that this includes negative emotions such as anger. Krause, for instance, cites with approval the anger that members of ACT UP publicly directed at Americans’ failure to take the AIDS crisis seriously.4 Similarly, Gutmann and Thompson praise the intense outrage that Senator Carol Moseley Braun voiced in response to the proposed renewal of a patent on the Confederate flag insignia.5 According to these theorists, then, deliberators needn’t necessarily refrain from publicly expressing their anger. But anger has come under attack. Rage and resentment pervade contemporary politics. And, to many, this pervasiveness seems problematic. As a result, some

4  Krause (2008, 119). See also Young (1996, 124), Mansbridge (1999a, 223), and Chakravarti (2014). 5  Gutmann and Thompson (1996, 135–8).

Voicing Anger  51 philosophers have revitalized a historical tradition of scepticism towards anger. Publicly expressing anger, these philosophers suggest, is morally undesirable. Perhaps the most influential argument for this view, an argument which is commonly echoed in broader public discourse, asserts that expressing anger is counterproductive: although anger is characteristically an aversive reaction to injustice, expressing anger is likely to exacerbate existing injustices. Thus, even if some emotions have an important role to play in public discourse, public speakers should refrain from voicing deep anger. Justice would be better served, critics conclude, by expressions of forgiveness, love, or even meekness.6 I believe that this scepticism is misplaced. In what follows, I will argue that anger can and should play a key role in democratic public discourse. My argument proceeds as follows. After offering a working definition of ‘anger’ (Section 2), I  provide an overview of the counterproductivity objection, and identify two complementary strategies for defending anger’s productivity in the face of this objection. Each will be taken up in turn (Section 3). The first part of my defence theorizes an important positive consequence of publicly expressing anger. I suggested, in Chapter  1, that expressing emotion plays a vital epistemic role, insofar as it helps to communicate experiences of politically significant phenomena. Here, I develop this insight in the specific case of anger. Because of anger’s characteristic phenomenology, communicating anger to one’s listeners can play an indispensable role in helping them register and understand existing injustices. This epistemic role is of crucial importance in divided societies, where the injustices suffered by some groups are often invisible to, or misunderstood by, others (Section 4). This initial argument is nonetheless insufficient to stave off the counterproductivity objection. To complement it, the second part of my defence examines what constitutes the correct context for evaluating the productivity of angry speech. Specifically, I argue that the consequences of angry discourse should be assessed from a broad, ‘systemic’, perspective. This reframing renders concerns about anger’s overall productivity or counterproductivity substantially more tractable than they initially seem (Section  5). Having defended and clarified the place of anger in public speech, I conclude by revisiting Ellison’s invisible narrator (Section 6).

2.  Defining Anger To define our subject more precisely, let us outline some key properties of the emotion I am calling ‘anger’. Like all emotions, and unlike moods, anger is an 6 On forgiveness, love, and meekness, see respectively McGary (1989), Nussbaum (2016), and Pettigrove (2012).

52  Democratic Speech in Divided Times intentional attitude, in the philosophical sense of the term—an attitude that is directed at a particular object (its content).7 Anger, put differently, is about something: A is angry that B betrayed him; C is angry that she was not promoted. Secondly, anger is a cognitive attitude: an important part of what anger does is aim accurately to represent certain features of the world. More specifically, philosophers commonly suggest that anger represents the content it is directed at as involving a moral violation or injustice. Because anger has this cognitive dimension, we can assess it as being more or less correctly directed, or fitting. Anger is fitting, first, only if its content actually involves an injustice; and, second, only if its intensity is proportionate to the severity of the injustice it is purporting to represent. Rage is not a fitting response to a minor moral violation, such as your borrowing my favourite pen without permission.8 Third, anger is phenomenologically distinctive: there is a distinctive qualitative experience that is involved in being angry at something. This experience is partly characterized by certain bodily feelings, such as an increased heart rate, feeling hot, or trembling. However, anger’s distinctive phenomenology also concerns the content it is directed at. A common observation among philosophers of emotion is that although emotions are directed at objects that are already given to us through perception, imagination, memory, or belief, emotions do not leave the representation of these objects unchanged. Instead, they are ‘distinctive ways of seeing a situation’.9 In Section 4, I will say more about how anger characteristically affects the way we experience its content, and why this matters politically. Finally, besides having a cognitive dimension, anger has a conative aspect. Not only is anger an attitude that aims to represent something about the world, but it also moves us to action. For example, A’s rage at the unfair treatment of black people in the criminal justice system may motivate A to retaliate against law enforcement agencies, or to protest for their reform.10 With this picture of anger in place, we can define this chapter’s focus more precisely. As we will see shortly, Martha Nussbaum has mounted an influential challenge to anger, which centres on its conative dimension: anger is problematic, she claims, because it disposes us to retributive or vengeful actions, actions aimed at inflicting payback on the perpetrators of injustice.11 But even Nussbaum allows 7  On emotions’ intentionality, see Deonna and Teroni (2012, 3–6), Brun et al. (2008, 11–12), and Nussbaum (2015, 42–3). 8 On the fittingness conditions of anger, see Cogley (2014, 201–3), Chakravarti (2014, 3–6), Srinivasan (2018, 128–9), and Nussbaum (2015, 42–3). 9  Jones (1996, 11). 10  This picture of anger is compatible with many theories of the emotions, including the perceptual theory (e.g., Döring (2007), Tappolet (2011)), and the attitudinal theory (e.g., Deonna and Teroni (2012, ch. 7)). And although, as I discuss later, it is unclear whether the evaluative judgment theory (e.g., Nussbaum 2015) can successfully explain anger’s distinctive phenomenology, advocates of this theory typically acknowledge that emotions are phenomenologically distinctive. For discussion, see note 56, this chapter. 11  Nussbaum (2015, 45–8).

Voicing Anger  53 for some forms of anger, which she labels ‘transitional anger’, and of which Martin Luther King is the paradigmatic illustration. While transitional anger’s most distinctive attribute is that it is not retributive,12 Nussbaum’s discussion of King makes it clear that it has further distinguishing properties. A second characteristic is that although, like anger generally, transitional anger is directed at a morally deficient object, it swiftly transforms into a forward-­looking attitude, which is oriented towards advancing justice. Thus, someone experiencing and expressing transitional anger does not dwell on past injustices.13 Third, although transitional anger sometimes moves individuals to non-­retributive violence (for instance, in self-­defence), it habitually motivates non-­violent actions.14 Finally, transitional anger has a distinctive tone. Its typical expression is not fiery or harsh, but calm and self-­controlled.15 To avoid arguing against a straw man when defending the place of anger in public discourse, I will focus on non-­transitional forms of anger. To this end, this chapter focuses predominantly on the public speech of Frederick Douglass (1818–1895), who relentlessly opposed slavery after escaping from it, and Malcolm X (1925–1965), who famously militated against the oppression of black Americans. Both were heavily involved in the struggle for racial justice, and both were reputed for their intensely angry rhetoric. Crucially, the anger they expressed was not typically transitional anger. First, its conative dimension, or aim, could be retributive. Douglass, for instance, once asserted that slaveholders ‘deserve to have [their throats] cut’.16 And even when its aim was not retributive, their anger nonetheless typically differed from transitional anger. It was often agitated and harsh, sometimes shading into rage, it did not shy away from encouraging violence, and it often dwelled on past and current injustices. Consider, for example, Malcolm X’s insistent depictions of the ‘nightmare’ constituted by black Americans’ daily lives—a rhetoric which contrasted so starkly with King’s own forward-­looking rhetoric of ‘dreams’ that King repeatedly distanced himself from it.17 Besides these dialectical reasons, there is also a political reason for examining the value of non-­transitional anger. The political reason is that such anger is widespread in non-­ideal public discourse. Douglass and Malcolm X, with their deeply angry rhetoric, played influential roles in their respective struggles. What is more, the intensity of their anger echoes that of the three figures we started with. Baldwin, in particular, makes it clear that it is rage he needs to control and channel. So, while political philosophers have given due attention to the rhetoric of more moderate speakers, like Martin Luther King,18 a normative assessment of

12  Nussbaum (2015, 44). See also Nussbaum (2016, 212). 13  Nussbaum (2015, 52–4). 14  Nussbaum (2015, 52–4). See also Nussbaum (2016, 212, 221). 15  Nussbaum (2015, 52–4; 2016, 222, 228–30). 16  Cited in Oakes (2007, 100). 17  Baldwin (1986). 18  E.g., Dryzek (2010), Rawls (1997).

54  Democratic Speech in Divided Times democratic public discourse in non-­ideal conditions should pay greater attention to the more controversial, but nonetheless politically salient, case of non-­ transitional anger.

3.  The Counterproductivity Objection: An Overview 3.1  The Counterproductivity Objection One of the most common charges against publicly expressing non-­transitional anger holds that doing so is counterproductive: it is likely to amplify rather than alleviate existing injustices. This accusation pervades actual political discourse. Social movements that angrily denounce, say, police brutality against people of colour, discrimination against women, or abuse directed at the LGBT community are frequently told that their advocacy would be more effective if it were less shrill. But this concern also has a distinguished philosophical history. Perhaps most famously, Seneca asserts that anger is inconsistent with humankind’s ethical ends: ‘mankind is born for mutual assistance; anger for mutual ruin’.19 Accordingly, ‘anger has nothing useful in itself ’.20 The view that anger and its expression are counterproductive has also been revived by contemporary philosophers. Glen Pettigrove, for instance, suggests that because anger may interfere with interpersonal communication, blind angry individuals, and motivate oppressive actions, ‘the person who does not grow angry [. . .] will be better positioned to focus on promoting common goods’.21 Among contemporary theorists, however, the counterproductivity objection owes its most prominent restatement to Martha Nussbaum. In Anger and Forgiveness, Nussbaum contends that, because of its tendency to motivate payback, ‘anger is always normatively problematic, whether in the personal or in the public realm’.22 As we will see in the following section, Nussbaum is partly concerned with anger’s intrinsic moral standing. But when explaining what makes anger problematic in the public realm, she frequently emphasizes its bad consequences.23 First, Nussbaum claims, anger tends to be an irrational waste of energy, which distracts from the pursuit of justice. Anger typically motivates individuals to seek payback or vengeance. But seeking payback, Nussbaum claims, will rarely undo the wrong that prompted the anger. Hence, ‘if we ponder the futility of the

19  Seneca (1928[c. 45], I.5). 20  Seneca (1928[c. 45], I.9). See also Adam Smith (1982[1759], I. ii. ch. 4). 21  Pettigrove (2012, 347, 369–70). See also Silvermint (2017, 470–5). 22  Nussbaum (2016, 5). 23  Nussbaum (2016, 220).

Voicing Anger  55 payback wish [. . .] we quickly discover that non-­anger and a generous disposition are far more useful’.24 Second, Nussbaum continues, anger is likely to positively obstruct the pursuit of justice. By ‘breeding mistrust’ and ‘increas[ing] the anxiety and self-­ defensiveness’ of its targets, expressing anger undermines interpersonal cooperation.25 And interpersonal cooperation seems necessary for the wide-­scale promotion of justice. In the worst cases, by inciting payback and eroding trust, anger risks triggering a ‘cycle of blood vengeance’.26 This is no doubt what King also feared when condemning Malcolm X’s violently angry rhetoric, which he claimed could ‘reap nothing but grief ’.27 So, expressing anger seems deeply misguided if we truly care about combatting injustice. In the fight against injustice, ‘anger is politically futile’.28 Anger’s alleged counterproductivity has normative upshots for both formal and informal political arenas. For Nussbaum, formal political institutions ought to be governed by a spirit of non-­anger. Hence, the people who populate and uphold these institutions should exercise emotional self-­restraint and resist the siren song of anger.29 What does this mean for specific political institutions? In the judicial realm, court proceedings should not institutionalize victim impact statements, where wronged parties voice their grievances.30 In the realm of transitional justice, Truth and Reconciliation Commissions (TRCs) ought to avoid providing an outlet for non-­transitional anger. Instead, they should give expression to ‘generous forward-­ looking thoughts’.31 Accordingly, Nussbaum commends Desmond Tutu insofar as he discouraged narratives of anger during the South African TRC.32 And although Nussbaum never discusses formal legislative politics, her condemnation of public anger presumably applies there too: if publicly expressed anger truly impedes justice, this seems just as problematic when making laws as when interpreting them. Nussbaum likewise recommends that informal political contestation be governed by norms of non-­anger. Indeed, she categorically praises Martin Luther King, Mohandas Gandhi, and Nelson Mandela for repudiating anger. Their concerted efforts to refrain from expressing deep anger and to discipline away the anger that they sometimes felt, she claims, were crucial to the success of their

24  Nussbaum (2016, 228). 25  Nussbaum (2016, 230). 26  Nussbaum (2016, 1, 230–3. 27  Cited in Baldwin (1986, 403). 28  Nussbaum (2016, 236). 29  Nussbaum (2016, 197). 30 Nussbaum (2016, 194–5). Nussbaum (2016, 196–7) does cautiously consider allowing such statements post-­sentencing, but precisely because this would diminish their political impact. 31  Nussbaum (2016, 240). 32  Nussbaum (2016). However, Nussbaum goes further than Tutu: because she worries that even forgiveness might be a covert form of anger, she recommends that TRCs advance a narrative of unconditional generosity. For discussion of how the South African TRC discouraged expressions of anger, see Chakravarti (2014, ch. 2).

56  Democratic Speech in Divided Times social movements.33 Because of this, Nussbaum concludes that we ought, morally, to emulate them: ‘the revolutionary non-­anger of Mohandas Gandhi and Martin Luther King, Jr., is proposed not as a distant hope but as an immediate task, to be embraced here and now in the confrontation with injustice’.34 The point is clear: participants in informal political contestation who wish to combat injustice should refrain from expressing non-­transitional anger.

3.2  Consequentialist and Non-­consequentialist Strategies The most direct strategy for responding to this objection is to deny that expressing anger is necessarily counterproductive: often, it may have more good effects than bad effects. But this is of course hard to show. To sidestep this thorny issue, Amia Srinivasan therefore suggests an alternative strategy. According to Srinivasan, the question of whether or not political anger is desirable is not fully answered by an assessment of anger’s consequences. Even if it were true that overall, we have instrumental reasons not to get angry, there may still be intrinsic reasons to do so, reasons unrelated to anger’s consequences. ‘For any instance of counterproductive anger we might still ask: is it the fitting response to the way the world is? Is the anger, however unproductive, nonetheless apt?’35 What makes anger apt? Recall that anger has a cognitive dimension: it is directed at an object, which it represents as involving a moral violation or injustice. Anger is cognitively fitting only if its object actually involves an injustice and only if its intensity is proportionate to the severity of that injustice. For Srinivasan, aptness is the intrinsic moral value that accrues to anger in virtue of being a fitting cognitive response to injustice. The idea is that, just as there is intrinsic aesthetic value in appreciating beauty, so too there is intrinsic moral value in anger that correctly registers injustice.36 Because anger is apt in virtue of fittingly or correctly registering its object’s moral properties, its aptness does not depend on producing good consequences. Therefore, even if anger is instrumentally problematic, its intrinsic value suggests that we might yet be able to resist the conclusion that it is all-­things-­considered morally undesirable. One worry with this response is that, for Nussbaum, anger’s tendency to mo­tiv­ ate retributive action, or payback, is not merely instrumentally problematic. It also makes anger questionable from a non-­ consequentialist perspective. On Nussbaum’s view, either retribution is driven by the thought that inflicting 33  Nussbaum (2016, 227, 231–3, 236). Nussbaum (2016, 236) acknowledges that King and Mandela occasionally experienced and even expressed anger. But she insists that, crucially, this anger was fleeting: it was (and needed to be) swiftly followed by a transition to forward-­looking love and compassion. 34  Nussbaum (2016, 218). 35  Srinivasan (2018, 126). 36  Srinivasan (2018, 127–31).

Voicing Anger  57 suffering will somehow undo the injury which prompted one’s anger, or retribution inflicts suffering in order to degrade or ‘down-­rank’ the offender’s status. In the latter case, retribution seems immoral, displaying an obsession with rank. In the former case, retribution is irrational, since inflicting suffering cannot possibly undo the injury one has sustained.37 Therefore, either the desire for payback involves an immoral obsession with the relative status of others, or it displays an irrational disposition to produce futile suffering. Since both traits are vicious, the fact that anger’s conative dimension inherently disposes us to retributive action means, Nussbaum concludes, that anger is intrinsically immoral. This initial worry is doubtful, however, because it is unclear that such a tight connection really obtains between anger and retribution. When a friend betrays me, it is perfectly conceivable that my ensuing anger might motivate me simply to seek his recognition that he has wronged me. This desire for recognition has no essential connection to the desire for revenge. What this suggests is that anger is not inherently associated with payback. Though anger may sometimes cause revenge, it is not itself essentially vindictive. So, even accepting for the sake of argument that retribution is inherently immoral, it would not follow that anger is intrinsically immoral. Still, there is a more pressing worry for the non-­consequentialist defence of anger. Even if Srinivasan is right that anger has intrinsic moral value, the non-­ consequentialist strategy does not allow us to circumvent the counterproductivity objection altogether. As Srinivasan herself recognizes, the non-­consequentialist reasons for expressing anger do not have absolute force. In other words, the aptness of anger gives us a pro tanto moral reason to express apt anger, but this moral reason could in principle be overridden by countervailing moral reasons. The counterproductivity objection purports to give such countervailing moral reasons: if publicly expressing anger is indeed counterproductive, this gives us pro tanto moral reasons not to do so. And if expressing anger is highly counterproductive—if it is bound, as Nussbaum suggests, to impede crucial social reforms—then these consequentialist reasons against expressing anger may well outweigh the non-­consequentialist reasons Srinivasan invokes. To illustrate: if expressing rage really would have caused Abraham Lincoln to fail in his effort to have slavery abolished, and Martin Luther King to fail in his leadership of the civil rights movement, then one might plausibly think that, overall, caring about justice required that they refrain from expressing anger. Correspondingly, if expressing anger really is extremely counterproductive, it might seem that enraged orators like Douglass and Malcolm X were overall wrong, from the perspective of justice, to express themselves as they did.

37  Nussbaum (2015, 41–51).

58  Democratic Speech in Divided Times So, although I agree that anger has non-­consequentialist value, we should not stop there. To strengthen the case for publicly expressed anger, the rest of this chapter therefore aims to tackle the counterproductivity objection more directly.

3.3  Two Types of Consequentialist Strategies One might worry that directly resisting the counterproductivity objection’s consequentialist claim is essentially an empirical task, which should devolve to social scientists. But philosophers can also play a part. For one thing, consider that when social scientists investigate the empirical consequences of public expressions of anger, they do not look randomly, but instead proceed with some theoretical hypotheses regarding what types of effects might reliably be related to anger. The task of developing such hypotheses is one that philosophers can fruitfully contribute to. By analysing emotions’ phenomenology, intentionality, and relation to epistemic values, philosophical accounts of emotions shed light on what kinds of consequences we can expect emotions to produce. Accordingly, one philosophical strategy for tackling the counterproductivity objection consists in identifying and conceptualizing possible positive consequences of expressing anger. A second way in which philosophers can contribute is by examining what the correct context for evaluating the consequences of angry speech is. The context of evaluation is an account of what alternative things are being compared along a given standard of evaluation. Suppose our aim is to motivate others to oppose pressing social problems (such as racial injustice). And suppose, moreover, that we have shown that publicly expressing anger sometimes motivates listeners to do so. A critic of anger might acknowledge this finding, but reply that it is beside the point. The question is not: does angry speech motivate individuals to oppose social problems better than no speech at all? Rather, the question is: does angry speech motivate individuals to oppose social problems better than the best form of non-­angry speech? Importantly, the criticism is not that angry speech produces a certain kind of (bad) consequence. Instead, it is that advocates of anger are taking the wrong things as their objects of comparison. In this way, examining the context of evaluation constitutes a distinct strategy through which philosophers can challenge claims about the productivity or counterproductivity of public anger. I will pursue both strategies in tandem. Section  4 articulates a vital and neglected epistemic consequence associated with public expressions of anger. In turn, Section 5 argues that critics of anger typically adopt the wrong context for evaluating public anger’s consequences.

Voicing Anger  59 Now, in arguing for these conclusions, my aim is of course not to show that publicly expressing anger will always be productive, all things considered. But it is to indicate that the counterproductivity challenge rests on at least two illicit moves: first, it overlooks a distinctive and important epistemic value of communicating anger; second, by adopting the wrong context of evaluation, it forces us to choose between these epistemic benefits and other values when doing so may be unnecessary. Establishing these conclusions will put us in a far better position to resist Nussbaum’s charge that ‘anger is always normatively problematic’,38 along with her recommendation that formal and informal political institutions be governed by norms of non-­anger.

4.  The Epistemic Value of Publicly Expressed Anger Expressing anger can contribute to producing various kinds of goods. It can motivate others to oppose injustice.39 It can help one retain one’s self-­respect.40 And it can enhance one’s perceived status.41 Though important, these positive consequences are not my focus here. Instead, I will specifically investigate the epistemic value of communicating one’s anger. In part, this is because this dimension is particularly relevant to my broader inquiry. At various junctures, I have emphasized that inclusive public discourse matters, among other things, because it contributes to pooling information that is relevant to good policymaking. In particular, inclusive public discourse can help to identify social problems, as well as solutions to those problems.42 This, I will now suggest, is true in the specific case of angry public discourse. Public expressions of anger have a distinctive ability to reveal injustices, and to promote understanding of injustices, when these may otherwise have gone unnoticed or been misunderstood. The view that anger and its public expression are epistemically valuable is not uncommon. Recall, for example, the invisible narrator’s assertion that his anger aims to make visible what had been overlooked, and Lorde’s reference to her anger as a ‘spotlight’. Furthermore, this view has been taken up by philosophers, particularly feminist philosophers, as part of a broader trend towards rehabilitating emotions within epistemology.43 Nevertheless, the epistemic value of publicly expressing anger has yet to be articulated adequately. Some philosophers who discuss anger’s epistemic worth 38  Nussbaum (2016, 5). 39  Lorde (1997, 280). 40  Bell (2009, 168); Borgwald (2012). 41  Tiedens (2001). 42  Introduction; Chapter 1, Section 5. 43  E.g., Frye (1983), Jaggar (1989), Bell (2009), Chakravarti (2014, ch. 5). On this epistemological trend, see Brun et al. (2008).

60  Democratic Speech in Divided Times are not primarily concerned with the value of publicly communicating anger.44 And most of the defences that are concerned with publicly communicating anger are largely programmatic, rather than precisely developed.45 These limitations have enabled critics of publicly expressed anger to acknowledge that it has some epistemic value, while downplaying the relative importance of that value. To pre-­ empt such replies, I will explore the epistemic value of publicly expressed anger in a way that makes it clearer why this epistemic role matters, and why it cannot easily be performed without anger.

4.1  The Theoretical Case Some epistemic benefits of publicly expressing anger are relatively uncontroversial. When public speakers express their demands with sincere anger—either by displaying the conventional physiological and vocal signs of anger, or by explicitly affirming that they are angry—we learn that they are angry. Given an understanding of what anger is, this reveals many things about their psychology. We learn that they are experiencing certain bodily feelings, like tenseness, that they have the sense that something is seriously unjust, and that they might be driven to action. Although providing this information is an important epistemic function, two worries arise. First, one might reply that this function could largely be performed without expressing full-­blown anger. Nussbaum, for instance, might again point to transitional anger. The entire cognitive content of transitional anger is ‘How outrageous. Something should be done about that.’46 So, expressing transitional anger, as King did, suffices to communicate the belief that a situation is unjust, and the disposition to act on this basis. Accordingly, although Nussbaum acknowledges that non-­transitional anger has some epistemic value as a signal of outrage and injustice, she readily downplays the significance of this value.47 Similarly, Pettigrove remarks that refraining from anger did not prevent Lincoln from expressing his moral opposition to slavery.48 For example, in a letter to a political opponent, Abraham Lincoln declares: ‘You think slavery is right and ought to be extended; we think it is wrong and ought to be restricted. For this, neither has any just occasion to be angry with the other.’49 Second, remember that we set out trying to illuminate the intriguing observations Ellison’s invisible narrator made about anger, observations which Lorde and Baldwin echoed. But the above account of publicly expressed anger’s epistemic value does not fully capture what they meant. The invisible narrator declares that 44  E.g., Jaggar (1989), Bell (2009). 45  E.g., Lorde (1997), Frye (1983), Chakravarti (2014). 46  Nussbaum (2016, 6). 47  Nussbaum (2016, 211–12). 48  Pettigrove (2012, 355). 49  Lincoln (1860, 152).

Voicing Anger  61 his anger teaches us something about the world, shining a light on nightmarish features we had overlooked. So far, however, the epistemic function of anger is merely that it teaches us something about the speaker’s psychology, not about the world she is looking at. Nor, moreover, does the present account make sense of Baldwin’s suggestion that useful rage works by making listeners feel what the speaker feels. To isolate an epistemic function of publicly expressing anger that can overcome these concerns, let us take a step back. Chapter  1 discussed how, in non-­ideal conditions marked by deep divisions, logical arguments from shared premises sometimes break down. Often, the reason they do so is that some social groups lack certain concepts or have erroneous views about the application conditions of certain concepts. This, in turn, frequently results from the fact that they lack relevant experiences of the world. In this context, I argued, emotionally charged narratives can remedy such communicative breakdowns, by facilitating the communication of one’s experiences and perspective to others. Applied to anger, the idea is that anger-­ infused discourse is epistemically important not just because it tells the audience that the speaker is angry, but also because it can help them imaginatively experience what it is like to be in the speaker’s shoes, how the world feels or appears from where they stand. Put differently, anger-­ infused narratives can enable the audience to empathize more fully with the speaker.50 How does this work? The first thing to notice is that publicly expressing anger can be a means of transmitting anger to one’s audience. Suppose A is describing an unjust state of affairs—say, slavery—to denounce it. But instead of simply enumerating facts about slavery, A expresses his intense anger, with the aim of arousing similar emotions in the audience. In doing so, A is making use of the propensity for human beings to resonate with, or ‘catch’, the emotions of others. As Chapter  1 mentioned, Hume famously observes this general tendency: in human beings, he suggests, ‘all the affections readily pass from one person to another’.51 This phenomenon is well documented in contemporary psychology, under the heading of ‘emotional contagion’. Human beings tend to mimic the expressions, vocalizations, and postures of others, in a way that leads them to experience the emotions of others. This process is typically automatic, involuntary, and largely unconscious.52 Importantly, this effect also holds with angry speech, specifically. 50  For the identification of empathy and imaginative perspective-­taking, see Matravers (2017, 1–2). 51  Hume (2009[1738], 3.3.1.7). 52  For discussion, see Barsade (2002) and Coplan (2011). Note that there can be other, more conscious, ways of coming to share others’ emotions. Sometimes, Goldman (2011) suggests, we first imagine someone’s perspective, then make an evaluative judgment on this basis, and this contributes to generating a relevant emotion. But I focus on emotional contagion for two reasons. First, as Coplan (2011, 8) suggests, it is more reliable as a mode of emotional transmission. Second, the fact that the contagious transmission of emotion is largely unconscious is important for my defence of anger: it means that individuals can catch others’ anger prior to judging that an injustice is involved in the

62  Democratic Speech in Divided Times One study, for instance, finds that when speakers raise their voices and accelerate their speech, ‘this is likely to raise the listener’s blood pressure and feelings of anger’.53 In short—and here we have an echo of Baldwin’s thoughts about rage— given the tendency for emotional contagion, infusing one’s narrative with anger can cause one’s audience to feel that anger. Consider, for example, Douglass’s indictment of slavery. Douglass, one historian records, was known for his fiery passion: ‘when he spoke, he roared, his booming baritone complemented by waving arms’. And the fire was infectious. Douglass deliberately used this anger to ‘to make people feel—viscerally—the bloody horrors of slavery’, ‘to provoke his listeners [. . .] to furious outrage’.54 Nowhere was this effect more clearly reported than in William Lloyd Garrison’s account of Douglass’s first public speech: I shall never forget [. . .] the extraordinary emotion it excited in my own mind— the powerful impression it created upon a crowded auditory [. . .] I think I never hated slavery so intensely as at that moment; certainly my perception of the enormous outrage which is inflicted by it [. . .] was rendered far more clear than ever.55

What the passage eloquently reveals is the contagious nature of Douglass’s anger, which spreads not just to Garrison, but to the audience generally. Publicly expressing anger about x, then, can help one’s audience experience anger about x. But why does this matter epistemically? As I noted earlier, philosophers of emotion commonly argue that emotional deliverances have a distinctive phenomenology, or felt quality. They are distinctive ways of seeing the object that they are directed at.56 Hence, conveying one’s anger to one’s listeners helps them imaginatively undergo a distinctive qualitative experience. More specifically, the audience’s perspective-­taking, or empathy, is rendered more complete

rele­vant situation. This matters because I will argue that transmitting anger can help one’s listeners register injustices that they had previously overlooked. 53  Siegman et al. (1990, 641). For tentative evidence of contagion from angry facial expressions, see Blairy et al. (1999, 35), Hess and Blairy (2001, 138–9), Friedman and Riggio (1981, 102–3). 54  Oakes (2007, 90–3). 55  Garrison (1845, iv), emphases added. 56  Nussbaum and Pettigrove dismiss the distinctive epistemic value of anger partly because they overlook this phenomenology. It is widely held that the distinctive phenomenology of emotion is irreducible to the phenomenology of a judgment, or even of a judgment/desire pairing. E.g., Deonna and Teroni (2012, 66–71), Goldie (2002, 73–4). But both Nussbaum (2015, 42–5) and Pettigrove (2012, 357–8) broadly take anger about x to be reducible to a judgment/desire pairing, involving roughly a judgment that x involves some moral violation, and a desire to retaliate against its perpetrators. Now, Nussbaum does try, elsewhere (2001, 65), to accommodate the rich phenomenology of emotions within her judgment theory of emotions. However, Ben-­Ze’Ev (2004, 454–7) and Deonna and Teroni (2012, ch. 5) forcefully rebut her attempt. Further, even if her account could accommodate emotional experience’s complexity, the problem remains that Nussbaum’s brief discussion of anger’s epistemic value does not integrate these insights. She simply stresses that anger signals the judgment that something is morally amiss and that something should be done about it (2015, 55–6).

Voicing Anger  63 by emotional contagion. As the audience tries to adopt speaker A’s perspective, their imaginative reconstruction is altered by the anger they have infectiously gained from A. Not only is Douglass’s audience imagining the facts about slavery he is reporting, but the felt quality of their imaginings is coloured by the anger he has transmitted to them. For our purposes, the fact that the audience’s imaginative reconstruction of the speaker’s perspective is coloured by anger matters only if this ‘colouring’ is epistemically valuable—that is, only if the phenomenology of anger enriches our understanding of the anger’s object. To appreciate how it does so, let us consider two aspects of what philosophers of emotion have argued characterizes the phenomenology of emotional representation. The first relevant feature is that emotions are sources of salience: they draw our attention to certain aspects of a situation, and thereby place some properties of that situation into the foreground.57 In this, we have a way to understand Lorde’s reference to anger as a ‘spotlight’. Emotions help us navigate complex environments by selecting features of the environment and highlighting them. In doing so, they ‘rende[r] previously ignored features and previously unknown patterns salient’.58 Which features do emotions call our attention to? Recall that the cognitive fittingness of an emotion depends on whether its object involves certain evaluative properties: loss for grief, danger for fear, injustice for anger, etc. Hence, one intuitive proposal is that a given emotion puts a spotlight on aspects of its object that are liable to ground those evaluative properties. Take the case of fear. If I am walking home at night and am afraid, my fear highlights features of the environment that may make it dangerous: the street’s emptiness, the absence of lighting, and the footsteps behind me.59 By extension, we should expect anger to highlight patterns of behaviour that are liable to ground injustices. Putting these observations together, we have a first sense in which anger’s phenomenology is epistemically valuable. Anger renders salient properties of our situation that are liable to ground injustices. Because the environments we navigate are extremely complex, and because they often contain obstacles that conceal or distract us from injustices, we may otherwise fail to notice these properties. Thus, perceiving a situation with anger can make us notice an injustice that we would otherwise have overlooked. Furthermore, having our attention directed to the properties that ground a given injustice can help us see why something is unjust. Hence, the salience role of anger is epistemically valuable not only because it can yield knowledge that an injustice is occurring, but also because it can facilitate a greater understanding of the nature of that injustice. 57 On the salience role, see Elgin (2008, 43–6), Deonna and Teroni (2012, 122), and Brady (2013, 16–25). 58  Elgin (2008, 45). 59  Elgin (2008, 43–4).

64  Democratic Speech in Divided Times The distinctive phenomenology of emotions makes them epistemically valuable in a second respect: emotional representations of objects allow us to register evaluative properties in a way that can be more fine-­grained than our evaluative concepts would allow.60 Julien Deonna and Fabrice Teroni clearly expound this second point. In the same way that we can visually discriminate thousands of shades of colour for which we simply lack the corresponding concepts [. . .] we may surmise that the sensitivity to evaluative properties that [emotions] authorize is more fine-­grained than the discriminations that evaluative judgments provide for. The idea would be, for instance, that the intensity of one’s fear co-­varies with the degree of danger one faces, something our comparatively coarse evaluative judgments may prove unable to capture.61

To reformulate, the suggestion is that emotional experience is similar to perceptual experience in the following way: it can involve discriminations of evaluative properties that are more fine-­grained than our existing evaluative concepts would allow, just as visual perception allows us to discriminate between more shades of blue than we currently have concepts for. This claim seems intuitively true, insofar as it accurately captures our experience of emotion. But there is a further reason to accept it. The fact that emotional experiences can involve such fine-­grained evaluative discriminations helps explain why emotions play an essential role in the process of acquiring and  mastering evaluative concepts.62 Emotions could not play this widely acknowledged developmental role unless the evaluative nuances offered by emotional experiences could be more subtle than those allowed by our existing concepts. Again, this seems analogous to the case of visual perception. The fact that we can visually discriminate between many shades of blue prior to having access to corresponding concepts is part of what explains how we first develop and master numerous colour concepts.63 What does this mean for anger? The idea is that our experiences of anger can make us perceive or sense injustices that our pre-­existing conceptual frameworks did not allow us to grasp. In turn, this emotional sensitivity facilitates the development of more nuanced moral concepts. As with the salience role of 60  This point is made most explicitly by theories that identify emotions with perceptions of evalu­ ative properties. E.g., Döring (2007), Tappolet (2011). For other theories that acknowledge these phenomenological resemblances between emotion and perception, see Deonna and Teroni (2012, 66–7), Elgin (2008, 36–7), and Brun et al. (2008, 15). In her general analysis of emotion, Nussbaum (2001, 65) uses language reflecting this analogy. 61  Deonna and Teroni (2012, 66–7). 62  On the role of emotions in conceptual development, see Deonna and Teroni (2012, 84), Brun et al. (2008, 20–1), Tappolet (2011, 126–7), and Jaggar (1989, 166–8). 63  Peacocke (2001).

Voicing Anger  65 emotions, then, emotional sensitivity to fine-­grained evaluative differences can not only help us first recognize that injustices are occurring, but can also advance our understanding of those injustices, by enriching our conceptual resources for thinking about them. To illustrate, consider the example, which Fricker discusses at length, of Carmita Wood. Wood was continually subjected to unwanted sexual advances by one of her male colleagues. Though she felt unable to say why at the time, Wood’s deep indignation and bemusement made her sense that she was being treated wrongly. Reflecting on these responses in a consciousness-­raising group then enabled her and others to develop the concept of sexual harassment.64 Thus, Wood’s emotional sensitivity allowed her to recognize that she was being treated wrongly before she could even name the wrong in question. And that in turn helped her acquire the conceptual resources needed to understand this wrong. Armed with the concept of sexual harassment, Wood was better equipped to explain how she had been wronged, and reliably to recognize other wrongdoings of the same type. In sum, recent developments in the philosophy of emotion yield a theoretical account of two ways in which publicly expressing anger, by inspiring anger in one’s interlocutors, helps them recognize and understand injustices when they would otherwise have struggled to do so: first, by rendering certain morally relevant properties salient which they had previously overlooked; second, by yielding perceptions of injustice that are more fine-­grained than their existing moral concepts would allow. These two roles can come apart. Someone might already have extremely refined moral concepts, but might fail to apply them correctly because some non-­moral properties of their environment are not salient to them. Conversely, someone might be aware of all the relevant non-­moral properties in their environment but have very crude moral categories, which prevent them from seeing or fully understanding existing injustices.

4.2  Two Historical Illustrations The foregoing account of the epistemic value of publicly expressing anger has been very abstract. To make it more concrete, let us consider two historical examples of political anger. The first focuses on Frederick Douglass’s denunciation of slavery. Before proceeding, note what this example is intended to illustrate. We have already seen, in the previous section, how Douglass used his gestures, tones, and rhetoric to express anger and thereby contagiously rouse his audience to

64  Fricker (2007, 149–51).

66  Democratic Speech in Divided Times anger. This, recall, was vividly reported in Garrison’s testimony, where he described the ‘extraordinary emotion’ Douglass’s speech excited in him. What I want to illustrate now is how the anger Douglass transmitted to his audience altered their perception of slavery in the epistemically beneficial way outlined in the previous section. Ideally, to do so, we would consider Garrison’s own testimony regarding how the anger he had acquired from Douglass modified his moral perception of slavery. However, Garrison’s testimony is not very precise when it comes to describing exactly how the anger Douglass’s speech excited in him changed his felt experience of slavery. He merely indicates that this anger made his perception of slavery’s wrongness ‘far more clear’. By contrast, Douglass reports in extremely rich detail how experiencing anger transformed his own perception of slavery. So, to illustrate my theoretical account of anger’s epistemic value, I will focus on Douglass’s testimony regarding his own anger. By considering how Douglass’s anger affected his own perception of slavery, we can learn about how being roused to anger by Douglass’s angry political speech may similarly have enhanced Garrison’s (and other audience members’) understanding of slavery’s injustice. With this clarification in mind, consider Douglass’s account of when, still a slave, he learned to read: The more I read, the more I was led to abhor and detest slavery, and my enslavers. [. . .] I loathed them as the meanest and the most wicked of men [. . .] Liberty! The inestimable birthright of every man had, for me, converted every object into an asserter of this great right. It was heard in every sound, and beheld in every object. It was ever present, to torment me with a sense of my wretched condition. [. . .] My feelings were not the result of any marked cruelty in the treatment I received; they sprung from the consideration of my being a slave at all. It was slavery—not its mere incidents—that I hated. [. . .] The feeding and clothing me well, could not atone for taking my liberty from me. The smiles of my mistress could not remove the deep sorrow that dwelt in my young bosom.65

Douglass clearly experiences intense anger: he ‘abhor[s]’, ‘detest[s]’, ‘loath[es]’, and ‘hate[s]’ slavery and slaveholders. But he is not just telling us that he is angry. Instead, he is also reporting his felt experience of anger at slavery, how slavery appears from his angry perspective. Accordingly, he stresses how every object’s appearance was ‘converted’ by his anger, and repeatedly employs perceptual language. What does Douglass’s anger teach him about slavery? First, Douglass’s anger draws his attention to how all things, including non-­human animals, are free and 65  Douglass (1855, 159–61).

Voicing Anger  67 independent in virtue of an ‘inestimable birthright’. By rendering salient the ubiquity of freedom, then, Douglass’s anger highlights his own degraded status. In turn, this helps him better understand the nature of slavery’s wrongness. The idea is that it is the status of slavery that is unjust, the very condition of depending on an arbitrary master, independently of that master’s oppressive actions. In Douglass’s words: it is ‘slavery—not merely its incidents’—that is wrong. Indeed, as he reports, by the standards he had previously been accustomed to, Douglass’s masters at this time were relatively kind. Thus, Douglass’s angry perspective helps him perceive more precisely the ground of slavery’s wrongness, which concerns the relative standing or status of slaves. This is quite a fine-­grained understanding of the injustice of slavery and of the value of freedom, which arguably anticipates contemporary conceptualizations of freedom as non-­domination, and which Douglass (as well as his audience) may otherwise have been unable to grasp. Hence, Douglass’s testimony regarding his anger vividly illustrates how anger can both highlight morally relevant properties of one’s environment (e.g., the comparative independence of other living beings), and enable a more fine-­grained understanding of injustice than may have been allowed by one’s prior moral categories (e.g., by making one sense the relational injustice of slavery, domination). We may surmise that being roused to anger by Douglass’s angry political speech had a similarly beneficial phenomenological effect on Douglass’s audience. This helps us understand more precisely what Garrison may have been referring to when asserting that the anger gleaned from Douglass’s speech made his understanding of slavery’s injustice ‘far more clear’. One might object that Douglass’s anger here is too close to ‘transitional’ anger to have much dialectical force against Nussbaum. But Douglass’s anger in fact differs from the standard case of transitional anger. His reference to loathing, abhorrence, and hatred suggests a harsh and fiery form of anger, rather than a calm and restrained one. Moreover, he focuses on the injustice itself, rather than on the remedy. This difference is crucial: because transitional anger characteristically does not dwell on the injustice itself, it is ill-­suited to helping us understand the nature and depth of that injustice. Admittedly, though, Douglass’s anger here does seem qualified in some respects. In particular, the conative dimension of his anger here does not seem to involve anything like retribution or violence. Hence, let us turn to Malcolm X for an even more intense expression of rage. In his speech ‘The Ballot or the Bullet’, Malcolm X denounces black Americans’ lack of economic and political opportunities: We’re trapped, trapped, double-­trapped, triple-­trapped. Any way we go, we find that we’re trapped [. . .] So today our people are disillusioned. They’ve become disenchanted. And in 1964 you’ll see this young black man, this new generation,

68  Democratic Speech in Divided Times asking for the ballot or the bullet. That old Uncle Tom action is outdated. The young generation don’t want to hear anything about ‘the odds are against us’. What do we care about the odds? [. . .] When we open our eyes today and look at America, we see America not through the eyes of someone who has enjoyed the fruits of Americanism. We see America through the eyes of someone who has been the victim of Americanism. We don’t see any American Dream. We’ve experienced only the American Nightmare. We haven’t benefited from America’s democracy. We’ve only suffered from America’s hypocrisy. And the generation that’s coming up now can see. And are not afraid to say it. If you go to jail, so what? If you’re black, you were born in jail.66

Like Douglass, Malcolm X’s speech expresses deep anger. And like Douglass, he is reporting what it is like to be enraged at America’s racially discriminatory practices, what one sees when one looks through the eyes of an outraged black American. Unlike Douglass, however, his anger does involve the conative dimension Nussbaum is wary of: it explicitly threatens violence (‘the bullet’) which might well be retributive, born of a desire simply to retaliate (‘what do we care about the odds?’). Let us examine what Malcolm X’s expression of deep non-­ transitional anger tells us about how anger affected his perception of racial injustices. As in the Douglass example, doing so will by extension help us learn something about how being contagiously roused to anger by Malcolm X’s fiery political speech may have been epistemically beneficial for his audience. Malcolm X’s rage reveals, in place of the American Dream, a vision of the ‘American Nightmare’. Just as Douglass’s anger draws his attention to the freedom of other beings, so Malcolm X’s anger renders salient the pattern of deceit that black Americans have experienced: an economic ‘Dream’ that is not genuinely accessible to them; a political ‘democracy’ that they are effectively excluded from; and a civic freedom that seems no different from ‘jail’. By putting a spotlight on this pattern of disappointments, Malcolm X’s rage foregrounds important properties of the injustice black Americans encounter, properties that otherwise may have been obscured by the ideology of the American Dream. First, seeing the pattern suggests that these exclusions are not accidental or isolated but systemic, built into the principal political and economic institutions of American society. Moreover, experiencing racial exclusions as patterned and systemic also conveys a sense of their inescapability (‘we’re trapped’), and, consequently, of hopelessness at the prospect of achieving reform from within the American political system. Via its salience role, Malcolm X’s anger therefore yields a fuller understanding of the injustice at hand.

66  Malcolm X (1964).

Voicing Anger  69 Rendering these properties salient, finally, arguably helps refine the audience’s prior evaluative categories: it contributes to adjusting dominant conceptions of what kinds of political agency count as reasonable. By highlighting patterns of exclusion, Malcolm X’s anger helps make rational sense of the apparently unreasonable attitude which consists in embracing violent action, even when the odds of success are unfavourable. If reforming the American system from within is hopeless, and if one’s situation is no better than jail, then violence aimed at putting one’s opponents in their place may seem the best option. In sum, these examples illustrate my theoretical account of how experiencing anger can enhance one’s understanding of existing injustices and of the reactions they elicit. When we put this together with the fact that publicly expressing anger can transmit anger to one’s audience by contagion (as happened with Garrison), this indicates how publicly expressing anger can enhance the moral understanding of one’s audience.

4.3  Objections to the Epistemic Productivity of Anger I have already considered some worries concerning anger’s epistemic productivity in Chapter  1, when recommending giving a greater public role to emotionally charged forms of speech more generally. In particular, I raised the common worry that emotions can be manipulated by skilled orators, who use impassioned appeals to excite misguided emotions. In the case of anger, we are all too familiar with this phenomenon. Demagogues sometimes express anger at immigrants or racial and religious minorities, in near-­total disregard of facts, and thereby arouse misguided rage in their listeners. In her sociological study Strangers in their Own Land: Anger and Mourning on the American Right, for instance, Arlie Russell Hochschild argues that Donald Trump’s rise to power was facilitated by his ability to provide an outlet for the anger of white working-­class Americans, and by his success at turning this anger against vulnerable scapegoats. His success at doing so, she suggests, was itself enabled by widespread factual misconceptions about these vulnerable groups.67 The upshot is that, even if angry discourse can be epistemically beneficial when the expressed anger is fitting or well directed, making space for anger in the public sphere risks opening the door to dangerous expressions of unfitting or misdirected anger—especially against a background of political ignorance. And since unfitting anger represents just situations as unjust, or innocent people as responsible for injustice, its public expression seems epistemically misleading, not epistemically useful.

67  Hochschild (2016, ch. 15, appendix C).

70  Democratic Speech in Divided Times There are several things to note in response. In the first place, unfitting or misdirected anger may sometimes retain some epistemic value. Fittingness comes in degrees. While some instances of anger are wholly misdirected—for example, a white supremacist’s rage at having black neighbours—unfitting anger is often only partly so. Some occurrences of anger, for example, might correctly represent a situation as involving injustice while representing the wrong group as blameworthy for this injustice. If the injustice in question is not known or understood by the society at large, expressing this partly unfitting anger may contribute to advancing our understanding in the way I have theorized. In fact, Hochschild’s study itself brings out this point. While the anger her subjects express is arguably directed at the wrong agents (e.g., minority groups), Hochschild vividly depicts how listening to their anger rendered salient an underappreciated pattern of hardship (rising joblessness, pollution-­ related illnesses, environmental decay, cultural vilification, etc.). What is striking is how, notwithstanding their mixed fittingness, these angry narratives effectively bring into view an unjust pattern of broken socio-­economic promises—in a way that, in some respects, echoes Malcolm X’s disillusion with the American Dream. In this way, expressions of partly misdirected anger can lead us to register and examine previously overlooked injustices, whose real source we may subsequently identify. This is simply to say that we should not recoil too quickly at the idea of admitting public expressions of unfitting anger: anger that is partly misdirected may retain some of the epistemic value I have theorized. What about wholly unfitting anger—e.g., the white supremacist’s rage at having black neighbours—which is solely epistemically misleading? There may be some strategies internal to democratic public discourse for mitigating its epistemic harms. In Chapters 3 and 4, I will examine utterances that express severely misguided viewpoints—viewpoints that either vilify vulnerable groups (Chapter 3) or that promote misinformation about political issues (Chapter  4). And I will develop an account of how we should go about countering these utterances with more speech. Specifically, I will suggest that authoritative state-­backed speech that positively and continuously reaffirms the truth can go some way towards blocking the harms that these misguided forms of speech would otherwise produce. This suggestion arguably extends to the epistemic harms involved in deeply misguided expressions of anger. By positively and persistently affirming a correct vision of the world, authoritative speakers can expose as ignorant or prejudiced the factual presuppositions upon which wholly unfitting anger is based, and thereby loosen the influence of this anger on the public. Now, for reasons that we will examine in these two chapters, there may be contexts where this discursive strategy is unlikely to be effective. Even so, a final and more fundamental response is available. In Chapter 1, in response to the concern about misdirected emotions, we saw that non-­emotional forms of discourse are

Voicing Anger  71 actually vulnerable to a similar problem.68 In conditions where reliable information is scarce or complex, one of the most successful forms of sophistry involves endowing fallacious arguments or misleading testimony with the appearance of scientific rationality or expertise. We will, in fact, revisit this problematic phenomenon in Chapter 4. What matters, in the meantime, is this: we standardly do not think that, simply because dispassionate argument and expert testimony can be misused in public discourse, we should forego these forms of speech altogether. Admittedly, this last response is an unhappy one for democratic discourse more generally: it concedes that political manipulation, which subverts otherwise epistemically useful modes of discourse, may be widespread in conditions of public ignorance. In Chapter 6, therefore, I will consider head-­on the democratic problem posed by public ignorance. However, the present response does suggest (1) that the problem at hand is not a special problem afflicting angry speech and (2) that just as we do not reject argumentative reasoning and expert testimony simply because they can be abused in non-­ideal conditions, so too we should not abandon the use of anger in political discourse. But publicly expressed anger faces further objections. While the previous worry concerned the possibility that anger might be misdirected, another goes further and interrogates the claim that fittingly directed or apt anger is epistemically desirable. Perhaps it is true that such anger can render certain facts more salient, namely those that are liable to ground injustices. But, according to Pettigrove, empirical research suggests that anger also obscures important facts. ‘When angry, people are more likely to see what they take to be hostile stimuli than they are to notice features of their environment they do not take to be hostile.’69 Relatedly, ‘people who have become angry are generally less responsive to counterevidence’.70 So, even when it is directed at real injustices, anger risks yielding a distorted perspective on the world. However, I am not recommending that all public discourse be angry. Rather, I am defending the view that angry discourse should play an important role in democratic public speech. The idea is not that angry discourse should supplant the countervailing perspectives that non-­angry speech communicates, but that it should complement them. It enriches countervailing perspectives, bringing into view features of the social context that they typically overlook or struggle to accommodate, but it does not replace them altogether. Therefore, the fact that anger highlights neglected, morally relevant facts while obscuring others is relatively unproblematic: non-­angry speech can communicate those other facts. Note that, in saying that angry discourse is an epistemic complement, I am not downplaying its importance. Casting a spotlight on injustices might seem

68  Chapter 1, Section 5.3.

69  Pettigrove (2012, 363).

70  Pettigrove (2012, 365).

72  Democratic Speech in Divided Times unimportant in conditions where injustices are scarce, or where injustices are fully recognized and understood by all. But non-­ideal conditions are not like this. Contemporary democracies typically involve grave intergroup injustices. And those injustices are often concealed by ideological barriers (as when the ideology of the American Dream concealed structural barriers to black Americans’ social mobility) or distorted by hermeneutical deficits (as when the conceptual resources needed to understand the wrong of sexual harassment were lacking). In conditions such as these, the angry perspective’s fixation on potential sources of injustice serves as a crucial epistemic corrective. The reason Pettigrove overlooks the possibility that anger might play such a complementary epistemic role is partly that his focus differs from mine. He is primarily concerned with determining whether anger as a standing disposition of character is more virtuous than non-­ angry dispositions, like meekness. By contrast, I am suggesting that, among other forms of speech, public discourse should involve expressions of anger that are liable to arouse episodes of anger in listeners. While having a standing disposition to anger is incompatible with having a standing disposition not to get angry, experiencing an episode of anger towards x seems compatible with experiencing non-­angry attitudes towards x at another time. Episodes of anger need not be, nor typically are, permanent. So, angry speech and non-­angry speech might alternate, getting their listeners to see the world in different but complementary lights. A different objection questions the transmission of anger. My account of anger’s distinctive epistemic contribution depends on the claim that publicly expressing anger can help arouse anger in listeners. I have already cited evidence supporting the phenomenon of emotional contagion. However, one might worry that contagion, and the perspective-­taking it facilitates, have limits that are particularly salient in the non-­ideal conditions with which I am concerned. After all, doesn’t resonating with the speaker’s anger require that the audience identify with, or at least trust, the speaker? If it does, my argument may seem inapplicable in non-­ideal contexts, marked as they are by deep social divisions. As it stands, this objection is overstated. Expressing anger is not wholly ineffective in divided circumstances. As we have seen, Douglass’s rage successfully induced profound anger in Garrison, a white abolitionist. Furthermore, the transmission of anger is sometimes possible even between individuals with significantly different perspectives. Famously, Republican vice-­president Dan Quayle once declared that he had vicariously gained important insights into racial in­just­ ice from Malcolm X’s fiery autobiography: ‘I can see the hate that was there; I can see the bigotry; I can see it from his perspective.’71

71  Quayle (1992).

Voicing Anger  73 Nevertheless, once qualified, the objection is problematic. There is presumably some point at which social divisions and the attending mutual distrust become too great to allow anger to be contagiously transmitted. If group A deeply distrust group B, so that they doubt the sincerity or competence of members of B, it seems unlikely that they will resonate with their angry speech. For example, when prejudicial gender stereotypes represent women as epistemically untrustworthy, those who accept those stereotypes typically experience women’s anger not as a signal of some serious injustice to be investigated, but rather as a symptom of women’s supposedly hysterical natures. Instead of resonating with the anger, they experience it as alienating or absurd.72 Now, Chapter  5 will examine more closely how far antecedent trust and goodwill really are required for fruitful interpersonal communication. But for now, I assume—along with advocates of the counterproductivity objection, such as Nussbaum—that some degree of pre-­existing trust is needed for the effective transmission of anger, and that expressing anger will not increase the amount of trust there is. What can we say about cases where this required trust is lacking? First, Marilyn Frye notes that historically, distrust has fruitfully been tackled by challenging the demeaning stereotypes that underpin it. In the case of gendered distrust, ‘the struggles and victories of abolitionists, suffragists, prohibitionists, and other reformers made it relatively safe for women to get angry, publicly, on behalf of great moral causes’.73 Thus, taking measures to transform gender roles—for example, by altering media representations of women—can continue to reduce the distrust that prevents many women’s anger from being heard. Second, even when distrust persists, the possibility of ‘networked’ angry speech weakens its significance. Suppose that C deeply distrusts A, so that A cannot transmit her anger to C. But B trusts A, and C trusts B. In such conditions, A can communicate her angry perspective to B.  And B, in turn, can relay that angry perspective by expressing it to C. In this way, networked angry speech helps to circumvent the communicative barriers raised by distrust. Still, these replies have limits. While they may salvage the epistemic utility of publicly expressed anger in the long term, they do not fully ease our more immediate worries. Just as the affirmative policies Frye recommends may be slow in effecting change, so too relaying angry speech may be arduous. Contrary to the simple schema outlined above, the chain of networked angry speech can be very long in divided settings. Suppose Garrison tried to relay Douglass’s rage. Though Garrison was possibly less distrusted by slaveholders than Douglass, it seems unlikely that they trusted him enough to resonate with his angry speech. And slightly more progressive politicians, who trusted Garrison, might have come to

72  Frye (1983, 90–1); Bourgeois and Hess (2008); Lewis (2000).

73  Frye (1983, 93).

74  Democratic Speech in Divided Times distrust him because of his association with Douglass’s message. Hence, Garrison might have to convey the angry perspective to listeners who are even closer ideologically. Progress could be excruciatingly slow. What is worse, a final concern is that even among those who trust the speaker enough to resonate with her anger, publicly expressing anger may have countervailing bad effects. This was the point of Nussbaum’s counterproductivity objection: as we saw in Section  3.1, Nussbaum warns that expressing anger ‘increases the other party’s anxiety and self-­defensiveness’.74 Thus, voicing anger risks impeding subsequent communication and coordination with one’s listeners. Consider two ways in which this could happen. In the first scenario, P is entirely swayed by the angry perspective. But because the anger was intense, and expressed blame and perhaps openness to violence against P’s social group, P feels too unwelcome or afraid to coordinate with the speaker. Perhaps P is a white man who resonates with Malcolm X’s rage towards racial injustices, but is made anxious by his references to white people as ‘hypocrites’ and ‘devils’. Call P the ‘anxious ally’. Alternatively, perhaps Q has imaginatively experienced Malcolm X’s rage, but feels that it does not offer an entirely balanced picture of the situation. However, for the reasons just mentioned, she worries that, were she to mention her lingering doubts about the accuracy of this perspective, her interlocutors would shun her rather than deliberate with her. Call Q the ‘sceptical ally’. Despite its promise, then, the epistemic defence of anger’s productivity remains problematic, as shown in three types of cases. In some cases, the problem is that anger hardly seems to have good epistemic consequences. These are cases where the listener does not trust the speaker enough to resonate with her anger. In the other two types of cases—the anxious ally and the sceptical ally—the problem is that despite its good epistemic consequences, it is unclear whether these consequences outweigh anger’s countervailing bad trust-­related consequences. These are cases where the listeners trust the speaker enough to vicariously experience her anger, but (because of that same anger) not enough subsequently to engage in practical coordination or deliberation. In what follows, I will show that we can mitigate these remaining problems by reframing our assessment of anger’s consequences.

5.  The Systemic Defence of Publicly Expressed Anger Section 4 offered a defence of publicly expressed anger that focused on the kinds of consequences it can produce. I will now complement this defence by examining

74  Nussbaum (2016, 230).

Voicing Anger  75 which context of evaluation is appropriate for assessing anger’s consequences. The context of evaluation, recall, is an account of what alternative things we should be comparing along a given standard. Reconceiving the context of evaluation in light of the systemic conception of public discourse renders more tractable the three worries outlined at the end of Section 4.

5.1  The Systemic Approach Revisited As we saw in Chapter 1, the systemic approach insists that we should not think of public discourse as occurring within a single unitary arena that strives to include all relevant perspectives. Rather, we should think of it as occurring within a network composed of many discursive arenas that vary in size, function, and formality, and that are connected more or less directly to one another. Some arenas, like legislative houses, engage in deliberation whose primary function is to produce coercive decisions, and tend to be regulated by constraining formal norms (such as the requirement that, in the final stages of deliberation, participants appeal to shared reasons). By contrast, more informal arenas—such as rallies organized by social movements—are characteristically more concerned with helping participants form considered opinions.75 I have already argued, in Chapter  1, that the systemic approach to public discourse is attractive on various grounds. The case of anger brings out several more benefits of this approach. In the first place, notice that the systemic view of public discourse was implicit in my earlier discussion of networked angry speech. Networked angry speech, remember, occurs when A transmits an angry perspective to B, who then relays it to C, and so on. This process presupposes the existence of a plurality of discursive arenas (one including A and B, one with B and C, etc.) that are connected in virtue of the fact that their constituencies overlap. Thus, the systemic view has already played a role in my defence of publicly expressed anger. The second benefit, and the one I will be discussing in the remainder of this chapter, is that the systemic view renders more tractable the worries outlined at the end of Section 4, by inviting us to adopt a more adequate context of evaluation. To see this, consider how debates surrounding the public expression of anger are often framed. Opponents and advocates of anger often proceed by examining the properties of angry speech and the properties of non-­angry speech, and asking: which of the two better advances our aims? Is King’s discourse more effective at promoting justice than Malcolm X’s? Nussbaum clearly illustrates this tendency when she asks: ‘Imagining justice anthropomorphically, should justice

75  Chapter 1, Section 4.1.

76  Democratic Speech in Divided Times get angry at offenders?’ Whereas Nussbaum answers negatively,76 advocates of anger such as Srinivasan respond positively.77 But what both sides accept, here, is a particular context of evaluation: they compare an instance of angry speech to an independent instance of non-­angry speech, and they ask which of the two best advances our aim of combatting injustices and other social problems. The systemic approach brings into view what is wrong with this framing. From a systemic perspective, what we ultimately care about are the properties of the system as a whole. We want the overall system to be epistemically effective and accountable to the concerns of the people. That is, we want it to be good at identifying social problems that people face (such as grave injustices) and at producing policy outcomes that are responsive to these problems. Consequently, when evaluating parts of the discursive system, what fundamentally matters is not whether they exhibit these desirable properties when taken independently, but rather whether they contribute to promoting these properties at the systemic level.78 This difference is important. A part of the system that lacks a morally desirable quality when taken on its own may nevertheless contribute to promoting this same quality at the systemic level. As we noted in Chapter 1, for example, party conventions are generally not open to all, and their debates exclude some experiences and values. However, when seen as part of a broader system that involves other party conventions, such conventions can make the discursive system more epistemically inclusive. Having a plurality of partisan conventions facilitates the articulation and development of significantly different perspectives, which can then confront each other in national contests. Let us apply these lessons to the context of evaluation for angry speech. The principal lesson is that, at the end of the day, we should be concerned with the qualities of the discursive system as a whole. As a result, what we should be comparatively assessing is not an instance of angry speech and an instance of non-­ angry speech. Rather, it is a system in which a significant range of arenas welcome angry speech, and a system in which all arenas are governed by norms of non-­anger. The correct question is not: should we have Martin Luther King or Malcolm X? It is: should the discursive system contain some important arenas—rallies, debates, etc.—where speakers like Malcolm X are invited to publicly express their anger? Or should we instead prefer a system where, in all arenas, speakers strive to emulate King’s non-­anger? This systemic context of evaluation creates new possibilities for defending publicly expressed anger. As we have just seen, although partisan rhetoric 76  Nussbaum (2016, 173). See also Pettigrove (2012, 355). 77  Srinivasan (2018, 133, 142–3), for instance, disputes Nussbaum’s particular anthropomorphic representation of anger, and suggests instead that anger aptly reflects a commitment to justice. See also Bell (2009, 176–80). 78  Chapter 1, Section 4.1.

Voicing Anger  77 excludes certain perspectives when taken in isolation, conventions involving partisan rhetoric can contribute to greater epistemic inclusiveness when considered as part of a broader system. By analogy, even if, on its own, publicly expressed anger erodes the trust that is necessary for subsequent coordination and deliberation, it does not immediately follow that a system where some (but not all) discursive spheres welcome angry speech lacks this trust. So far, I have only shown that once we adopt a systemic context of evaluation, the fact that angry speech taken in isolation can have negative effects does not necessarily entail that a system that gives angry speech a prominent place in some arenas will produce these negative effects more than one that does not. But this only establishes the abstract possibility that angry speech might not be counterproductive. In the following section, I will flesh out this possibility, by revisiting the three problem cases introduced at the end of Section 4. Before proceeding, we should pause to note what it might mean, concretely, to have a system where some arenas welcome anger and others do not. In the case of transitional justice proceedings, this might involve having a division of labour between criminal trials—where speakers are discouraged from expressing negative emotions like anger—and Truth and Reconciliation Commissions— where speakers are invited to voice their angry grievances.79 In the case of formal legislative debates, such a system might distinguish between earlier deliberative stages, where discourse charged with intense anger is welcome, and later deliberative stages, where the presiding officer not only requires that arguments be framed in terms of shared considerations, but also demands that they be expressed in an orderly fashion. In the informal public sphere, the idea is that there should not simply be rallies, conventions, and debates where speakers (perhaps encouraged by a moderator) refrain from intense expressions of anger. The discursive system should also include a wide range of rallies, conventions and debates where expressions of deep anger are welcome. Alongside the non-­angry televised discussion between King and the segregationist James Kilpatrick, there should also be the Cleveland rally where Malcolm X was invited to deliver ‘The Ballot or the Bullet’, or the Harvard or Oxford Union debates, which gave him a platform to voice his intense anger. It is clear that a critic such as Nussbaum would reject this. As discussed in Section 3.1, Nussbaum considers anger to be ‘always normatively problematic, whether in the personal or in the public realm’.80 Thus, as we saw, she recommends that TRCs give expression to generosity rather than anger; and, accordingly, she praises the way in which Tutu discouraged resentful narratives during the South African TRC. Similarly, recall that in her discussion of informal political

79  Chakravarti (2014, 44–9); Mihai (2016, 28).

80  Nussbaum (2016, 8).

78  Democratic Speech in Divided Times contestation, Nussbaum upholds the non-­anger of King and Gandhi as an ideal that agents concerned with promoting justice should strive for here and now.

5.2  The Systemic Evaluation of Anger According to Mansbridge, different ‘parts of a system may have relationships of complementarity [. . .] Two venues, both with deliberative deficiencies, can each make up for the deficiencies of the other.’81 Let us consider how this complementary relationship functions in a system that contains sites where speakers express anger as well as sites where speakers refrain from expressing anger. More specifically, how does it help address the three problem cases outlined in Section 4? The benefits of the systemic perspective are more obvious in the two cases where, although angry speech successfully communicates certain experiences of injustice, it erodes trust in a way that makes subsequent coordination and subsequent deliberation more difficult. Take the anxious ally, who does not trust speakers in anger-­welcoming arenas enough to coordinate with them in the struggle against injustice. Here, the existence of arenas governed by norms of non-­anger offers a solution: such arenas give him the opportunity to engage in practical coordination against injustice. Imagine once more the white man who resonates with the anger Malcolm X expresses at the Cleveland rally but feels unwelcome, because of that anger, to coordinate with participants in this rally. Equipped with the epistemic insights about injustices gleaned from the anger-­ infused rally, he could then coordinate to oppose those injustices within meetings governed, as King’s were, by norms of love and openness. A similar observation holds with the sceptical ally, who, having vicariously experienced the angry perspective, seeks to engage in further deliberation regarding its accuracy, but is reluctant to deliberate with angry speakers. What she seeks is a venue where she can expect her interlocutors to listen to her thoughts and doubts about racial injustice. Arenas governed by expectations of non-­anger provide just such a venue. Indeed, the meetings, rallies, or church-­ gatherings organized by King’s followers adopted norms of love and non-­anger partly to facilitate dialogue between estranged social groups. In these two cases, the relationship of complementarity takes an additive form. Each venue independently performs a necessary function: the anger-­welcoming venues provide a distinctive epistemic good effect, by transmitting anger-­infused experiences; and the anger-­discouraging venues provide the trust and openness that facilitates further deliberation and coordination. Thus, when determining

81  Mansbridge et al. (2012, 3).

Voicing Anger  79 the overall properties of the system that contains sites that welcome anger as well as sites that discourage it, we can add their independent contributions, and thereby appreciate how it is preferable to the system that shuns anger. Understood additively, however, the complementary relationship between anger-­welcoming and anger-­discouraging venues is unhelpful when it comes to the remaining problem case. This is the case where listeners do not trust the angry speaker enough to resonate emotionally with her angry speech. Here, the problem is that angry speech taken independently fails to produce its distinctive epistemic good effect. Consequently, simply adding the independently achieved good effects of anger-­welcoming and anger-­discouraging venues will not reveal what is advantageous about a system that contains angry speech. To address this last problem case, we must consider the interactive dimension of the complementary relationship. This dimension concerns how different parts of a system interact to produce good effects that no part would produce independently.82 How might anger-­ welcoming and anger-­ discouraging discursive arenas interact fruitfully, and thereby help address the last problem case? One suggestion is that the presence of rallies and debates suffused with angry speech may help speakers in venues governed by norms of non-­anger acquire the trust of wary listeners. As Section 4.3 discussed, part of the reason why expressions of anger can make one seem untrustworthy is the existence of prejudicial stereotypes that associate anger with unreason. In such circumstances, the contrast between anger-­welcoming and anger-­discouraging venues can make speakers in the latter appear more reasonable and trustworthy. And this, in turn, can enable them to convey a perspective that they otherwise could not have communicated. By way of example, consider the deliberative venues inhabited, respectively, by Lincoln and Douglass. These, James Oakes argues, had contrasting but interdependent roles in the abolitionist struggle: the function of the Congressional debates Lincoln participated in was to effect policy changes; the purpose of venues like The North Star—the anti-­slavery newspaper published by Douglass and Martin Delany—was to denounce slavery’s injustice without restraint or compromise. Importantly, part of what made Lincoln’s contributions seem reasonable to other members of Congress, Oakes suggests, was the contrast between the legalistic calmness that characterized Congressional debates and the fiery moral discourse promoted in venues such as The North Star.83 In a similar vein, Lewis Baldwin contends that, for many white Americans, part of what made it conceivable to engage with civil rights activists who embraced non-­anger was the contrasting rage that animated Malcolm X’s rallies.84 Malcolm X explicitly acknowledged this: ‘if the white people realize what the 82  For a distinction that mirrors the distinction between additive and interactive complementarity, see Bächtiger and Parkinson (2019, 106–8). 83  Oakes (2007, 90–5). 84  Baldwin (1986, 408).

80  Democratic Speech in Divided Times alternative is, perhaps they will be more willing to listen to Dr. King’.85 Without the contrast with the deep rage expressed in black nationalist rallies, many white Americans may have deemed King’s vision of interracial love too extreme to be worth listening to. Oakes and Baldwin are highlighting a similar kind of interaction between parts of the discursive system. By altering perceptions of reasonableness, they suggest, the presence of anger-­ welcoming sites helped speakers in anger-­ free sites articulate their non-­angry viewpoints and receive uptake in a way that would otherwise have been impossible.86 Now, one concern is that, in this interactive relationship, the viewpoints whose successful communication is facilitated by the presence of anger-­ welcoming arenas are not the angry perspectives themselves. So it seems that the distinctive epistemic value of anger is lost. An initial reply to this concern is that the information conveyed by the angry perspective need not be entirely lost in the promoted non-­angry perspectives. A message infused with grief might serve to highlight the loss that results from injustice. And a message of unconditional love, like King’s, might mention the existence of injustice, insofar as what makes it loving is its willingness to move beyond a history of oppression and to embrace its perpetrators. Nevertheless, it is undeniable that some of the distinctive information carried by the angry perspective is lost. Because they are different emotions, grief and love highlight different evaluative properties than anger. Grief may highlight the loss resulting from injustices, but it does not highlight the injustices themselves. And while a loving message may mention injustice, it does not put a spotlight on those injustices or on their perpetrators’ responsibility. For instance, the point of King’s passing reference to the ‘sweltering oppression of Mississippi’, in his Dream speech, is less to communicate a sense of what it is like to be oppressed, and more to introduce a vision of what it is like to move beyond the divisions that produce oppression. Hence, examining the interactive relationship between anger-­welcoming and anger-­ discouraging venues in the discursive system only partly defuses the challenge posed by the last problem case. In this case, the listener will not resonate with the angry perspective and will not experience its distinctive phenomenology. However, this qualification should not obscure our main result: that even in this case, welcoming anger in some sites still has the potential to produce an epistemic benefit, albeit indirectly. To reiterate: the presence of anger-­welcoming sites may

85  Cited in Baldwin (1986, 398). 86  For discussion of how Malcolm X and King, as well as Douglass and Lincoln, facilitated this by strategically distancing themselves from each other, see Baldwin (1986, 402–3) and Oakes (2007, xvii–xviii).

Voicing Anger  81 facilitate the communication, in other sites, of valuable non-­angry perspectives, such as King’s loving vision. By way of concluding this systemic evaluation of angry discourse, an im­port­ ant qualification is needed. When exactly the productive interdependencies I have outlined obtain is, at bottom, an empirical matter which cannot be settled definitively without input from social science. Accordingly, while the examples of the abolitionist struggle and of the civil rights movement arguably illustrate such interdependencies, my point is not that they will definitely always obtain. Rather, as I explained in Section 3.3, I am making the weaker but nonetheless important point that switching to a systemic context of evaluation renders the productivity worries raised in Section 4.3 distinctly more tractable. Recall how it does so. By considering anger and non-­anger in isolation from the broader system of public deliberation, critics of anger force a choice upon us: either we can have angry speech’s epistemic benefits; or we can have non-­angry speech’s practical benefits for communication and coordination. But we cannot have both. The conceptual shift recommended by the systemic approach reveals that this is a false dilemma. Indeed, the systemic approach brings into view the possibility of complementary relationships, additive or interactive, between spheres that welcome anger and spheres that discourage it. In fact, we have seen that some of the benefits Nussbaum credits solely to non-­angry discourse (e.g., communicating a loving perspective) may even depend on such complementary relations. Therefore, invoking the systemic approach allows us to say this: even if Nussbaum were right that angry speech taken alone has a corrosive effect on trust which impedes important types of practical coordination and communication— an assumption I revisit in Chapter 5—it would not necessarily follow that giving anger a significant place in the discursive system requires foregoing non-­anger’s practical benefits. Indeed, the division of labour between different arenas means that not every arena has to fulfil every function. To justify shunning public expressions of anger as categorically as Nussbaum does, the counterproductivity critic would have to show that anger-­welcoming and anger-­free sites cannot complement one another in this way. This is something they have not shown, and which—on the evidence supplied by past struggles against injustice—seems highly contentious.

6. Conclusion I have offered a defence of the claim that there is an important place for public expressions of anger in democratic public speech. Those who deem expressions of anger counterproductive typically overlook anger’s distinctive epistemic contribution, and adopt the wrong context for assessing its consequences.

82  Democratic Speech in Divided Times Publicly expressing anger matters partly because, in virtue of anger’s particular phenomenology, doing so can produce a distinctive kind of epistemic good. As philosophical research on emotion indicates, experiencing anger helps render morally significant facts salient and contributes to enriching our moral concepts. Therefore, insofar as it induces listeners to imaginatively experience anger, expressing anger enables them to register overlooked injustices, and to develop a finer understanding of those injustices. In this, we have a systematization of the compelling suggestion with which we started: that expressing anger is there to teach the audience something, by casting a spotlight on ‘what [is] really happening’. This function matters greatly in non-­ideal settings, where members of social groups are often radically ignorant of the injustices endured by other groups. Taken alone, this first defence encounters several worries. But the force of these worries softens once we adopt a more adequate context for evaluating the consequences of public anger. In line with the systemic approach to public discourse, we should not ask whether isolated expressions of angry speech have better consequences than isolated expressions of non-­angry speech. Instead, the relevant question is whether a system that gives a key role to angry speech (among other kinds of speech) is more productive than a system that does not. By bringing into view various complementary relationships that can obtain between anger-­ welcoming and anger-­free venues, the systemic perspective renders the above worries more tractable. This second part of my defence captures another intriguing claim with which we began. We now have an answer to why it matters that Ellison’s invisible narrator be ‘dedicated and set aside’ for the purposes of expressing his anger: anger is supposed to play specific functions within a broader system, which complement and depend upon other forms of speech. This philosophical defence should guide our future empirical research on publicly expressed anger’s (counter) productivity. First, given the foregoing account of angry speech’s epistemic value, social scientists should further investigate under what precise conditions publicly expressing anger tends to rouse listeners to anger.87 Second, since the systemic perspective reminds us that we cannot conclude much from evidence regarding the effects angry speech has in isolation, they should study more fully the additive and interactive relations of complementarity that may obtain between angry and non-­angry speech. In the meantime, I have offered some preliminary evidence for thinking that this avenue of research is promising. In particular, by analysing how it illuminates historical pairings such as Malcolm X and King, and Douglass and Lincoln, I hope to have given a sense of how, when skilfully channelled within the discursive system, rage offers powerful tools for combatting injustice.

87  For preliminary research, see Bourgeois and Hess (2008).

3

Countering Public Hate Speech 1. Introduction Emotionally charged narratives, including narratives infused with anger, should play an important role in the democratic public discourse of divided societies. As we have seen in Chapters  1 and  2, such narratives are epistemically vital. They enable members of different social groups, who encounter radically different constraints and enablements, to communicate politically important experiences and perspectives to one another. This epistemic function matters greatly, not least because it contributes to making another desirable discursive norm more practicable in divided contexts: namely, the norm that participants in the final stages of formal deliberation should appeal to shared reasons. When emotionally charged narratives succeed in making important new considerations shared, the shared reasons constraint then allows these newly shared considerations to be used when justifying political demands in the latter stages of formal deliberation. In tandem with a discursive norm that welcomes emotionally charged narrative into public discourse, the shared reasons constraint is therefore less restrictive, and more appealing, than it would otherwise be. Yet this praise of personal narratives might seem to be too inclusive. In strongly divided democracies, some citizens have highly prejudiced beliefs about others. For such citizens, telling others how they see the world may involve publicly expressing profoundly disrespectful views—that is, views that deny others’ standing as equal persons. Racists or homophobes, for example, might publicly deliver narratives that communicate their hateful viewpoints. So, disrespectful views could enter public discourse at the discursive stage where citizens express narratives aiming to make their distinctive perspectives shared.1 Unless they are further qualified or supplemented, then, the norms advanced in Chapters  1 and  2 may open the door for public utterances whose content is deeply disrespectful. That deeply disrespectful views may make their way into public discourse is of course not just a theoretical problem. Whether it is Danish cartoonists publishing 1  This is true even if the narrative ultimately fails to make the abhorrent considerations shared. While such a failure would mean that the shared reasons constraint will not allow the consideration to be used to justify political demands during the final stages of formal deliberation, the abhorrent narrative would still be expressed at the prior stage where deliberators aim to publicize their distinctive experiences.

Democratic Speech in Divided Times. Maxime Lepoutre, Oxford University Press (2021). © Maxime Lepoutre. DOI: 10.1093/oso/9780198869757.003.0004

84  Democratic Speech in Divided Times representations of the prophet Muhammad as a bomb-­throwing terrorist,2 British journalists assimilating migrants to ‘cockroaches’ and to a ‘plague of feral humans’,3 or then-­presidential candidate Donald Trump referring to Mexican immigrants as ‘rapists’ and ‘criminals’,4 one thing leaves no doubt: disrespectful and degrading speech is rampant in actual democratic political discourse. What should we do, then, about public speech that expresses deeply disrespectful viewpoints? There is no real controversy that such utterances are morally undesirable. Not only are they offensive but, as we will see, disrespectful public utterances like those cited above can also give rise to significant harms. Accordingly, deliberative theorists of democracy widely agree that, ideally, participants in deliberation should respect one another. Gutmann and Thompson, for example, maintain that ‘mutual respect’ should govern deliberative inter­ actions.5 Likewise, in Young’s communicative ideal, interlocutors should frame their exchanges ‘in terms of basic respect’.6 So, by contrast with the discursive norms mentioned in Chapters 1 and 2,7 the salient question here is not whether public speakers have moral reasons to refrain from engaging in deeply disrespectful speech. That much is uncontroversial. Rather, the salient question is: given that so many actual speakers violate this moral norm, what kind of response is warranted by these moral reasons? Do the moral reasons that oppose disrespectful public speech recommend implementing formal legal norms that coercively expel such speech from democratic public discourse? Or should we instead respond to disrespectful speech internally to demo­crat­ic discursive processes, by countering it with more speech? My examination of these questions will focus on an especially abhorrent type of disrespectful speech: hate speech. As we will see in the following section, hate speech picks out an intensely disrespectful kind of speech, as opposed to a mildly disrespectful one. As a result, the public presence of hate speech is distinctively worrying for the ideal of inclusive public discourse: it threatens to turn such discourse into a vehicle for harm and injustice, rather than a vehicle for opposing these social problems. How, then, should we respond to the presence of hate speech in public discourse? In addressing this question, we should take as our starting point Jeremy Waldron’s The Harm in Hate Speech, where he develops what is perhaps the most forceful and sophisticated existing argument for adopting legal norms that 2  Asser (2010). 3  Usborne (2015). 4  Reilly (2016). 5  Gutmann and Thompson (2004, 79–81). 6  Young (2000, 48). See also Krause (2008, 166) and Mansbridge et al. (2010, 66). 7  As we saw in Chapter 1 (esp. Section 2.1), debates about the shared reasons constraint are largely about whether there is a moral duty to appeal to shared considerations, where this moral duty might be unenforced or only enforced informally (e.g., through verbal disapproval). A similar consideration applies to debates about the place of emotions in public discourse, such as the debate on anger discussed in Chapter 2.

Countering Public Hate Speech  85 coercively ban hate speech. Drawing on insights from law, political philosophy, feminist thought, and critical race theory, Waldron contends that public hate speech severely erodes the dignity of its targets. This dignitarian harm, he suggests, should dispose us legally to restrict public hate speech. The principal strength of Waldron’s analysis lies in its articulation of precisely how public hate speech harms the dignity of its targets. There is no denying, therefore, that if unchallenged, public hate speech is deeply harmful. But what is less persuasive, I wish to suggest, is the claim that coercive legal bans constitute the best way of responding to this harm. Like many other proponents of bans, Waldron relies on an overly optimistic picture of his preferred solution (bans), and an overly unsophisticated picture of the main alternative (countering hate speech with more speech, or ‘counterspeech’). Accordingly, by developing a more refined understanding of counterspeech, and by offering a more nuanced vision of bans, I will argue that counterspeech seems a morally preferable way of upholding the dignity of targets of hate speech. In this light, we should be significantly less inclined to implement legal norms banning public hate speech. Instead, we should embrace robust norms of demo­ crat­ic counterspeech. Note, however, two caveats. First, my defence of counterspeech will remain somewhat pessimistic. In other words, for some of the objections to countering hate speech with more speech, I will proceed less by suggesting that counterspeech fully overcomes them, and more by showing that they are at least as problematic for bans. Second, although I focus primarily on the dignitarian account of what makes public hate speech harmful, there are other ways in which such speech can harm. For example, hate speech is sometimes said to incite violence. For this reason, my argument in this chapter is limited in scope. That being said, it does not follow that my inquiry is irrelevant to these other harms: as a further line of investigation, we should explore to what extent the strategy I deploy for defending counterspeech can successfully be extended to other speech-­based harms. Because many of the other harms associated with hate speech involve claims about the causal effects of hate speech, I expect that this further line of investigation will be, in substantial part, empirical. My argument will be structured as follows. I begin by offering a working def­in­ ition of hate speech (Section 2) and outlining how hate speech, thus understood, harms its targets (Section 3). The rest of the chapter will then argue that counterspeech, not legal bans, is the more appropriate tool for countering this speech-­ based harm. As a preliminary step, I show that legal bans are morally costly. What this means is that we should adopt bans only if counterspeech proves less effective than bans at blocking the harm in hate speech (Section 4). In turn, I articulate the norms that should govern counterspeech for it to be maximally effective. Counterspeech should be driven, not merely by private actors, but also by the

86  Democratic Speech in Divided Times state (Section  5). And it should consist, not merely of ‘negative’ opposition to hateful viewpoints, but rather of ‘positive’ reaffirmations of egalitarian ideals (Section 6). To reiterate, my claim is not that positive, state-­driven counterspeech will always be effective. It is rather that, in light of the obstacles that stand in the way of countering hate speech with more speech, this conception of counterspeech is the most promising. And insofar as its effectiveness remains limited, the sources of these limits are at least as problematic for legal restrictions.

2.  Defining Hate Speech The definition of ‘hate speech’ is highly contested. By ‘hate speech’, I will refer, roughly, to speech that communicates or otherwise expresses the inferiority of other members of society. More specifically, hate speech emphatically rejects the basic standing of its targets as equals, typically on the basis of their membership in a vulnerable social group.8 Two things are worth stressing here. First, hate speech includes not just verbal speech, but all platforms (including printed images) that might be used to express the inferiority of a target group. Second, what makes an utterance hate speech is not the speaker’s emotion, but rather the content that it communicates or expresses. And, in principle, it is possible to communicate that another group is fundamentally inferior without experiencing hatred or resentment towards that group. This abstract characterization is intended as an approximate guide, rather than as a precise analysis. So, to make matters clearer, consider a few illustrations of the kinds of speech I am concerned with. Hate speech might include newspaper articles falsely attributing essential dangerousness to a minority group (‘Muslims are terrorists’). Or it might take the form of leaflets portraying some groups as subhuman (for instance, depictions of black people as ape-­like). Sometimes, hate speech explicitly expresses the exclusion of a minority group from the political community, as in Pegida’s infamous posters asserting ‘Rapefugees Not Welcome’. We have seen, when examining the systemic approach to public discourse, that public speech can take place in many different kinds of arenas and contexts. What is more, the same utterance can perform different functions, and have different effects, depending on where it occurs. This is also true of hate speech. Now, instead of considering hate speech across all discursive contexts and arenas, my

8  This definition is inspired by Waldron (2012, 27–8, 34–41) and Brettschneider (2012, 1). In legal scholarship, it echoes the UN’s (1965) condemnation of speech ‘based on ideas or theories of su­per­ ior­ity of one race or group of persons of one colour’.

Countering Public Hate Speech  87 discussion in this chapter will be restricted to the following kinds of contexts and arenas. First, I will be focusing, as in this book more generally, on hate speech that occurs in relatively stable democracies. That is, I bracket contexts of political instability where hate speech is likely to induce a breakdown of minimal peace. While this would constitute an extremely powerful reason for legally banning it,9 I will rather be concerned with cases, like France or the United States, where hate speech is diffuse, without threatening to produce an imminent political breakdown. The second restriction is that I will be especially concerned with hate speech that has a significantly public character. Here, I mean ‘public’ in quite a broad way: it includes not just communications occurring in formal political debates, but also those published in newspapers, posted on the Internet, or plastered in common spaces. Very often, Waldron notes, such communications become ‘a permanent or semipermanent part of the visible environment’.10 Furthermore, this is commonly speech that is not simply directed at the targets of the hate speech’s content, but addresses a broad audience, which includes third parties. Hence, although the distinction between public and non-­public hate speech remains vague, this restriction tends to exclude live, face-­to-­face encounters where, for example, the speaker directs a slur at a particular person with the deliberate intention to wound, or cases where someone threatens a particular individual by, say, spray-­painting a swastika on their door.11 I focus on public hate speech, thus defined, for two reasons. The first has to do with the harm in hate speech. I will be concerned with the way in which hate speech harms its targets by assaulting their dignity. And this particular harm, Waldron suggests, arises most readily from public instances of hate speech. The second reason for this focus is political. The political contrast that is often drawn is between the US, which tends to oppose hate speech bans, and the rest of the world, which tends to embrace them. But even in the US, live, face-­to-­face ‘fighting words’, as well as targeted threats and harassment, are generally legally prohibited.12 So the main political debate concerns not whether, but which kinds of hate speech should be prohibited. And hate speech bans in public discourse are politically far more divisive. This characterization might still seem too broad. Many actual hate speech re­gu­la­tions require not just that the expressed content of the regulated speech profoundly contradict the basic standing of vulnerable groups, but also that it be uttered in circumstances such that it is likely to produce certain effects, like 9  Tirrell (2012). 10  Waldron (2012, 37). 11  Heinze (2013, 590–1), Brettschneider (2012, 1), Langton (2012, 76–7), and Waldron (2012, 37) also bracket such face-­to-­face assaultive forms of hate speech. Brison (1998, 313–15) treats them separately. 12  Chaplinsky v. New Hampshire (1942); Virginia v. Black (2003).

88  Democratic Speech in Divided Times violence or antipathy.13 Now, there is enormous variation in characterizations of hate speech. In part, this is due to the fact that how we define hate speech depends on what kinds of speech-­related harms we are concerned with. Accordingly, I  adopt this relatively broad account because (1) it is adapted to the particular speech-­related harm I will be focusing on (see Section 3), and (2) it is informative given my more general aim: since I am fundamentally interested in democratic public discourse, I am especially concerned with speech that is abhorrent because it expresses views that contradict the moral underpinnings of democracy (i.e., participants’ status as equal citizens).

3.  The Harm of Hate Speech For some, the harm involved in hate speech warrants forcibly excluding or banning it from the public discourse of democratic societies.14 According to their opponents, countervailing considerations, such as freedom, democracy, or truth, demand that we tolerate it.15 Waldron articulates a framework for understanding the harm in hate speech which, he claims, supports the former option. Waldron starts from the strong intuition that justice requires treating human beings with dignity. In the first place, this means that citizens’ status as equal members of society in good standing—their dignity—should be recognized and upheld by other citizens.16 However, drawing on Rawls’s idea of a well-­ordered society, he argues that justice requires something more. Citizens must also know that their peers uphold their good standing. Absent the assurance that their dignity is safe, citizens cannot fully enjoy their good standing. Indeed, this assurance is necessary for citizens to pursue their aims and to participate in civil and pol­it­ ical life without fear or shame.17 Because the public assurance of dignity is an essential component of justice, Waldron suggests that we should pay greater attention to ‘political aesthetics’. We should care, in other words, about what our society literally looks and sounds like.18 In particular, we should pay attention to what a society that allows public hate speech would look like to targets of hate speech. Imagine, Waldron urges, the ‘marching feet and chants of neo-­Nazis in Skokie, a Grand Wizard’s speech at a Ku Klux Klan rally’.19 Or imagine, similarly, a Muslim citizen confronted with newspapers and posters depicting Muslims as terrorists. This is clearly not what a 13  The UK’s Public Order Act of 1986, for instance, prohibits insulting or abusive speech when it is likely to stir up hatred. See also Mill’s (2006[1859], 64–5) justification for restricting speech. 14 E.g., Waldron (2012), Delgado (1993), Matsuda (1993), Brison (1998), and Quong (2011, 305–12). 15  On freedom or autonomy, see Baker (1996), Scanlon (1972), and Dworkin (1981). On democracy, see Heinze (2016), Dworkin (2009), and Post (1990). On truth, see Mill (2006[1859]). 16  Waldron (2012, 59–60). 17  Waldron (2012, 65–9, 81–9). 18  Waldron (2012, 71–7). 19  Waldron (2012, 71).

Countering Public Hate Speech  89 well-­ordered and just society looks and sounds like. When a society looks and sounds like this, it cannot be public knowledge that all citizens uphold the dignity of all others. This reveals a powerful harm generated by public hate speech. In publicly denying the status of its targets as social equals, public hate speech affects the way the society looks in such a way as to undermine the public assurance that their status is secure. Hearing racist vituperations of far-­right groups on the radio, for example, may well produce fear and a sense of alienation in its targets. In add­ ition, public hate speech reaches out to other hateful persons, to assure them that they are not alone. In doing so, it replaces the public assurance of dignity with the assurance of hatred.20 How exactly does hate speech relate to this harm? Dignitarian accounts of the harm in hate speech commonly argue that hate speech causes harm to the dignity of its targets. On this understanding, hate speech causally influences how other members of society view and act towards the target group, so that they stop upholding their status as equals.21 While Waldron’s account is consistent with the causal claim, he insists that there is an even tighter link between hate speech and harm to dignity. Hate speech constitutes an assault on dignity.22 In publicly saying ‘You are not one of us, and we want you out’, hate speech undermines the public assurance that the target group are respected by their peers. Communicating this is part and parcel of the act of refuting the public assurance of dignity, and of assuring other citizens with hateful views that they are not alone.23 Three features of this account of the harm of hate speech are particularly appealing. First, it rests on intuitively appealing moral foundations—the idea that justice requires not just respecting the dignity of human beings by upholding their status as equal members of society, but also publicly assuring them that their equal status is secure. Second, the framework is methodologically appealing. Waldron embraces non-­ideal theory by demanding that we imagine what hate speech looks like in the real world, particularly to those on the receiving end. This commitment to taking the experiences of targets seriously is a welcome corrective to scholarship that frequently adopts an abstract and idealized conception of hate speech. Finally, because it focuses on the harms constituted (as opposed to caused) by hate speech, Waldron’s account is dialectically powerful. It sidesteps the fiercely contested empirical question of whether the causal relations between hate speech and its purported harms really obtain.24 For all of these reasons, Waldron’s account of the harm in hate speech constitutes a promising starting point for thinking about the problem posed by public

20  Waldron (2012, 94). 21  E.g., Delgado and Stefancic (2004). 22  Waldron (2012, 166–7). 23  For other constitutive accounts, see, e.g., Langton (2012) and Maitra (2012). 24  Waldron (2012, 166).

90  Democratic Speech in Divided Times hate speech. The crucial question is what follows from this profound harm. Public hate speech assaults the dignity of its targets. How should we uphold the assurance of dignity in the face of this assault? For advocates of bans, such as Waldron, the answer is that coercive hate speech bans are needed. By forcefully excluding hate speech from the public sphere, bans prevent hate speech from creating an environment of fear and alienation, where members of vulnerable groups cannot be assured of their dignity. An alternative, however, would be to allow public hate speech, and to counter it with more democratic public speech. In the rest of this chapter, I will argue that, when governed by the right norms, this counterspeech strategy is a prima facie preferable response to the harm of public hate speech.

4.  The Moral Cost of Legal Restrictions The first thing to emphasize is that bans—legal prohibitions that impose coercive sanctions, such as fines or incarceration, on hate speech—come at a significant moral cost. The most commonly invoked moral cost is that having one’s speech suppressed significantly undercuts one’s freedom or autonomy. Although this point can be made with different conceptions of autonomy or freedom, I will emphasize, in line with previous chapters, how it applies with freedom understood as non-­ domination. Domination, recall, is subjection to arbitrary or unchecked power. Accordingly, non-­domination is secured insofar as individuals are capable of contesting the power to which they are subjected, so that it is forced to track their interests and concerns.25 In Chapter 1, I argued that the ability of participants to contest and thereby check political power is enhanced when they can publicly appeal to the considerations they deem most significant.26 This remains true even when those considerations are hateful ones. Hence, banning public hateful con­ sid­er­ations diminishes the freedom as non-­domination of citizens who possess hateful political views.27 Importantly, Waldron himself recognizes these freedom- or autonomy-­based costs. Bans, he claims, have a direct bearing on freedom to publish, sometimes on freedom of the press, very likely on freedom of the Internet [. . .] And this matters to individuals. 25  Lovett and Pettit (2009, 14). 26  Chapter 1, Section 2.2. 27  For other freedom- or autonomy-­related objections, see Brettschneider (2012, 76–80), Scanlon (1972), Dworkin (1981), and Baker (1996, 992). Note that this objection cannot justify protecting all kinds of hate speech. Shouting a slur at someone in the street, for instance, is not an attempt at contesting political power. But, as stated in the introduction, I am bracketing such assaultive, face-­to-­face forms of hate speech. By contrast, the examples of public hate speech mentioned in the introduction (e.g., ‘Muslims are terrorists’) sometimes are raised to support political demands.

Countering Public Hate Speech  91 Often the messages that racists or Islamophobes are stopped from expressing are the very messages [. . .] that matter most to them. For them, other aspects of political expression pale into insignificance [by comparison].28

But Waldron is unfazed by this observation. He responds that these con­sid­er­ ations are outweighed by considerations having to do with the dignity of targets of hate speech. All things considered, then, bans remain morally justified. We could go even further on behalf of advocates of bans. Lacking assurances that one is a member of society in good standing risks depriving one of the psychological conditions needed to participate politically without fear or shame. By undermining the public assurance of dignity, hate speech may therefore undercut its targets’ ability to contest exercises of political power. Therefore, even within the value of freedom, understood as non-­domination, it is plausible to think that the balance of reasons favours bans. I want to grant these responses to the freedom-­ based argument, not merely for the sake of argument, but because I find such claims about the harmfulness of hate speech persuasive. But even if this is true, we should still be worried about bans. Even if the freedom-­based costs of bans may be outweighed by their benefits, they remain significant. Part of what is non-­ideal in actual circumstances is that many citizens uphold hateful viewpoints. Over the past decade, for example, far-­right parties, whose rhetoric is often saturated with xenophobia and racism, have grown in power across Europe and the United States. In some cases, moreover, supporters of such hateful ideologies are socio-­economically disaffected citizens who feel, sometimes justifiably, that their voices have been ignored by mainstream parties.29 Thus, coercively suppressing hate speech seems to come at a substantial moral cost: it further diminishes the freedom of many relatively disadvantaged citizens, and thereby fuels the exclusionary processes that, often, contributed to producing their hateful disaffection.30 What this means is that even though advocates of bans are right that the claims of targets of hate speech outweigh those of citizens who possess and wish to express abhorrent views (‘hateful citizens’, for short), the trade-­off between the two is significantly worse than they typically acknowledge. This realization should strongly motivate us to seek a solution that weakens the trade-­off—that is, that blocks the harmful features of hate speech while limiting the costs to hateful citizens. Bans should be a last resort.

28  Waldron (2012, 148–9). 29  E.g., Willsher (2015). 30  Why does it matter that those affected by bans are often relatively badly off? On a consequentialist interpretation, ignoring those who have already been ignored is problematic because it risks alienating them further, and breeding more hatred. But the explanation needn’t be consequentialist. Instead, one might think this matters because one shares the basic prioritarian intuition that losses of welfare or freedom are more morally problematic when they afflict those with overall lower levels of well-­being.

92  Democratic Speech in Divided Times Now, in observing that many of those whose freedom would be restricted by bans are relatively disadvantaged and numerous, my intention is to bring out how important it is to find an alternative response to hate speech. But my more fundamental point does not hinge on this sociological observation. Even where hateful citizens tend to be well off (as they of course often are), there is still a freedom-­ based cost to legally restricting their political speech. Accordingly, it remains true that if we can find an alternative solution that blocks the harm in hate speech without imposing this cost on hateful citizens, we should prefer that solution to legal bans. This takes us to counterspeech. For legal bans to be justified despite their moral cost, it must be the case that counterspeech is insufficient or otherwise inadequate as a way of counteracting the harm in hate speech. In Sections 5 and  6, I will consider what I take to be the two most forceful arguments for holding this view, and I will argue that a more sophisticated conception of counterspeech can avoid both to a significant degree. Moreover, insofar as the more sophisticated conception of counterspeech that I advocate faces lingering problems, these problems can be turned—arguably with even greater force—against legal bans.

5.  State Counterspeech 5.1  The Burdens of Counterspeech According to a historically influential understanding of counterspeech, the demo­ crat­ic public sphere should be akin to a ‘marketplace of ideas’,31 where all members of society have the opportunity to freely exchange their views. This means that degrading views can be expressed by hateful members of society. But it also means that other members of society, who disagree with these hateful standpoints, can in turn counter hate speech by expressing their own ideas. Thus, it is said that the free marketplace of ideas responds both to the interests of hateful citizens (since they can freely express their opinions) and to the interests of the targets of hate speech (since hate speech can be countered through the speech of other citizens). Understood in this way, the counterspeech strategy faces a powerful objection— namely, that it places the burden of responding to hate speech on targets of hate speech. According to Ishani Maitra and Mary Kate McGowan, if the state does not intervene to legally restrict hate speech, and instead stands by as members of society engage in a free marketplace of ideas, the task of responding to hate 31  E.g., Milton (1644), Mill (2006 [1859]), Holmes (1919), Brandeis (1927), and Schulzke (2016, 237–9). A version of this view is criticized by Waldron (2012, 67), Matsuda (1993, 36–7), Delgado (1993, 95), Nielsen (2012, 154), Brown (2015, 246), and MacKinnon (1993, 72–6).

Countering Public Hate Speech  93 speech through more speech will fall, disproportionately, on targets of hate speech. And this, in turn, is problematic for at least two reasons. First, it suggests that the counterspeech strategy is unfair. Targets of hate speech are the ones who sustain the harms of hate speech. It therefore seems wrong that they should have to bear the costs of countering hate speech as well.32 Second, if targets are the ones tasked with responding to hate speech, counterspeech is likely to be ineffective at preserving the public assurance of dignity. For one thing, empirical evidence indicates that targets of hate speech in fact rarely respond to hate speech in public settings. This is, in part, because they sometimes fear retaliation. But it is also because doing so is extremely demanding, especially in non-­ ideal conditions where hate speech is not only harmful but also widespread.33 But there is a more fundamental reason why having targets of hate speech respond is unlikely to be effective at protecting the public assurance of dignity. The difficulty, as philosophers of language often put it, is that targets of hate speech frequently lack the authority needed to block the harm in hate speech.34 If what is harmful in public hate speech is that it asserts that a given group is unwanted by the society at large, then being able to refute this assertion requires being able to speak on behalf of the society at large. But targets of hate speech are typically members of vulnerable minority groups. And, what is more, hate speech contributes to eroding their social standing.35 As a result, they are often not in a position to speak, authoritatively, on behalf of society at large.

5.2  The Idea of State Counterspeech The problem with this first argument against counterspeech is that it relies on an inadequate account of who should engage in counterspeech. The objection assumes, in line with the marketplace of ideas metaphor, that the responsibility for engaging in counterspeech lies essentially with private citizens. In doing so, it sets up a choice between either having the state legally prohibit hate speech or having the state step back and let private citizens respond. But this framing is misleading. Even when the state refrains from banning hate speech, it can be active as a speaker endowed with expressive powers. This, I believe, is the more appropriate way of thinking about counterspeech: the task of speaking out against hateful perspectives should be spearheaded by the state. As Corey Brettschneider and Katharine Gelber have observed, states possess numerous tools for speaking back against hate speech.36 To begin, the state can 32  Maitra and McGowan (2012, 9–10); McGowan (2018, 183). 34  Maitra and McGowan (2012, 9–10). 35  Langton (2018b). 36  Brettschneider (2012); Gelber (2012).

33  Nielsen (2012).

94  Democratic Speech in Divided Times speak back by having public officials affirm the ideal of human dignity, not just through public pronouncements in formal deliberative forums, but also by naming public spaces, dedicating monuments, and enacting public holidays. In add­ ition, state officials can also speak back by denouncing particular instances of public hate speech and doing away with national monuments, holidays, and other public symbols, that express degrading perspectives. Of course, public officials can only do so much. But the state can also speak back in more indirect ways. First, as Brettschneider notes, it can fund private associations, like civil rights groups, whose functions include publicly responding to occurrences of hate speech.37 Second, the state can speak against hateful views through public education, which teaches children about historical events such as the Holocaust, and which exposes them to democratic values.38 How does having the state use its numerous expressive tools to speak out against hate speech address the concerns raised in the last section? First, and most obviously, state counterspeech is less unfair than approaches that rest exclusively on private counterspeech. While the ‘state counterspeech’ strategy is consistent with the targets of hate speech also speaking back, the burden of responding does not fall primarily on them. Rather, it falls first and foremost on the state, either directly or via the third parties it sponsors. Second, state counterspeech does not suffer from a deficit of authority. State officials, unlike many private citizens, have a highly authoritative status, in virtue of which they can successfully counteract hate speech’s assault on the public assurance of dignity. When state officials denounce hateful utterances, they are thereby using their authority to reassure targets of their good standing. They are saying: ‘We are sorry that you had to hear that. It’s true, some individuals unfortunately still believe such things. But on behalf of the people generally, rest assured that we stand by you; that they, not you, are the outliers; and that we will ensure that what they say remains an empty threat.’ To make this last point more vivid, consider how Christiane Taubira—then the French Minister for Justice—intervened in response to David Rachline—a Front National Senator—in the French Senate in 2015. After Rachline criticized Taubira’s advocacy of humane treatment for criminals, to whom he referred as ‘thugs’ and ‘executioners’, Taubira replied to his various claims, and concluded as follows: ‘You belong to a political family [. . .] that systematically seeks scapegoats [. . .] and that looks for invaders everywhere [. . .] [O]ne shudders at the thought of

37  Brettschneider (2012, 110–11). 38  Brettschneider (2012, 96–104). Countering hate speech through public education does entail that teachers, in their capacity as teachers, are required to refrain from expressing hateful viewpoints. As Brettschneider (2012, 40) explains, ‘there is no general right to be employed by a public school’. Teachers (like other government employees) are hired on the condition that they perform certain functions—including repudiating hateful views in the classroom. Therefore, the state would be justified in firing teachers who violate this condition.

Countering Public Hate Speech  95 what would happen if it took power—but we will fight to make that absolutely impossible.’39 The overwhelming majority of the Senate rose to their feet to applaud this response. In her response, Taubira does several things: she takes the time to answer; she condemns far-­right ideologies and their scapegoating practices; she promises to keep those exclusionary ideas away from power; and she does this in broad daylight, to wide acclaim. If we want hate speakers to appear publicly isolated, so that their speech cannot undermine the assurance of dignity, public counterspeech like Taubira’s seems to realize those aims in an exemplary fashion. At this juncture, an important point of clarification is needed. It is tempting— particularly when one thinks of speech predominantly through the lenses of deliberation—to think of state counterspeech primarily as a persuasive tool. On this view, the point of having the state respond to hate speech is to persuade hate speakers, or (more likely) third parties, that hateful perspectives are deeply misguided. This, for example, is how Brettschneider tends to think of state counterspeech.40 But though persuasion is important, it is not the only, nor indeed the main, function of state counterspeech. The harm generated by hate speech, as Waldron sees it, is not primarily that it persuades people to embrace hateful views. It is, instead, that it alters the public appearance of society, in a way that makes it impossible for targets of hate speech to feel assured of their good standing. From this perspective, what state counterspeech must do is dispel this appearance, and thereby reassure targets of their good standing. That is what Taubira’s intervention brilliantly illustrates. Even if her utterance fails to persuade anyone who did not already reject hateful views that they ought to do so, it contributes to blocking the harm of hate speech by emitting an authoritative assurance that the targets of hate speech are, and will remain, in good standing. The same might be said of public education. One aim of exposing students to democratic values is of course to persuade them to accept these values. But that is by no means its only aim. Importantly, doing so also contributes to assuring students who belong to vulnerable groups that they are valued members of society. So far, my argument gives us some reason to think that state counterspeech is preferable to legal prohibitions. Unlike private counterspeech, state counterspeech is not unfair, and it has the authority needed to uphold the public assurance of dignity in the face of hate speech. At the same time, state counterspeech performs this crucial function without coercively suppressing hateful citizens’

39  Sénat Vidéo (2015). 40  Brettschneider (2012, ch. 3). The persuasive function is also popular among theorists of non-­ state counterspeech, such as Mill (2006[1859]), Brandeis (1927), Schulzke (2016), and Howard (2019b).

96  Democratic Speech in Divided Times public utterances—and so, without incurring the distinctive moral costs of legal prohibitions. This, however, does not mean that state counterspeech will always be perfectly effective at preserving the public assurance of dignity. Perhaps the most obvious limitation of state counterspeech is that, in non-­ideal conditions, the state is sometimes part of the problem. That is, the state sometimes uses its expressive powers wrongly.41 Accordingly, state officials may sometimes put counterspeech to bad ends: instead of appealing to values such as dignity and equality to condemn genuine instances of hate speech, they may use their speech to advance ends that are contrary to those values, and to denounce speech that actually aims to promote those values. The public reception of Black Lives Matter brings this concern to life. Over the past ten years, the social movement has campaigned regularly against structural racism directed at black Americans, and has notably protested the violent treatment of black Americans by law enforcement officers. In response, in 2016, soon-­ to-­be-­president Donald Trump and Rudy Giuliani condemned the slogan ‘Black Lives Matter’ as racist on the fallacious grounds that it implied that non-­black lives did not matter.42 Here, we have a case of misdirected counterspeech by the state: authoritative public figures used the language of racial equality to denounce a public slogan that was neither hateful nor inegalitarian, and that instead promoted racial equality. This is a real limitation of state counterspeech. Yet it does not give us any reason to embrace legal bans instead of, or in addition to, counterspeech. If the state is badly motivated, this will presumably affect not just its use of expressive p ­ owers, but also its use of coercive powers.43 And so, just as the state may speak out against praiseworthy utterances and fail to speak out against hateful utterances, so too it may legally prohibit praiseworthy utterances and fail to prohibit hateful utterances. Therefore, insofar as this concern applies to state counterspeech, it also applies to legal bans. In fact, there is reason to think that this concern cuts even more deeply against bans than it does against state counterspeech. If a legal ban prohibits the egalitarian speech of Black Lives Matter activists, and if this prohibition is successfully enforced, then the egalitarian speech in question is suppressed altogether. By contrast, misdirected counterspeech does not disable all speech-­acts associated with the targeted utterance. State counterspeech clearly keeps the targeted speaker from doing some things with his or her words. Indeed, as we saw earlier in this section, the whole point of state counterspeech is to prevent targeted utterances from doing certain things (such as undermining the public assurance of dignity). But even so, counterspeech leaves other speech-­acts intact. In voicing 41  West (2014, 1039). 42  Twohey (2016); Weigel (2016). 43  For discussion of this concern, see Howard (2019a, 105).

Countering Public Hate Speech  97 her concerns, the target of state counterspeech can still successfully disclose her worldview, share her experiences, and expose others to new arguments. Furthermore, the target of state counterspeech can still contest the state’s claim that her views are hateful or unjust. For example, when leading politicians fallaciously condemn the slogan ‘Black Lives Matter’ as inherently racist, members of Black Lives Matter still can (and do) contest that claim. Hence, the risk that the state will misuse its powers seems even more problematic for bans than it does for counterspeech. While misdirected state counterspeech salvages some valuable speech-­acts associated with the targeted utterance, misdirected bans do not. By implication, this risk gives us no reason to replace or supplement state counterspeech with bans. Still, the risk of misuse is neither the only nor the most fundamental problem for state counterspeech. In the rest of this chapter, I will consider what I take to be a deeper concern. Although this concern will ultimately fail to justify embracing legal bans, we will see that it does recommend another refinement to the practice of counterspeech: namely, the use of positive counterspeech.

6.  Positive Counterspeech 6.1  The Salience of Counterspeech We have encountered a number of challenges for counterspeech: first, that it can be overly burdensome for targets of hate speech (Section 5.1); and second, that it can be misused by ill-­motivated actors (Section 5.2). But there is a more fundamental worry. Even if counterspeech is driven by the state, and even if the state directs it at genuine instances of hate speech, there is a reason to think that counterspeech risks being, not just ineffective, but worse than ineffective. In other words, state counterspeech risks maintaining or even exacerbating the harmful effects of hate speech. The main argument motivating this worry comes from philosophy of language. Philosophers of language have long observed that utterances can affect the conversations to which they contribute in many different ways. For instance, they can share new information with interlocutors; they can call accepted views into question; they can enact new norms for discussion; they can make new ideas salient to other members of the conversation, and so on. Some of these conversational effects can easily be reversed by means of a further utterance. A teacher who verbally introduces a new norm for classroom discussion (‘From now on, you must raise your hand before speaking’) can easily verbally reverse that norm (‘Never mind. You may speak whenever you like’). But feminist philosophers of language have argued that not all conversational effects are so easily reversed. And, in particular, McGowan has argued that the harmful

98  Democratic Speech in Divided Times conversational effects of oppressive speech (a category that includes hate speech) are ‘especially difficult to undo’.44 Two examples will help bring out this concern. Suppose a public figure asserts  ‘Xs are lazy parasites’. McGowan’s point is that verbally countering this assertion—for instance, by saying ‘That’s false! Xs are not lazy parasites!’—seems an ineffective way of reversing the initial utterance’s damaging conversational effects. Likewise, responding to a misogynist’s claim that women are submissive by saying ‘But women are not submissive!’ seems misguided, to McGowan.45 The intuitive problem, in both cases, is that challenging the hateful view somehow ends up consolidating its place in the public conversation, and maintaining its harmful effects. What accounts for this phenomenon? In an attempt to elucidate McGowan’s suggestion, Robert Simpson has argued that the explanation has to do with salience. Generally speaking, while we can easily make an idea or association of ideas salient to others by mentioning it and drawing their attention to it, it is much more difficult subsequently to make it less salient.46 This is because rejecting an association of ideas may inadvertently make one’s interlocutors pay even more attention to it. This salience-­based asymmetry helps explain why the harmful effects of hate speech are so resistant to counterspeech. When a hateful speaker expresses a vilifying association (say, between being an X and being a lazy parasite), they thereby make that association more salient to listeners. In this context, trying to repudiate the vilifying association risks amplifying its salience.47 That is, exclaiming ‘Xs are not lazy parasites!’ may challenge the claim that Xs are lazy parasites. But even as it does so, it risks magnifying the place of this stereotype within the public conversation. This is problematic, because there is a clear connection between the salience of a hateful proposition, on the one hand, and the harmfulness of that proposition, on the other. Public hate speech harms, in part, by attacking people’s assurance that they are members of society in good standing. The salience of hateful pro­ posi­tions contributes directly to this attack. When a proposition is salient to participants in a conversation, that proposition is a relevant topic of discussion. By 44  McGowan (2018, 189; see also 2009, 403). McGowan is building on Lewis’s (1979) observation that raising the standards for using vague terms like ‘flat’ is easier than lowering them. Stanley (2015, 157–61) and Langton (2018a, section  6) echo the point that oppressive conversational effects are harder to reverse. 45  McGowan (2019, 119–120). 46  Simpson (2013b, 572). 47  Simpson (2013b, 569–70). While McGowan and Simpson focus on oppressive speech generally, this phenomenon also arises in more specific debates. When discussing hate speech that takes a coded form, Stanley (2015, 159) worries that verbally countering such speech requires exposing the code, and thereby raising the salience of its hateful contents. Moreover, examinations of slurs have shown that slur terms preserve their bad effects even when they are negated (‘A is not an X’). E.g., Langton et al. (2012, 756–7). Here too, salience may have an explanatory role: negating a slur involves repeating it, which increases the salience of the degrading stereotype it expresses.

Countering Public Hate Speech  99 implication, when hateful propositions are salient in the public sphere, this means that the public conversation is partly about these hateful propositions. What is at issue in the public conversation, then, is whether or not, say, immigrants are parasites, women are submissive, or black people are lazy. Since these propositions are clearly inconsistent with the good standing of the targeted groups, these groups can no longer rest assured of their good standing. Instead, their standing is precisely what is at issue in public debate. The problem, in sum, is this. Verbally repudiating hate speech risks increasing the salience of its content. And the more salient a hateful proposition is, the more it endangers the public assurance of dignity. Thus, verbally repudiating hate speech to assure people of their good standing may inadvertently end up maintaining or even exacerbating the dignitarian harm of hate speech.48 This problem is at least as problematic for state counterspeech as it is for non-­ state counterspeech. Because state speech is highly authoritative and widely diffused, what the state says is typically highly salient. As a result, having the state repudiate vilifying propositions risks increasing their salience all the more. In other words, the objection under consideration seems especially problematic for state-­driven counterspeech because it works by harnessing the power of counterspeech against it. And since the state has an unparalleled ability to set the agenda of public discourse, it also has an unparalleled tendency to strengthen the salience of vilifying representations when challenging them. What this means is that, unlike in Section 5, we cannot alleviate the problem at hand simply by appealing to the notion of state counterspeech. The natural response to this is pessimism about counterspeech. McGowan famously concludes that counterspeech is as futile as ‘trying to unring a bell’.49 Likewise, Jason Stanley ultimately suggests that it may be ‘impossible’ to respond to vilifying speech without reinforcing the damaging salience of its contents.50 And this, Simpson adds, in turn has an unhappy upshot. For activated as­so­ci­ ations ‘to become u  n-salient,’ he observes, ‘we will simply have to change the subject [. . .] or allow time to elapse’.51

6.2  The Salience of Legal Restrictions The suggestion that (state) counterspeech risks maintaining or exacerbating the salience of hate speech is concerning. But legal restrictions, too, risk increasing the salience of hate speech in the public sphere. And if legal restrictions do not 48  Waldron (2012, 87) hints at this problem when he complains that speech-­based responses to hate speech are ‘evidence of a problem’, and therefore constitute the wrong way of upholding the assurance of dignity. 49  McGowan (2009, 403). 50  Stanley (2015, 161). 51  Simpson (2013b, 569).

100  Democratic Speech in Divided Times avoid or mitigate the problem of salience, then this problem does not give us any reason to replace or supplement counterspeech with legal restrictions. There are two main reasons why bans risk increasing the salience of hate speech. The first is that legal restrictions on hate speech have an expressive dimension. It is widely held that when the law prohibits conduct, this expresses a public message of condemnation towards that conduct. By implication, legal restrictions on hate speech publicly express opposition to hateful utterances. And this message of opposition is expressed by the state. Thus, in virtue of their expressive dimension, legal restrictions on hate speech in fact resemble state counterspeech—with the important exception that, unlike state counterspeech, bans coercively suppress speech and so incur a distinctive moral cost.52 The problem is that, insofar as legal restrictions resemble a (morally costly) form of state counterspeech, they too risk increasing the salience of hate speech. By expressing op­pos­ ition to hate speech, bans strongly implicate the existence of hateful utterances, and run the risk of drawing greater attention to those utterances. The second reason has to do with enforcement. The enforcement of legal restrictions on hate speech commonly leads to highly publicized trials that put a spotlight on hate speakers, their utterances, and the bad associations they promote.53 Indeed, hate speech trials—particularly criminal trials—are highly public events that often involve rehashing public hate speech in the process of determining whether or not it infringed hate speech laws. For example, Eric Heinze contends that the ‘high-­profile prosecutions’ of Holocaust deniers in France, Austria, and Germany have inadvertently fuelled a significant amount of ‘media hype’ surrounding them. ‘Holocaust deniers in the US rarely enjoy the celebrity their counterparts have won in Austria, Germany or France,’ Heinze concludes, ‘precisely because US law, and therefore the US media, mostly ignore them.’54 Perhaps this is too quick. One of the most important aims of legal restrictions is to deter hate speech. If it turns out that hate speech laws are highly effective at doing so, then one might think that bans do after all help reduce the salience of hate speech: they reduce the incidence of public hate speech; and, in turn, if public hate speech is very rare, then public prosecutions of hate speakers are bound to be as well. For this reason, Waldron suggests that legal restrictions operate in a distinctively ‘silent’ way, compared to state counterspeech.55 They pre-­emptively

52  I discuss the expressive power of legal restrictions on hate speech in greater detail in Lepoutre (2020). There, I argue that although legal restrictions have significant expressive power, state counterspeech can generally match this expressive power. And since bans come at a special moral cost, we should opt for state counterspeech. On the expressive dimension of laws generally, see McAdams (2015). And on the expressive dimension of hate speech regulations, see, notably, Simpson (2013a, 724) and Brown (2015, 240–2). 53 See, e.g., Heinze (2013, 599–600), Simpson (2013a, 724), and Gelber and McNamara (2015b, 656–7). 54  Heinze (2013, 600). 55  Waldron (2012, 87–96).

Countering Public Hate Speech  101 eliminate hate speech, and in so doing prevent it from negatively affecting the public appearance of society. This line of thought is unconvincing for two reasons. The first is empirical. There is, for the moment, little reason to believe that legal restrictions actually reduce the incidence of hate speech. Part of the problem here is an absence of evidence. We do not know, in any systematic way, whether countries that have hate speech restrictions contain less hate speech than countries that do not.56 And even if there were such an observed difference, it would be extremely difficult to establish that bans are causally responsible for it. This is because there are countless other social, cultural, and political differences between countries that ban hate speech, and countries that do not. What is more, the tentative evidence that we do have is not altogether promising. Heinze, for instance, observes that, despite adopting increasingly punitive and comprehensive bans, some European states have experienced a rise in hate speech.57 Similarly, Katharine Gelber and Luke McNamara report that, in Australia, the incidence of hate speech has hardly decreased (and in some contexts has actually increased) since the introduction of hate speech laws.58 But there is a second and more basic problem with the thought that if bans are effective as deterrents, then they do not raise the salience of hate speech. Bans suppress or prevent hate speech only to the extent that they are seen and heeded by prospective hate speakers. So, the case for thinking that bans are effective as deterrents depends on bans being highly visible. And insofar as they are highly visible, so too is their message of opposition to hate speech. But that message, as I mentioned above, strongly implicates the existence of prospective hate speakers, and therefore runs the risk of drawing people’s attention to the views they wish to express. So even if bans were effective at deterring hate speech, it would not follow that they do not raise the salience of hateful propositions. More generally, then, the worry that state counterspeech could increase the salience of hate speech—and with it, the dignitarian harm of hate speech—does not give us any reason to adopt legal restrictions. Legal restrictions seem just as vulnerable to this problem.

6.3  The Idea of Positive Counterspeech But we should go further than this comparative point. In absolute terms, the problem under consideration constitutes a serious blow for our attempts at countering public hate speech with more speech. In what follows, I will therefore

56  Brown (2015, 246). 57  Heinze (2013, 609). 58  Gelber and McNamara (2015b, 644–5).

102  Democratic Speech in Divided Times suggest that we can in fact mitigate this problem of salience by further refining the idea of counterspeech. Specifically, the objection under consideration overlooks the distinction between what we might call ‘negative’ and ‘positive’ types of counterspeech. One way to counter hate speech is to explicitly negate its content. ‘It is false,’ a pol­it­ ician might announce, ‘that Xs are cockroaches.’ Similarly, a civil rights activist might respond to hate speech by insisting that ‘Ys are not lazy parasites!’ Call this negative counterspeech. Because it involves repeating the hateful proposition in the process of negating it, negative counterspeech risks strengthening its salience. Accordingly, this kind of counterspeech threatens to weaken the public assurance of dignity. But there is another way of verbally countering hate speech. As we have seen, the difficulty of reversing hate speech’s attack on the public assurance of dignity results from an asymmetry having to do with salience. It is easier to make ideas, and associations of ideas, salient than unsalient. So far, we have focused exclusively on how this asymmetry applies to hateful ideas. However, the asymmetry applies to non-­hateful ideas as well. Other things being equal, it is also easier to make non-­hateful associations of ideas—for example egalitarian associations— salient than unsalient. What this suggests is that the problem under consideration does not show that all kinds of counterspeech are bound to be ineffective. Rather, it recommends a distinctive kind of counterspeech. To avoid reinforcing the salience of hateful associations, we should steer clear of counterspeech whose form is primarily negative. Instead, we should privilege positive forms of counterspeech, which aim to put forward and render salient egalitarian associations of ideas. On this approach, countering hate speech is less about directly contesting a distorted vision of the world, and more about affirming a correct vision of the world that is inconsistent with the hateful ideas at hand. Positive counterspeech is useful in two respects. In the first place, because positive counterspeech entails or implicates, but does not explicitly assert, that the hateful speech in question is wrong, it avoids increasing its salience—and, by extension, it avoids reinforcing its harmful effect on the public assurance of dignity. At the same time, because positive counterspeech makes egalitarian as­so­ci­ ations salient instead, it makes the asymmetry of salience work for good rather than bad: once salient, egalitarian representations will be comparatively difficult to render unsalient. Notice that resorting to positive counterspeech does not amount to simply changing the subject. The egalitarian vision that positive counterspeech affirms bears on the same issue as negative counterspeech. This is guaranteed by the fact that the content of positive counterspeech is supposed to entail (or at least implicate) the wrongness of the targeted hateful utterance. So, positive counterspeech expresses a similar proposition as negative counterspeech, but under a different

Countering Public Hate Speech  103 mode of presentation. Consider: ‘The star of King’s Row was married twice’ and ‘The 40th US President was married twice’ make the same claim about the world (namely, that Ronald Reagan was married twice) while highlighting different fa­cets of that claim (one highlights Reagan’s youth film career, the other his pol­it­ ical role). Analogously, positive and negative counterspeech tell us something similar about the state of the world, while emphasizing different features of that state: while negative counterspeech reactivates hateful associations in the process of negating them, positive counterspeech activates egalitarian associations. By way of illustration, consider first of all Martin Luther King’s ‘I Have a Dream’ speech. In it, King takes a stand against a deeply racist public culture, which is rife with statements professing that black Americans are morally, culturally, and intellectually inferior to whites. Famously, King does so by vividly depicting a vision of racial equality. In this vision, the American nation ‘will rise up and live out the true meaning of its creed: “We hold these truths to be self-­ evident, that all men are created equal”’; and, accordingly, ‘the sons of former slaves and the sons of former slave-­owners will be able to sit down together at the table of brotherhood’.59 King’s utterance is very different from saying: ‘That’s false! Black Americans are not immoral, uncivilized, or stupid!’ While his vision of racial equality entails—or at the very least, strongly implicates—the falsehood of these racist stereotypes, it is first and foremost a positive representation of racial equality. This difference matters. Because it does not repeat the racist stereotypes, King’s positive speech does not enhance their salience in the public sphere. Instead, it draws his audience’s attention away from these stereotypes, and towards a different—and incompatible—set of associations: namely, between the American dream and the values of equality and interracial unity. Another example of positive counterspeech involves the Ku Klux   Klan in Missouri. In 1999, in response to the instalment of a road sign advertising the KKK, nearby residents erected another sign close to it, which read: ‘Freedom of speech protects all people, even if they are wrong.’60 Consider how this sign operates as an instance of positive counterspeech. The sign does not explicitly cite the KKK’s ideology and say that it is wrong. Instead, it counters the KKK in a more ‘positive’ manner. In the first place, it draws people’s attention away from the KKK’s hateful views, and towards an egalitarian ideal: an ideal of equal tol­er­ation, according to which all have the right to freedom of speech. At the same time, although the sign does not explicitly negate the KKK’s ideology, it does oppose the KKK in two ways. First, the value of toleration that the sign puts forward is inconsistent with the KKK’s own racial intolerance. Second, the sign also implicates the wrongness of the KKK’s ideology in a subtler manner. Given its close juxtaposition to the KKK sign, the specification that freedom of speech protects

59  King (1963).

60  Richards and Calvert (2000, 553).

104  Democratic Speech in Divided Times people even if they are wrong implies that the KKK and what it stands for are wrong. In this, we have an ingenious manifestation of positive counterspeech: the road sign makes a positive ideal of equal toleration salient; and in doing so, it opposes the KKK without rendering its distorted views salient. Some forms of state-­driven counterspeech operate in this positive manner. The state, as Section 5.2 discussed, can notably oppose public hate speech by erecting monuments and enacting holidays that honour egalitarian social movements and their leaders. Martin Luther King Day is a clear example of this practice. Expressive acts such as these are in the first instance positive celebrations of social and political equality, which strongly implicate the wrongness of hateful pro­posi­ tions, without reiterating them. Thus, although negative forms of state-­driven counterspeech—for example, having state officials verbally negate hateful utterances—may be distinctively vulnerable to the problem of salience, this is not true of all forms of state counterspeech. The upshot is clear. When the state uses its vast expressive powers to oppose hate speech, and to thereby uphold the good standing of all members of society, it should, as far as possible, employ positive forms of counterspeech. In recommending positive counterspeech, I am of course not suggesting that this form of counterspeech is a perfect remedy for the problem of salience. For some special forms of hate speech, it may not be possible to respond in a ‘positive’ way, which does not increase the salience of the hateful proposition being countered. Moreover, even when positive counterspeech is available, it may not fully undo the harms generated by hate speech. I will return to both concerns in the next chapter. For now, however, these concerns do not undermine my main conclusion here: that, insofar as we wish to counter hate speech with more speech, we ought to do so using counterspeech that is not only driven by the state, but also positive in form. Even if it is true that positive counterspeech does not eliminate the difficulties stemming from the salience of hate speech, that would be an unreasonable expectation. After all, neither do legal restrictions. And though positive counterspeech does not eliminate the difficulties at hand, it nonetheless alleviates them. It does so, to reiterate, by drawing our attention away from hateful ideas, and towards egalitarian ideas that entail or implicate the wrongness of hateful pro­ posi­tions. Thus, positive counterspeech repurposes the asymmetry of conversational salience, and puts it to a good end.

7. Conclusion Public expressions of deeply disrespectful views—such as public hate speech— assault the basic social standing of their targets. When this kind of speech is

Countering Public Hate Speech  105 widespread, democratic public discourse therefore risks becoming a vehicle for harm and social exclusion, rather than a remedy for these social problems. How, then, should we counteract such utterances? This chapter has offered a defence of the view that public hate speech should be countered via more public speech, rather than via legal regulations. Central to my argument is the claim that attempts at countering public hate speech with more public speech should be governed by two important discursive norms: they should be driven by the state, rather than merely by private actors; and they should take a positive, rather than merely a negative, form. Understood in this way, counterspeech constitutes a comparatively desirable tool for reassuring targets of hate speech of their good standing. For one thing, counterspeech does not involve coercively suppressing speech, and it is therefore less morally costly than legal bans. At the same time, state-­driven, positive counterspeech seems at least as effective as bans at blocking the harm in hate speech. Such counterspeech will of course not always be effective at upholding the assurance of dignity. But throughout, I have argued that, insofar as its effectiveness has limits, these limits also apply—often more strongly—to bans. All in all, then, there are good reasons for thinking that, when combatting public hate speech, counterspeech is a preferable solution to bans. Yet deeply disrespectful speech, such as public hate speech, is not the only form of speech that poses a threat to democratic public discourse. Some speech ­threatens to lead to great social harms and injustices without necessarily conveying disrespectful views. This is, most notably, the case of public misinformation. To complement the present inquiry, the next chapter will therefore turn to the challenges involved in countering public misinformation. This next examination will deepen the present chapter’s conclusions in two ways. First, we will see that some of the discursive strategies I have introduced to counteract public hate speech—in particular, positive counterspeech—can also be of use when responding to public misinformation. Second, the case of misinformation will expose new challenges for these discursive strategies; and, in doing so, will help us further refine them.

4

Countering Public Misinformation 1. Introduction Public discourse is rife with misinformation about political matters. We are all familiar, for example, with the proliferation of conspiracy theories. After 9/11, so-­ called ‘truthers’ falsely claimed that the terrorist attack on the World Trade Center was in fact an inside job. And throughout Barack Obama’s presidency, countless rumours alleged (incorrectly) that his birth certificate was a forgery, that he was actually born in Kenya, and that, as a result, he was ineligible for the US presidency. Perhaps most outrageously, the 2016 Pizzagate theory falsely asserted that high-­ranking Democratic officials were running a child sex ring from a pizza joint in Washington, DC.1 Yet public misinformation can spread falsehoods without positing conspiracies. Many of the most influential cases of misinformation simply promote false claims about the content or expected consequences of policies. At the time of the Affordable Care Act’s (ACA) passage, for instance, it was widely and incorrectly reported that the ACA would create ‘death panels’ whose role was to decide whether elderly or disabled Americans were eligible for medical care. In other respects, the ACA benefited from misinformation. When selling his bill to the public, Obama falsely insisted that, under its provisions, everyone who liked their current healthcare plan would be able to keep it. Public misinformation also played an infamous role in the United Kingdom’s 2016 decision to leave the European Union (EU). Most notably, the Leave campaign repeatedly affirmed that leaving the EU would save the UK (and its National Health Service) 350 million pounds every week.2 Not even science is safe from public misinformation. Misinformation regularly surrounds scientific questions that bear closely on matters of policy. To this day, climate deniers publicly maintain that man-­ made climate change is not happening—or at least, that there is no scientific consensus on whether it is happening. Likewise, ‘anti-­vaxxers’ vigorously insist (in the face of all credible scientific evidence) that the MMR vaccine causes autism. And, in the midst of the 2020 COVID-­19 pandemic, social media came to be saturated with false rumours

1  BBC (2011, 2016); Smith and Tau (2011).

2  Holan (2009); Jacobson (2013); Blitz (2018).

Democratic Speech in Divided Times. Maxime Lepoutre, Oxford University Press (2021). © Maxime Lepoutre. DOI: 10.1093/oso/9780198869757.003.0005

Countering Public Misinformation  107 about medical treatments, including rumours suggesting that the virus could be cured by eating garlic.3 In brief, there is no shortage of utterances publicly spreading falsehoods about matters that are relevant to politics. Nor are these many falsehoods simply circulating among a small minority of people. Political misinformation can, and often does, spread on an enormous scale. This is particularly true on social media, where false information can be shared at great speed and across immense networks. In the month preceding the 2016 US presidential election, for example, American adults on average encountered between one and three false stories about the candidates.4 The pervasiveness of misinformation about political matters poses a significant threat to the value of democratic public discourse. Misinformation may, on the face of it, seem less immediately abhorrent than hate speech. While hate speech is invariably offensive, misinformation can sometimes come across as merely benighted or ridiculous. But it is in fact no less dangerous. Like hate speech, misinformation risks turning public discourse into a source of injustice and social harm. The present chapter will investigate how we should go about responding to this problem. For reasons that I will touch on below, legal restrictions are far less popular as a remedy for political misinformation than they are as a remedy for public hate speech. Accordingly, my focus here will almost exclusively be on the use of ‘more speech’ (or ‘counterspeech’) as a remedy for misinformation. In other words, unlike in Chapter  3, I will not be offering a detailed comparative assessment of legal restrictions and counterspeech. Instead, I will predominantly be examining what norms should govern our speech when it targets public misinformation about political matters. My investigation will unfold as follows. I begin by offering a working definition of misinformation and outlining the problem it poses for democratic public discourse (Section 2). Next, I briefly explore why legal restrictions are commonly taken to be unpromising as a response to this problem (Section 3). I then turn to counterspeech, and diagnose a challenge for countering misinformation with more speech: in virtue of the distinctive way misinformation affects listeners, misinformation can continue to influence listeners’ beliefs after it has been corrected (Section 4). To overcome this ‘stickiness’, I will argue that our speech-­ based response to misinformation should be both positively framed (Section 5) and diachronic (Section 6). This argument builds on Chapter 3’s prescriptions in two ways. First, it extends those prescriptions, by showing that ‘positive’ forms of counterspeech—which 3  Van der Linden et al. (2017); Godlee and Marcovitch (2011); Mikkelson (2020). 4  Allcott and Gentzkow (2017). On the scale of misinformation, see also Vaidhyanathan (2018, ch. 7) and Cassam (2019, ch. 2).

108  Democratic Speech in Divided Times I  initially introduced as a response to public hate speech—are also useful as a remedy for political misinformation. At the same time, my argument here will also qualify and supplement Chapter 3’s conclusions. That is, my discussion will also expose limitations of positive counterspeech, which apply both when countering misinformation and when countering hate speech. And, by developing the idea of ‘diachronic’ counterspeech, it will help to overcome these limitations in both cases.

2.  The Problem of Misinformation By ‘misinformation’, I am referring to speech that disseminates or promotes falsehoods. And I understand ‘political matters’ in a relatively loose sense. Thus, misinformation about political matters is a broad category. Among other things, it includes speech that promotes falsehoods about the content of policies (e.g., ‘the ACA will implement death panels’); about technical or scientific claims that bear on the evaluation of policies (e.g., ‘man-­ made climate change is not happening’); about political figures (e.g., Obama was born in Kenya); and about groups of people (e.g., ‘the Democratic Party are collectively running a child sex ring’). Understood in this way, many instances of hate speech arguably count as a species of misinformation about political matters. After all, hate speech often rejects the equal standing of its targets by disseminating falsehoods about groups of people—for example, by claiming that immigrants are parasites; that Muslims are terrorists; and so on. But hate speech and misinformation can come apart in two respects. First, not all hate speech involves promoting or disseminating falsehoods. Some instances of hate speech reject the equal standing of other groups via utterances that are true (e.g., the leader of a racist party who shouts ‘We want you out!’). And other instances of hate speech, such as imperatives, are not capable of being true or false (e.g., ‘Get out of this country!’). Second, not all misinformation about political matters takes the form of hate speech. This should be obvious from the examples with which we started. Misinformation about vaccines, the Affordable Care Act, Brexit, climate change, and so on, does not deny the equal standing of other groups. Yet even though it does not, it can still be hugely problematic for democratic public discourse. How is non-­hateful public misinformation problematic? There are at least two ways, which correspond to the two functions of public discourse that I have been emphasizing. The first problem is epistemic. A key argument for inclusive public speech, as discussed in previous chapters, is that exchanging arguments and voicing personal narratives seems an effective way for diverse groups to pool the

Countering Public Misinformation  109 politically relevant knowledge that is distributed between them.5 Pooling knowledge, in turn, can help people identify better policies, politicians, or parties. The pervasiveness of misinformation about political matters turns this argument on its head. If public discourse is saturated with misinformation, it may lead us to pool falsehoods rather than knowledge—and by extension, it may lead us mistakenly to accept unjust or harmful policies, politicians, or parties. If, because of public misinformation, people come to doubt the reality of man-­made climate change, or to doubt the safety of vaccines, they may end up voting against policies aimed at mitigating climate change, and refusing to vaccinate their children. The second problem has to do with domination. As we have seen, another important function of inclusive public discourse is that it allows voters to hold policymakers accountable. Participants can publicly challenge policymakers, and thus make them more responsive to their concerns.6 In this way, public discourse protects subjects of political power from domination. But misinformation risks undermining this function. For people effectively to challenge policymakers, they need to have a sense of how policymakers are exercising power. If misinformation deprives people of this knowledge, then it prevents them from holding policymakers accountable. Hence, it exposes them to domination. Notice that, unlike the first problem, this second problem applies even when misinformation does not result in bad policies, politicians, or parties being selected. Take Obama’s false claim that the Affordable Care Act would allow voters who liked their current healthcare plan to keep it. Assuming that the ACA is a good policy, this use of misinformation might get people to accept a better policy than they otherwise would have. Yet it remains morally problematic. By inducing false beliefs in voters, this piece of misinformation erodes their control over political decisions. Insofar as they are misinformed about what the ACA consists in, they are not in a position to ensure that it tracks their concerns. To disentangle these two problems raised by public misinformation from the distinctive problem raised by public hate speech (namely, that it attacks the dignity of its targets), my investigation of misinformation in this chapter will focus almost entirely on non-­hateful forms of misinformation. Still, as advertised above, my investigation will nevertheless have some implications for countering public hate speech, and I will flag these implications where relevant. Before proceeding, a final clarification is needed. I have defined misinformation solely in terms of its content—namely, the fact that it disseminates or promotes falsehoods about political matters. But one might be tempted to adopt a more restricted focus. In particular, commentators often restrict their attention to cases of misinformation where the speaker knows that what they are saying is false. In her influential examination of ‘fake news’, for example, Regina Rini stresses that 5  Introduction; Chapter 1, Section 5; Chapter 2, Section 4. 6  Introduction; Chapter 1, Section 2; Chapter 3, Section 4.

110  Democratic Speech in Divided Times fake news is ‘known by its creators to be significantly false’.7 There can be good reasons for this focus. First, those who knowingly spread misinformation may be more blameworthy than those who do so inadvertently. So, if one is interested in assessing the culpability of speakers, it might make sense to focus on the former kind of case. Second, intentional misinformation is a common phenomenon. The false stories that circulated during the 2016 US presidential election (such as the story claiming that Pope Francis had endorsed Donald Trump) often were created by deceitful actors who intended to spread falsehoods. For my purposes, however, a focus on intentional misinformation (or ‘disinformation’) is too narrow. Many influential false stories are created by people who sincerely believe them. Quassim Cassam is adamant, in his analysis of conspiracy theories, that producers of such theories often wholly accept them.8 It is entirely conceivable, for example, that the conspiracy theorists who started the rumour that 9/11 was an inside job actually believed that this was the case. Moreover, even when a piece of misinformation does originate with an actor who knows that it is false, it is often disseminated by people who do not. For instance, even if many of the fake news stories that emerged during the 2016 US presidential election were manufactured by deceitful agents, the speakers who shared them over and over again on social media did so, often, because they believed the stories.9 The important point is this. Many claims that introduce or disseminate political falsehoods are voiced by agents who actually believe what they say. And, crucially, this does not necessarily make these utterances less dangerous for public discourse. When a conspiracy theorist spreads his delusions, and when someone who has been duped by a fake news story shares it on Facebook, this risks inducing false beliefs about politically important matters in others. Insofar as unintentional misinformation has this effect, it—like intentional misinformation—risks eroding the epistemic and anti-­domination functions of public discourse. As a result, the rest of this chapter will consider both intentional and unintentional cases of public misinformation.

3.  The Difficulty of Restricting Misinformation In response to the problems posed by public misinformation, one direct solution might be to adopt legal restrictions that prohibit misinformation about political

7  Rini (2017, 45). See also Vaidhyanathan (2018, ch. 7), Gelfert (2018, 108), and Aikin and Talisse (2018). For a contrasting view, see Pepp et al. (2019). 8  Cassam (2019, chs. 1, 3). Pepp et al. (2019, 73) make a similar observation in the context of fake news. 9  Allcott and Gentzkow (2017).

Countering Public Misinformation  111 matters, and that inflict penalties (such as fines or incarceration) on those who violate this prohibition. Thus far, liberal democracies have largely opted against this solution. Legal bans on the creation or dissemination of politically relevant falsehoods are relatively common in non-­democratic or non-­liberal states that involve little freedom of the press (such as Singapore, Russia, China, and Egypt). Yet this is by and large not the case in liberal democracies.10 There are exceptions, of course. In 2018, for example, France adopted a law empowering judges to suppress politically relevant misinformation in the run-­ up to elections. But this remains quite uncommon.11 Moreover, where they have appeared in liberal democracies, legal prohibitions on misinformation have been hugely controversial.12 To be clear, liberal democracies typically do suppress some subcategories of misinformation. Chapter  3 noted that many such democracies legally regulate hate speech (which, as we saw in the previous section, often involves a form of misinformation). And nearly all liberal democracies prohibit defamation, or speech that unfairly damages the reputation of others. Still, many of the most damaging cases of politically relevant misinformation (e.g., misinformation about vaccines, climate change, healthcare, Brexit, and so on) constitute neither hate speech nor defamation. What this means is that much of the misinformation that I am concerned with in this chapter is generally permitted in liberal democracies. This situation could conceivably change in the future. Concerns about public misinformation have grown in recent years, as social media has allowed dangerous falsehoods to proliferate on a greater scale and at greater speed than previously. Perhaps, then, other liberal democracies will soon follow in France’s footsteps. But there are indications pointing in the opposite direction. A recent report issued by the European Union, for example, explicitly cautions member states against the legal regulation of misinformation.13 Why is this the case? Why, in particular, have liberal democracies seemed more reluctant to ban (non-­hateful) political misinformation than they have been to ban public hate speech? In its most straightforward form, a legal ban on misinformation prohibits the dissemination or promotion of politically relevant 10  Other states that have enacted such restrictions include Bangladesh, Burkina Faso, Cambodia, Cameroon, Ivory Coast, Kazakhstan, Malaysia, and Myanmar. For overviews of these restrictions across the world, see Funke and Flamini (2019) and Library of Congress (2019). 11  See, e.g., Funke and Flamini (2019) and Library of Congress (2019). It was widely reported that Germany’s Network Enforcement Act (NetzDG) prohibits fake news. But, as Funke and Flamini (2019) observe, this is in fact misleading. The NetzDG strengthens the enforcement of existing speech-­related laws, including laws against hate speech and defamation. However, it does not expand the range of utterances that are illegal. The UK also has a law relating to false statements during elections. But it only targets statements about the personal conduct or character of politicians. As a result, Rowbottom (2012, 526–30) gives a pessimistic assessment of its ability to address the problems created by public misinformation. 12  McAuley (2018); Financial Times Editorial Board (2019). 13  European Commission (2018, 19–20, 30).

112  Democratic Speech in Divided Times falsehoods. Understood in this way, the idea of restricting misinformation seems deeply problematic. In the first place, ill-­motivated states could use such a ban to suppress legitimate dissent, whether or not that dissent is accurate. This is not a hypothetical problem. In Singapore, as in many other states that have introduced anti-­misinformation restrictions, these restrictions have routinely been used to prosecute civil activists and members of the opposition.14 Second, even if the state does not misuse restrictions on misinformation, they nevertheless risk having a severe chilling effect on public discourse. That is, people are likely to refrain from making public claims unless they are absolutely sure that these are correct.15 This, in turn, threatens to stifle many fruitful exchanges—particularly if the give and take of reasons in public discourse is one of our most important tools for finding out which views are correct in the first place. Of course, as we saw in Chapter 3, legal restrictions on hate speech also run the risk of being misused, and of chilling public speech.16 But a prohibition on politically relevant misinformation raises a comparatively greater threat of misuse and chilling. This is because such a prohibition bears on a far greater range of utterances. A hate speech ban only bears on claims having to do with the equality or inequality of other groups. By contrast, a generic ban on misinformation bears on any claim that is relevant to political decisions. Accordingly, it gives ill-­ motivated states greater scope to punish dissent; and it produces more situations where speakers are likely to engage in unwarranted self-­censorship. To avoid these problems, we might try to make the legal prohibition of misinformation more precise. Rather than prohibiting the dissemination of any falsehood that bears on political matters, we might instead prohibit the dissemination of specific falsehoods that are deemed particularly dangerous. This is not unlike what restrictions on public hate speech do. They specify certain contents that cannot publicly be uttered—namely, the denial or rejection of other social groups’ basic social standing. To the extent that legal restrictions on misinformation, too, are made specific in this way, they are less likely to be misused or to chill legitimate speech. Yet this proposal is in fact less applicable to the case of politically relevant misinformation than it is to the case of hate speech. For Waldron, a necessary circumstance that makes it possible safely to specify the hateful utterances that are to be banned is ‘the cessation of serious controversy’ on questions of basic social equality. Take the case of racial equality. According to Waldron,

14  Palma et al. (2020). See also Funke and Flamini (2019) for further empirical evidence of this phenomenon. 15  Rowbottom (2012, 525). 16  On the problem of misuse, see Chapter  3, Section  5. On the chilling of political speech, see Chapter 3, Section 4.

Countering Public Misinformation  113 the fundamental debate about race is over—won; finished. There are outlying dissenters, a few crazies who say they believe that people of African descent are an inferior form of animal; but for half a century or more, we have moved forward as a society on the premise that this is no longer a matter of serious contestation.17

In other words, for Waldron, the falsity of claims alleging that some racial, ethnic, gender, or other groups are fundamentally inferior to others is by and large ‘settled’ in contemporary liberal democracies. These claims have been discussed for a very long time. The evidence needed to discredit them has repeatedly been canvassed. Many of our social institutions (including national celebrations and school syllabi) are premised on their falsity. And there is overwhelming acceptance, among the public, of their falsity.18 This, in Waldron’s estimation, makes it safe to specify them as claims that can be regulated. We can disagree with Waldron about just how socially settled questions of equality are in contemporary liberal democracies. In Chapter  3, we saw that Waldron is sometimes insufficiently sensitive to the extensiveness of public hate speech, and (relatedly) to the moral costs of regulating it.19 Thus, there may well be more than just ‘a few crazies’ who are tempted by racist or hateful ideologies. But one thing is nonetheless clear. Comparatively speaking, the basic equality of different racial, ethnic, or gender groups is far more ‘settled’ than the falsity of much of the damaging misinformation with which I am concerned in this chapter. By comparison with racial or gender equality, the truth about how much the UK pays the EU, about what provisions the Affordable Care Act contains, or about what treatments are effective against COVID-­19 is hardly settled. These issues are comparatively novel, as a result of which there have been fewer opportunities for discussing them, for establishing the truth of the matter in these domains, and for premising our social practices on this truth. In consequence, many of us could conceivably learn something new from discussing them. The upshot is this: because much of the political misinformation that corrupts public discourse bears on issues that are far from settled, it is often difficult safely to specify the particular pieces of misinformation that should be prohibited. Of course, there is another possible way of making legal prohibitions on misinformation more restricted. Instead of limiting such prohibitions to a set of pre-­specified falsehoods, we could restrict them based on the intentions of the speaker. In particular, legal bans on misinformation might prohibit utterances that knowingly or deliberately disseminate politically relevant falsehoods. This is,

17  Waldron (2012, 195). 18  Krysan and Moberg (2020). 19  Chapter 3, Section 4. See also Saul (2019) for a discussion of whether norms surrounding racial equality are currently eroding in the United States.

114  Democratic Speech in Divided Times indeed, what most existing restrictions on misinformation (including the French law prohibiting election-­time misinformation) tend to do. The problem with this alternative proposal is not simply, as Rowbottom notes, that it can be very difficult to discern and prove the speaker’s true intentions.20 The more fundamental difficulty is that, as we saw in the previous section, public misinformation can have devastating consequences even when (as is often the case) the agents who spread it genuinely believe it. So, a legal restriction that focuses exclusively on intentional misinformation fails adequately to address the dangers of public misinformation. In sum, legal restrictions on misinformation risk being either dangerously over-­inclusive, or (if we try to make them less inclusive) insufficiently effective. This may explain why, to date, liberal democracies that are willing to ban hate speech have nevertheless been reluctant to implement such restrictions. Now, it may be that, in the future, creative new forms of restrictions will succeed in steering between these two challenges. But for the time being, legal restrictions do not seem to be a particular promising solution to the problem at hand. For the rest of this chapter, I will therefore focus instead on speech-­based strategies for countering public misinformation.

4.  The Stickiness of Misinformation How, then, should we counter public misinformation with more speech (or counterspeech)? The most straightforward way would be to publicly cite pieces of misinformation, authoritatively declare that they are false, and explain why this is the case.21 Speech-­based responses to misinformation in fact often take this form. In the run-­up to the UK’s 2016 EU referendum, the UK Statistics Authority publicly singled out the claim that the UK pays the EU 350 million pounds each week, and branded it ‘misleading’.22 Likewise, to counter the rumour that the Affordable Care Act would implement death panels, Barack Obama pointedly cited and criticized it in town halls across the US.23 Much in the same spirit, the Centre for Disease Control continuously maintains a web page affirming, in bold letters, that ‘there is no link between vaccines and autism’.24 And, during the COVID-­19 pandemic, fact-­checking websites consistently ran headlines stating that ‘garlic isn’t a cure for coronavirus’.25

20  Rowbottom (2012, 532, 534). 21  Cassam (2019, ch. 4) likewise recommends the point-­by-­point dismantling of conspiracy theories. 22  Dilnot (2016). 23  Lee (2009). 24  Center for Disease Control and Prevention (2015). 25  Spencer (2020).

Countering Public Misinformation  115 These are not isolated examples. The standard practice today when considering controversial public statements consists in publicly citing them, stating whether or not they are false, and articulating reasons for this verdict. This is illustrated, most notably, by the proliferation of fact-­checking websites. However, this popular and intuitive way of countering public misinformation with more speech is problematic. As we saw in Chapter 3, it is easier to make an idea or association of ideas salient, than to subsequently make it less salient. This, recall, is because rejecting an idea or association may inadvertently make one’s interlocutors pay even more attention to it. While Chapter 3 applied this insight to the case of hateful ideas, it extends readily to misinformation about political matters. When we respond to a piece of misinformation by publicly citing it and stating that it is false, we risk reinforcing its salience in the public sphere. This is particularly true when an authoritative source, such as a prominent expert body, issues the correction. The more the Center for Disease Control asserts that ‘vaccines do not cause autism’, the more salient the misleading association between vaccines and autism risks becoming. Similarly, authoritatively insisting that the UK does not send the EU 350 million pounds each week risks drawing attention to that very rumour. At first sight, it might be unclear why this is a problem. This is because it might not be immediately obvious why the salience of misinformation is a bad thing. In the case of hate speech, Chapter 3 argued there exists a tight connection between salience and harm. The public salience of hateful propositions (e.g., Xs are lazy parasites) means that these propositions are up for debate in the public sphere. Since these propositions are typically inconsistent with the good standing or dignity of targets of hate speech, the upshot is that their good standing or dignity is up for debate in the public sphere. Therefore, the mere salience of hateful propositions jeopardizes the public assurance of dignity. This, however, does not seem to be the case with non-­hateful misinformation. Unlike with vilifying or hateful stereotypes, the fact that a falsehood about political matters is up for debate or at issue in the public conversation does not inherently constitute a harm. Indeed, the fact that there is a debate about whether the MMR vaccine causes autism, or about whether the UK pays the EU 350 million pounds each week, does not in and of itself call anyone’s standing into question. So, if the salience of political misinformation is not harmful in the same way that the salience of hate speech is, what nevertheless makes this salience problematic? In answering this question, we can benefit from contemporary advances in cognitive science. Cognitive scientists have shown that there is an intimate connection between the cognitive fluency of an association of ideas—roughly, how familiar one is with that association—and one’s disposition to believe it. As

116  Democratic Speech in Divided Times Lewandowsky et al. summarize, ‘in general, fluently processed information [. . .] is more likely to be accepted as true’.26 This finding helps appreciate what is problematic about the salience of political misinformation. When an association becomes salient, that association is thereby activated for people: their attention is drawn to it, and they are prone to represent it. Thus, the salience of an association is tightly linked to people’s cognitive familiarity, or fluency, with it. If, as cognitive scientists argue, the fluency of an association in turn disposes people to believe it, then the problem becomes clear. The salience of politically relevant misinformation is problematic quite simply because it increases the likelihood that people will accept the misinformation. The upshot for counterspeech is troubling. Insofar as countering misinformation with more speech risks increasing its salience, doing so may fail to correct the false beliefs originally induced by misinformation. In fact, by reinforcing the salience of misinformation, counterspeech could even help to consolidate these false beliefs.27 This worry is not merely theoretical. Social psychologists have widely reported a ‘continued influence effect’, whereby attempting verbally to correct falsehoods is ineffective or worse.28 In one study, for example, Center for Disease Control flyers distinguishing myths from facts about vaccines have been shown to backfire: people who read the flyer end up being more likely to misidentify myths as facts—and to oppose vaccination—than people who do not.29 Similarly, Adam Berinsky finds that, although rehearsing and correcting the ACA death panel rumours (as Obama did) initially makes people somewhat more likely to reject those rumours, this effect largely disappears after a few weeks.30 Thus, misinformation risks being ‘sticky’, or resistant to counterspeech. In particular, what seems like the most natural—and certainly the most popular— speech-­ based strategy for countering misinformation risks maintaining or 26  Lewandowsky et al. (2012, 212). See also Alter and Oppenheimer (2009, 228), Berinsky (2017, 245–7), and Jerit and Zhao (2020). 27  This analysis sheds light on why the widespread ‘sharing’ of misinformation on social media is so dangerous. Sharing is not necessarily a sign of support. People often share news items that they oppose, and they sometimes even do so in the context of debunking those items. But what sharing always does is draw attention to the shared item. Indeed, in their philosophical analysis of sharing, Pepp et al. (2019, 83–4) argue that sharing is similar to pointing: its essential function is to make something salient to others. Given the relationship between salience and belief, this makes sharing misinformation problematic. At best, sharing misinformation to debunk it risks being ineffective. And in the worst case, it could even help to induce or to consolidate false beliefs. 28  For a review, see Lewandowsky et al. (2012, 113–21). In recent years, some social psychologists have questioned whether corrections can actually backfire, in the sense of exacerbating the effects of misinformation. For example, Wood and Porter (2019) and Guess and Coppock (2020) find no evidence of a backfire effect. But regardless of whether corrections can backfire, the result that has remained steady is that corrections often fail to significantly diminish the effects of misinformation. See, e.g., Guess and Coppock (2018, 13), Wood and Porter (2019, 150–1), and Pennycook et al. (2018, 1874–5). 29  Lewandowsky et al. (2012, 115). 30  Berinsky (2017, 254–60).

Countering Public Misinformation  117 reinforcing the salience of misinformation. And, as a result, it risks being unsuccessful at offsetting the damaging consequences of misinformation on people’s beliefs. In what follows, I will consider what norms should govern counterspeech, if we wish to avoid or mitigate this salience-­induced stickiness.

5.  Positive Counterspeech Revisited In Chapter  3, I distinguished between ‘negative’ and ‘positive’ forms of counterspeech, and argued that positive counterspeech can help to alleviate problems relating to the salience of hate speech. What I now want to suggest is that the idea of positive counterspeech can also help overcome the salience-­ induced stickiness of misinformation.

5.1  Positive Counterspeech and Misinformation Negative counterspeech is speech that explicitly negates the content of the utterance that it is countering. The popular fact-­checking approach outlined in the previous section is a paradigmatic illustration of negative counterspeech. Indeed, applied to a false claim, fact-­checking involves citing a falsehood and overtly negating it (e.g., ‘there is no link between vaccines and autism’). In virtue of operating in this way, this negative approach tends to maintain or reinforce the salience of the targeted falsehoods. Positive counterspeech works very differently. Rather than directly contesting a distorted vision of the world, it affirms a correct vision of the world that is inconsistent with the content of the targeted utterance. So, when positive counterspeech targets a piece of political misinformation, it does not explicitly invoke that piece of misinformation and assert that it is false. Instead, it puts forward a true proposition that entails or implicates its falsehood. This, in turn, should help make misinformation less sticky. The stickiness of misinformation results partly from the fact that people are more inclined to accept salient propositions. Now, because positive counterspeech does not explicitly assert that its target is false, it avoids increasing the salience of its target. In fact, positive counterspeech actively aims to draw the public’s attention away from false propositions and towards true propositions instead. What might positive responses to misinformation look like in practice? Take the claim that the UK pays the EU 350 million pounds a week, which could benefit the National Health Service instead. As we have seen, the negative strategy here is to say: ‘That’s misleading. We don’t actually pay 350 million pounds each week.’ By contrast, a more positive strategy would be to conduct a public

118  Democratic Speech in Divided Times education campaign that avoids mentioning the 350 million pound claim and instead draws the public’s attention to facts that entail its falsity. Such a campaign might publicize facts about what the UK receives from the EU (including, if applicable, how free movement in Europe helps staff the National Health Service).31 And it might also reveal facts about what the UK actually pays the EU (expressed, perhaps, as a proportion of overall spending, so as to highlight its comparatively small size). This response would entail that the UK does not pay the EU 350 million pounds each week, and that it is misleading to posit a contrast between funding the EU and supporting the National Health Service. In doing so, however, it would avoid repeating the 350 million pound figure, and would therefore avoid reinforcing its salience. Or consider the myth that garlic can cure COVID-­19. ‘Negatively’ publishing a headline asserting that ‘garlic isn’t a cure for coronavirus’ could inadvertently reinforce the notion that there is some kind of connection between garlic and curing coronavirus, or indeed bring this notion to the attention of people who had never heard of it. To avoid this risk, a positive approach might instead advertise known treatments for coronavirus, while emphasizing that these treatments are exhaustive. It might, for instance, say the following: ‘There are many rumours about how to treat COVID-­19. To date, the following are the only treatments recognized and recommended by health experts: remaining hydrated, taking paracetamol, etc.’32 This pronouncement rules out the possibility that garlic is a recommended treatment for coronavirus. But it does so without making this spurious treatment salient. On the contrary, it draws the public’s attention to actual treatments. On the whole, states have predominantly deployed negative forms of counterspeech when responding to public misinformation. Many, for instance, have set up government-­run fact-­checking websites or radio programmes.33 But state-­ sponsored positive counterspeech is not unheard of. One promising illustration comes from the UK’s Rapid Response Unit (RRU), a team of digital and media experts assembled in 2018 for the explicit purpose of countering public misinformation. In April 2018, the UK, the US, and France launched airstrikes on suspected chemical weapons facilities in Syria. A wave of online misinformation soon followed—including, most notably, allegations that the chemical attacks that prompted the airstrikes were in fact ‘false flag’ operations that had been staged by the US. To counteract these rumours, the RRU ensured that government pages explaining the actual nature and origin of the airstrikes appeared higher up in Google searches than they otherwise would have

31  Blitz (2018). 32  National Health Service (2020). 33 As Funke and Flamini (2019) observe, this has notably happened in Brazil, Italy, Spain, and Mexico.

Countering Public Misinformation  119 (particularly for searches that included the term ‘false flag’).34 This response nicely exemplifies how the state might try to implement positive counterspeech: it refrains from overtly citing and negating falsehoods; it instead promotes accurate information that is inconsistent with these falsehoods; and it uses online search algorithms to make this accurate information more salient than the problematic falsehoods.35 If the analysis that I offered in Section 4 is correct—that is, if the stickiness of misinformation results in part from its salience—then positive forms of counterspeech such as these are likely to be more effective at countering misinformation than negative forms of counterspeech. This, to be clear, remains a theoretical hypothesis, and it requires further empirical investigation. Still, there is already some evidence supporting it. Notice, first, that the worrying evidence of a ‘continued influence effect’ mentioned in Section 4 almost exclusively involved cases of negative counterspeech. The CDC ‘myth versus fact’ flyers about vaccines worked by reiterating myths and labelling them ‘False’. Moreover, Berinsky’s observation that correcting ACA death panel rumours is ineffective applies primarily in cases where the rumours were explicitly repeated beforehand. It is harder to find evidence examining clear cases of positive counterspeech. But, to the extent that this evidence exists, it provides grounds for cautious optimism. For one thing, when Berinsky’s subjects did not repeat the ACA death panel rumour, corrections actually tended to reduce their support for the rumour.36 This suggests that correcting rumours without increasing their salience (as positive counterspeech does) is comparatively effective. Relatedly, there is evidence that people are more likely to reject false reports about how a fire started if, instead of simply being told that the initial report was false, they are given an alternative narrative of how the fire started.37 Nyhan and Reifler have shown that this last result extends to political cases. When considering rumours regarding a politician’s resignation, they find that people are far  more likely to reject rumours that the politician accepted bribes if, instead of simply being told that the politician did not accept bribes, they are offered an alternative explanation for the politician’s resignation.38 On the basis of similar observations, Lewandowsky et al. hypothesize that, when correcting misinformation, it is better to ‘emphasize the facts you wish to communicate

34  Aiken (2018). 35  Note that this strategy does not amount to censoring or suppressing misinformation. Making government pages appear at the top of Google searches is consistent with allowing misinformation to maintain a high position in these searches. Admittedly, however, this is a matter of degree. Moving false rumours down a few places in the search results does not suppress them. But moving these rumours down to the hundredth page of results would make them so difficult to find that it is almost equivalent to removing them altogether. 36  Berinsky (2017, 258). 37  Johnson and Seifert (1994). 38  Nyhan and Reifler (2015).

120  Democratic Speech in Divided Times rather than the myth’.39 This is entirely in tune with the positive approach to counterspeech. So, there appear to be both theoretical and empirical reasons for thinking that positive forms of counterspeech can attenuate the stickiness of misinformation. To a greater extent than negative forms of counterspeech (such as fact-­checking), positive counterspeech can reduce the impact of misinformation on the public’s beliefs.

5.2  The Limits of Positive Counterspeech Still, even if positive counterspeech is comparatively promising as a remedy for misinformation, it nevertheless faces important limits. Two problems relate to the audience of counterspeech. First of all, if members of the audience distrust the speaker (e.g., because the speaker comes from a social group that they dislike), then they may refuse to listen to the speaker’s positive counterspeech. Second, even if the audience seriously listens to positive counterspeech, they may inadvertently process that speech in a way that is biased. In particular, they may be motivated to accept the information being communicated only to the extent that it is congenial to the interests of their social group.40 These are both serious challenges, which in fact generate difficulties, not just for positive counterspeech, but also for inclusive public speech more generally. Accordingly, the next two chapters will be devoted to answering these broad challenges. Chapter 5 will address the first problem by examining to what extent goodwill and trust are needed for effective communication, and how we might use public discourse to regenerate them. Chapter 6 will then turn to the influence of group identity on the way people process information, and assess whether or not this influence really is a problem—and when it is, how that problem might be tackled. But even setting aside these broad concerns for the time being, positive counterspeech faces two further limits. To begin, this approach may not always be available as a strategy for opposing misinformation. Positive counterspeech affirms a view that entails or implicates the incorrectness of the targeted utterance without repeating or invoking its contents. So, it is available as an alternative to negative counterspeech only insofar as there exists a way of presenting the negation of the falsehoods in question that does not explicitly invoke those falsehoods. For instance, in the case of the 350 million pound EU claim, positive counterspeech is possible because there is a way of saying what the UK actually 39  Lewandowsky et al. (2012, 123). See also Ecker et al. (2011, 570) and Jerit and Zhao (2020). 40  See, e.g., Jerit and Zhao (2020, section 4).

Countering Public Misinformation  121 pays the EU—for instance, citing the percentage of the UK’s GDP that goes to the EU—that is not simply the negation of the 350 million pound figure. For some forms of misinformation, however, there is no way of saying something that entails or suggests their falsehood besides invoking and rejecting their contents. Conspiracy theories seem especially liable to generate this difficulty. According to Cassam, such theories invariably suggest that things are not how they seem. As a result, ‘there is almost no explanation that is too bizarre for the Conspiracy Theorist’s taste [. . .] If how things are isn’t how they look, who is to say how strange the actual truth is.’41 Thus, conspiracy theories inherently tend to doubt the natural or obvious explanations of events, and to propound highly counter-­intuitive or improbable explanations instead. This, in turn, is problematic. Because conspiracy theories are typically so improbable, it is difficult verbally to oppose a conspiracy theory in a way that does not inadvertently invoke it. Consider again Birtherism, the theory that Obama’s birth certificate is a forgery, that he was born in Kenya, and that he was therefore ineligible for the US presidency.42 Negative counterspeech—saying that Obama was not born in Kenya—clearly reactivates the association between Obama and Kenya. But it is not immediately clear what the positive alternative would be. Vigorously affirming the true proposition that Obama was born in Hawaii would be a highly unusual thing to do, were it not for the concern that Obama was born outside the US. Put differently, because the salience of the Birtherist myth is the main reason why Obama’s birthplace is politically worth talking about, it is difficult to say anything on this topic without inadvertently invoking or reactivating the Birtherist myth. Given this background fact about salience, insisting that Obama was born in Hawaii risks, like negative counterspeech, calling attention to the concern that he was born in Kenya. The problem so far has been that genuine positive counterspeech—which opposes misinformation without reinvoking its contents—may sometimes be unavailable. But there is another problem. Even when we can use positive counterspeech to reverse the salience of misinformation (together with its effects on people’s beliefs) positive counterspeech might still come too late to stop misinformation from producing harms. This concern is especially relevant during elections.43 If, shortly before an election, misinformation deceives voters into believing that the Affordable Care Act will implement death panels, people may vote on the basis of this false belief before prospective counterspeakers have a chance to put forward a response. In a situation such as this one, counterspeech (positive or not) will fail to stop the harmful consequences of misinformation. This is presumably why, as Section 3 41  Cassam (2019, 22).

42  Smith and Tau (2011).

43  Rowbottom (2012, 523).

122  Democratic Speech in Divided Times explained, France implemented a legal restriction on misinformation that specifically applies in the run-­up to elections. During elections, there might not be enough time verbally to respond to the many pieces of misinformation that are circulating before they generate harms. The rest of this chapter will investigate how counterspeech might overcome these two limitations. Before proceeding, however, notice that the two limits in question are in fact not specific to the case of misinformation. It is also true that positive counterspeech could be unavailable for some forms of hate speech (such as fantastical conspiracy theories accusing minorities of white genocide). And positive counterspeech may also come too late when it responds to hate speech. Even if it successfully reassures targets of hate speech of their good standing, it cannot retroactively undo the fact that, before they received this reassurance, they may well have experienced severe psychological distress. So, although my focus in what follows will continue to be on politically relevant misinformation, it should also inform our attempts at verbally responding to hate speech.

6.  Diachronic Counterspeech To address the two limitations just outlined, we need to adopt a more expansive understanding of the temporality of counterspeech. Simply put, counterspeech should not merely be interpreted as speech that responds to misinformation after the fact.44 Rather, counterspeech is better understood as diachronic—as a continuous process, extended over time, which can precede as well as follow misinformation. This might seem strange. One might think that, if one is countering something, then that thing must already have happened. But this is not true. We can counter a poison retroactively, by using antidotes. Yet we can also counter it pre-­emptively, by inoculating people against it. In her influential analysis of ‘toxic speech’, Lynne Tirrell has argued that we can think about counterspeech along similar lines: we can counter harmful speech both through verbal ‘antidotes’ and through verbal ‘inoculations’.45 In line with Tirrell’s metaphor, the diachronic interpretation of counterspeech suggests that it is a mistake exclusively to focus on using our speech as a post hoc remedy for misinformation. In addition, counterspeech can build resistance against future instances of misinformation. This diachronic understanding of counterspeech is entirely consonant with the broader systemic approach to public discourse that I have been relying on in previous chapters. The systemic approach emphasizes that public speech takes place in many different arenas, and generates insights that move from arena to 44  See, e.g., McGowan (2018). 45  Tirrell (2018, 136–9). See also Richards and Calvert (2000, 569–74).

Countering Public Misinformation  123 arena over time.46 What this highlights is that public discourse is a temporally extended process. As part of this extended process, the impact of new utterances (such as utterances that spread misinformation) depends not just on how we later respond to them, but also on the context in which they first arise. And previous utterances play a fundamental part in defining this context. How can previous utterances pre-­emptively condition the discursive context to make it less hospitable to misinformation? They can do so, in short, by making members of the public more resilient in the face of misinformation. This aim is most clearly exemplified by Sweden’s creation, in 2018, of an agency explicitly devoted to bolstering the ‘psychological defence’ of its citizens.47 The stated purpose of this agency is to make sure that, when misinformation arises, the public is able to resist its appeal. There are at least two ways in which speech can pre-­emptively make the public more resistant to subsequent instances of misinformation. The first is simply to expound and widely diffuse politically important facts.48 This conditioning makes it more difficult for false propositions subsequently to gain a foothold in the conversation. Insofar as listeners are knowledgeable about a subject matter, they are less vulnerable to being swayed by distorted claims about politically relevant matters. For example, misleading claims about EU funding will be less readily accepted if they contradict propositions on these topics that are already part of the audience’s common ground. The second strategy is to warn the public about untrustworthy sources. This might involve identifying and exposing particular sources that are known to be untrustworthy. Or, alternatively, it might involve informing the public that many sources are untrustworthy, and giving them information that helps them distinguish trustworthy sources from untrustworthy ones.49 In either case, the aim of this strategy is to pre-­emptively undermine the perceived authority of untrustworthy sources, so that their false claims fail to affect people’s beliefs in damaging ways. This second idea is akin to the silencing effect that philosophers of language often ascribe to hate speech and pornography. Hate speech and pornography, it is often said, prevent minorities and women from contributing effectively to conversations. One of the ways in which they do so, as we saw in Chapter 3, is by stripping these groups of their authority.50 When hate speech persistently represents racial minorities as inferior, for instance, they may lose the social standing needed for their speech to be taken seriously. What I am suggesting here 46  This is particularly clear in Hertzberg’s (2018) articulation of the systemic approach. 47  Swedish Civil Contingencies Agency (2018). 48  For an example of this approach, see the UK’s Government Communication Service (2018). 49 The UK is currently experimenting with implementing this strategy in schools. See, e.g., Murray (2019). 50  Chapter 3, Section 5.1.

124  Democratic Speech in Divided Times is that this silencing effect can also be used for good, by pre-­emptively challenging the authority of those who would otherwise spread political misinformation. There is evidence that both of these pre-­emptive strategies can be effective. Cook et al., for instance, find that telling subjects about the scientific consensus on anthropogenic climate change neutralizes the causal impact of subsequent climate-­related misinformation.51 This corroborates the first pre-­emptive strategy outlined above. By giving people authoritative testimony that man-­made climate change is real, pre-­emptive speech introduces this proposition into the common ground of shared beliefs. This in turn makes it harder for subsequent climate-­ related misinformation to gain assent. There is also support for the second pre-­emptive strategy: casting doubt on the authority or credibility of unreliable sources. Lewandowsky et al. report that ‘misinformation effects can be reduced if people are explicitly warned [. . .] that information they are about to be given may be misleading’.52 For example, when people are warned that upcoming information about climate change may  come from non-­experts, and given some indication of how to identify non-­ experts, the influence of subsequent misinformation on their beliefs ­virtually disappears.53 In virtue of allowing pre-­emptive strategies such as these, the diachronic understanding of counterspeech helps alleviate the two limitations that I have been concerned with. The first limitation, remember, is that misinformation risks producing irreversible harms before it is verbally countered (negatively or positively). This worry, as we have seen, is especially salient when misinformation arises shortly before an election. Pre-­emptive strategies speak directly to this issue. Because they allow us to disarm misinformation before it arises, they eliminate the temporal gap during which a piece of misinformation has been uttered but not yet countered. Thus, they remove the interval during which this piece of misinformation might otherwise, unopposed, have generated harms. Second, pre-­emptive strategies can also help tackle the forms of misinformation (such as conspiracy theories) that cannot readily be opposed in a positive way. The problem with these utterances is that, once they gain traction, it is difficult to oppose them in a way that does not backfire. To circumvent this difficulty, a more promising approach is to pre-­emptively condition the discursive context so that they never gain traction to begin with. Now, if the only way to do so were to pre-­emptively deny conspiracy theories (‘You may hear next week that Obama was born in Kenya. But rest assured: he was born in Hawaii.’), one might worry that this strategy simply reproduces the initial problem: pre-­emptive denial risks inadvertently reinforcing the salience of falsehoods by invoking them. Crucially, however, one of the main pre-­ emptive forms of counterspeech involves casting doubt on the authority of speakers, rather than directly criticizing 51  Cook et al. (2017, 10). 52  Lewandowsky et al. (2012, 116). 53  Cook et al. (2017, 15). See also Van der Linden et al. (2017).

Countering Public Misinformation  125 the content of their utterances. The recommendation that emerges from the diachronic understanding of counterspeech, then, is that we should pre-­emptively warn people about, say, fake news sites spewing conspiracy theories. We can do this, to reiterate, by pre-­emptively identifying specific fake news sites as unreliable; or, alternatively, by warning people that such unreliable sites exist and teaching them how to identify these sites. As discussed above, such warnings can drastically reduce the audience’s vulnerability to misinformation. So, diachronic counterspeech is doubly helpful: not only does it show how counterspeech might wholly forestall the damaging effects of misinformation, but it also provides guidance for handling particularly resilient kinds of political misinformation. These two benefits extend to the case of hate speech. If counterspeech pre-­ emptively assures minorities of their good standing, or erodes the authority of prospective hate speakers, this can help to disarm subsequent instances of hate speech. That is, insofar as minority groups have authoritatively and continuously been assured that they are valued members of society, and that hateful citizens (though vocal) are the true social outliers, they may experience less psychological distress when later confronted by hateful utterances. And this pre-­ emptive approach applies equally well to hateful utterances promoting conspiracy theories for which positive counterspeech is unavailable. It is important not to overstate the present section’s argument. I am not claiming that diachronic counterspeech will always be entirely successful at defusing misinformation (or hate speech). Clearly, it will not. When a piece of misinformation is already widespread, it is quite simply too late for counterspeech to pre-­emptively disarm it. And even when pre-­emptive counterspeech remains an option, it is not foolproof. For instance, some speakers may be so authoritative that their verbal influence cannot fully be pre-­emptively disabled. This is a real problem in non-­ideal conditions, where irresponsible heads of state often actively promote misinformation. In highlighting diachronic counterspeech, I am therefore making a more restricted point: that (much as legal restrictions aim to do) counterspeech can in  principle defuse misinformation before it ever occurs—and so, before it ­produces harms; and that it can do so even in hard cases where positive counterspeech is unavailable. So, while it cannot fully obviate the limits introduced in the previous section, a diachronic approach to counterspeech nevertheless significantly alleviates them.

7. Conclusion Public misinformation about political matters poses a deep threat to the value of democratic public discourse. This problem, I have argued, is very difficult adequately to address via legal norms that restrict public misinformation. Perhaps

126  Democratic Speech in Divided Times for this reason, liberal democracies have thus far remained reluctant to adopt such restrictions—more reluctant, indeed, than to adopt legal norms restricting public hate speech. As a result, I have emphasized the importance of countering misinformation with more speech. But not just any kind of speech will do. The effects of misinformation on people’s beliefs often prove stubborn in the face of attempted corrections. To mitigate this stickiness, two discursive norms should govern counterspeech. In the first place, instead of negatively citing and rebutting false claims (as the popular practice of fact-­checking typically does), we should strive, where possible, to correct misinformation in a way that is positively framed. The second norm has to do with the temporality of counterspeech. To counter the most resilient forms of misinformation, and to limit the damage that misinformation occasions, counterspeech should be an extended process, which precedes as well as follows misinformation. Together with Chapter 3, these conclusions yield a relatively unified account of the norms that should govern our speech-­based responses to harmful speech. Albeit for different reasons,54 positive counterspeech is needed both when responding to politically relevant misinformation (much of which is not hateful) and when responding to hate speech (which need not involve misinformation). Likewise with diachronic counterspeech: for both hate speech and misinformation, there are important benefits to pre-­empting, as well as reacting to, the problematic utterances in question. There is nonetheless at least one difference between the two cases, relating to the agent of counterspeech. In the case of hate speech, I emphasized that it is crucial for counterspeech to be driven by the state. Comparatively speaking, I have said little about the state’s role in verbally countering misinformation. This is not to deny that the state can have an important role in doing so. After all, given its vast expressive resources, the state may help us disseminate policy-­related facts. Nevertheless, state involvement is especially important when it comes to countering hate speech. This is because hate speech typically works by attacking its targets’ social standing, and the state has distinctive authority when it comes to defining and upholding the social standing of its members. By contrast, for at least some categories of misinformation, non-­state bodies (e.g., independent scientific institutions) might have greater authority than the state. Thus, while it matters both in the case of hate speech and in the case of misinformation that counterspeech be positively framed, diachronic, and driven by the state, this last stipulation is especially pressing when it comes to opposing hate speech.

54 This, as we saw in Section  4, is because the salience of hate speech and the salience of misinformation are problematic for different reasons. The salience of hate speech risks undermining the public assurance of dignity. And the salience of misinformation risks, via a fluency effect, inducing false beliefs.

Countering Public Misinformation  127 More generally, this concludes the first part of this book: namely, my investigation of the norms that ought to govern public discourse in non-­ideal settings. I have considered whether speakers should offer shared reasons to one another; what the appropriate place of emotions is in public discourse, with emphasis on negative emotions such as anger; and, finally, how we should counteract harmful forms of speech, in particular hate speech and misinformation. What we need to consider now—and what will occupy me for the rest of this book—is whether the norms I have defended here (and the resulting normative picture of public ­discourse) are sufficiently realistic. In other words, can this normative picture of public discourse really guide us in contemporary divided democracies? Or have I underplayed the challenges facing democratic public speech?

5

The Problem of Goodwill 1. Introduction 1.1  The Presuppositions of Democratic Public Discourse What norms should govern democratic public discourse in conditions marked by widespread social divisions? The past four chapters have defended a set of discursive norms that are geared towards identifying and counteracting pressing social problems (such as group-­ based harms and injustices) in such non-­ ideal conditions. To prevent political power from being exercised in an arbitrary or dominating way, speakers should appeal to shared considerations in the final and most formal stages of public deliberation.1 But to offset the exclusionary tendencies of this first norm, public discourse should also leave significant room for non-­argumentative and emotionally charged forms of speech.2 This last suggestion might seem overly inclusive, by opening the door for intensely angry narratives and eventually hateful narratives. However, we should not wholly reject this implication. Angry public narratives perform a crucial epistemic function in divided democracies, by highlighting persistent injustices that may otherwise have been overlooked or misunderstood.3 As for public hate speech, there are strong (though defeasible) grounds for thinking that we should counter its group-­based harms, not via legal norms, but via public speech that is driven by the state,4 positively framed,5 and diachronic.6 In fact, this species of ‘counterspeech’ is doubly useful as a remedy for harmful public speech: it constitutes a promising strategy for repudiating not only public hate speech, but also the politically relevant misinformation that threatens to corrupt the epistemic and anti-­ domination functions of public discourse.7 The relevance and effectiveness of these discursive norms presuppose the existence of some non-­ideal conditions, including: that many people experience systemic injustices; that people are often ignorant of each other’s experiences; that some people embrace hateful views regarding others; and that the public sphere is rife with (intentional or unintentional) falsehoods. In offering a defence of 1  Chapter 1, Section 2. 4  Chapter 3, Section 5. 7  Chapter 4, Sections 5–6.

2  Chapter 1, Section 5. 5  Chapter 3, Section 6.

3 Chapter 2. 6  Chapter 4, Section 6.

Democratic Speech in Divided Times. Maxime Lepoutre, Oxford University Press (2021). © Maxime Lepoutre. DOI: 10.1093/oso/9780198869757.003.0006

132  Democratic Speech in Divided Times discursive norms whose desirability depends on such circumstances, my aim has been to develop a political ideal that is relevant to, and capable of guiding action in, our world. But at this point, one might question just how close this normative account of democratic public discourse really is to actual, non-­ideal conditions. In other words, one might worry that, despite my aims to the contrary, the desirability of the norms I recommend continues to presuppose background conditions that are significantly better than those that obtain in actual divided democracies. To address this worry, Chapters 5 to 7 will assess the extent to which my normative account of democratic speech is complicated by, respectively, (1) the lack of goodwill between different groups (or ‘affective polarization’), (2) the fact that members of society often know very little about politics, and (3) the fragmentation of the public sphere into insulated communities. Finally, Chapter  8 will consider whether ‘minimalistic’ political ideals, that do away with inclusive public discourse, are better able to navigate these problematic background conditions than the political ideal I have offered.

1.2 Outline The first potential concern is this. It seems plausible to think that fruitful public communication, where speakers exchange arguments and narratives, requires significant levels of goodwill between speakers. Yet goodwill often appears to be lacking in contemporary democracies, which are marked by deep distrust or ‘affective polarization’. Consequently, unless inclusive public discourse plays a role in generating the missing trust and goodwill—a suggestion that those who voice this concern typically also reject—then democratic public discourse seems largely unhelpful in actual conditions.8 This concern certainly should inform broadly ‘deliberative’ theories of democracy, such as my own, that emphasize the importance of having people from different walks of life exchange reasons and narratives with one another. Nevertheless, the present chapter will argue that, in significant measure, it is possible to stave off its  pessimistic conclusion. After presenting the problem of  goodwill (Section  2), I  mount a two-­faceted attack on its premises. First, I qualify the assumption that productive public discourse demands high levels of goodwill, by drawing together observations made in Chapters  1 and  2. Specifically, I argue that the division of labour between different arenas in the system of public discourse alleviates the motivational demands made on

8  E.g., Johnson (1998), Shapiro (2003, ch. 1), Parvin (2015).

The Problem of Goodwill  133 the  vast majority of members of society, and helps to circumvent these demands when they remain too high (Section 3). Even with this qualification, however, the problem of goodwill remains challenging. The second part of my argument therefore pushes back against another assumption: that trust and the attending goodwill cannot arise via non-­ideal public discourse (Section  4). In particular, I investigate how aspects of public discourse that are especially widespread in non-­ideal conditions—namely public hypocrisy (4.1), anger (4.2), and (most controversially) the occurrence of public hate speech (4.3)—have properties that can be made to contribute to generating trust. In saying this, I am not claiming that all of our trust-­building measures should be discursive, or speech-­based. Indeed, the trust-­building strategies that I theorize often involve a mutually supportive combination of speech-­based and non-­speech-­based tools. My point is the more nuanced one that inclusive and public speech remains useful even where antecedent trust and goodwill are largely lacking, because it offers important resources for generating trust and goodwill. Two preliminary observations are in order. The first concerns empirical evidence. The latter part of my argument considers the potential trust-­building effects of public speech. One might think that this is simply an empirical question, which should interest social scientists rather than philosophers. However, when social scientists gather data, they rely upon theoretical hypotheses that guide their research. As I stressed in earlier chapters, the conceptual insights developed by philosophers can contribute valuably to articulating and enriching such theoretical hypotheses. In this chapter, then, I articulate hypotheses informed by philosophical work on topics such as trust, anger, and hypocrisy. At the same time, I will strive to indicate what empirical assumptions these hypotheses depend on, how far these are consistent with existing evidence, and where— in light of these hypotheses—further empirical research may be needed. The second point to note is that the features of democratic public discourse that I will suggest have potential for generating trust and the attending goodwill are consonant with the discursive norms defended in Chapters 1 to 4. Therefore, this chapter’s defensive exploration of the problem of goodwill will lend further support to the picture of democratic speech that I have been advancing.

2.  The Problem of Goodwill The problem of goodwill is by no means new, and it has prominently been articulated by political theorists such as James Johnson and Ian Shapiro.9 But it owes its

9  Johnson (1998, 173–4); Shapiro (2003, 22–30).

134  Democratic Speech in Divided Times most sustained and general reformulation to Phil Parvin, whose argument can roughly be reconstructed as follows:10 (1) Democratic public discourse that is effective at solving or mitigating social problems (such as injustices) presupposes a high level of goodwill between participants, which itself cannot obtain unless they trust each other to a significant degree. (2) Contemporary democracies involve low levels of mutual trust. (3) Democratic public discourse is ineffective at building trust and the attending goodwill. (4) So, democratic discursive processes will not be effective at combatting social problems until greater trust arises, and will not play an important role in building that trust.11 To illustrate, consider a concrete way in which public discourse might require goodwill. Many accounts of deliberative democracy, including John Rawls’s or Joshua Cohen’s, require that interlocutors be civil towards one another. Civility, here, means respecting one’s interlocutors by appealing only to considerations that are suitably shared or public. In Chapter 1, I defended a similar norm, which I referred to as the ‘shared reasons constraint’, on the grounds that it helps to ward off problematic forms of domination. Now, according to Parvin’s interpretation, there is some evidence that Rawls and Cohen want the norm of civility to regulate the public reason-­giving of ordinary citizens and public officials alike.12 This demand for civility, Parvin suggests, is problematic in non-­ideal conditions. It seems unlikely that citizens will be civil unless they trust one another to reciprocate: plausibly, I will not commit to respectfully constraining the claims I make unless I feel that I can rely on my interlocutors also to fulfil a commitment to discursive civility.13 But actual levels of trust seem too low to sustain this 10  Parvin (2015). 11  I focus on whether public speakers have enough mutual goodwill to conform to certain discursive norms. But Parvin (2015, 408–9) is also sceptical that they will participate at all. Put differently, he is concerned with the rate as well as the quality of discussion. I only examine the quality issue, as there is evidence that rates of engagement in political discussion are actually quite high. In their comprehensive study, Jacobs et al. (2009, 4) find that ‘citizens engage in more extensive and meaningful public talking than previously suspected’. See also Neblo et al. (2010). According to Chambers (2012, 63n3), previous studies (e.g., Mutz  2006) reported different findings because they adopted an inadequate operationalization of discursive participation. Nevertheless, these recent studies find that the quality of public discussion is low. E.g., Jacobs et al. (2009, 4), Searing et al. (2007, 612). Thus, the pressing question seems to be less whether mass participation is possible, and more whether mass participation of the right kind is possible. 12  Parvin (2015, 409–12). 13  Here, I am assuming Hawley’s (2014, 9–12) account of trust, where A trusts B to x when A believes B to be committed to x-ing, and relies on B to x. There are competing accounts of trust: for instance, others suggest that it involves relying on others to x out of goodwill (e.g., Jones (1999, 68–9)), or out of a lack of ill-­will (e.g., Lenard (2012, 23)). My arguments here will not hinge on these differences.

The Problem of Goodwill  135 civility: ‘increased inequality, social fragmentation, and a weakening of traditional associational life in contemporary states has resulted in a weakening of common bonds of citizenship’.14 In the US, for instance, racial tensions between black and  white Americans are high,15 and the mutual dislike (or ‘affective polarization’) between Republicans and Democrats has increased dramatically since the 1960s.16 What is worse, since (according to the present objection) non-­civil public debate is incapable of producing trust and the attending civility, we will need to generate the missing trust via policies that are external to inclusive public discourse.17 It is worth emphasizing the significance of this problem. The primary conclusion is that without trust and goodwill, democratic processes of inclusive public discourse will be largely ineffective both at combatting social problems (such as injustices) and at generating the antecedent trust that would enable them to combat social problems. A corollary of this is that until we build that trust, our instruments for generating trust and combatting social problems will be external to public discourse. They will, in other words, be non-­discursive.18 And since we are allegedly far from having adequate levels of trust and goodwill,19 this means that inclusive public discourse will not be a useful instrument in the near future. Given that I am concerned with how inclusive public discourse can be a pro­duct­ ive tool for combatting social problems in non-­ideal conditions such as our own, this last upshot is devastating. How can we avoid these upshots? It is undeniable that non-­discursive trust-­ building strategies will be needed to improve public discourse. But this implication alone is not overly problematic. After all, deliberative theorists have long acknowledged that, in non-­ideal conditions, political talk must be supplemented by other forms of political action.20 And, in fact, Chapter 7 will explicitly examine a strategy for combatting intergroup dislike and for improving political discourse that is external to public discourse—namely, the use of so-­called ‘integrative’ pol­ icies. What we need to show, then, is not that non-­discursive strategies such as these are unnecessary. Rather, it is that inclusive discursive processes also have an important part to play in conditions marked by low trust and goodwill. To do so, I will challenge premises (1) and (3) in turn.

3.  What Does Fruitful Public Discourse Really Presuppose? The systemic understanding of public discourse helps appreciate that fruitful public discourse—public discourse that is effective at combatting social 14  Parvin (2015, 414–15). See also Putnam (1995). 15  Massey and Denton (1993, 91–6). 16  Iyengar et al. (2012, 412–21). 17  Parvin (2015, 418–20). 18  Parvin (2015, 420). 19  Parvin (2015, 414–15). 20  E.g., Fung (2005), Mansbridge et al. (2010), Anderson (2010).

136  Democratic Speech in Divided Times problems—is less demanding in terms of trust and goodwill than it initially appears. By embracing this more sophisticated understanding of public discourse, we can therefore substantially weaken premise (1). Since the argument here draws together observations made in Chapter 1—when evaluating the shared reasons constraint—and Chapter 2—when examining trust-­based obstacles to the contagious transmission of anger—my discussion will be relatively brief. The systemic approach to public discourse, recall, claims that public discourse occurs in a system composed of many arenas, which differ in size, function, and formality, and which are connected more or less closely to one another. Why does this matter? First, as I observed in Chapter 1, different discursive norms may be appropriate in different arenas. Participants in formal arenas, whose primary function is to produce coercive decisions, should appeal to reasons that are suit­ ably public or shared (at least in the final stages of discussion). But this requirement need not apply to informal arenas (such as newspaper opinion sections) whose main purpose is to help participants form considered opinions.21 Thus, different arenas need not be equally demanding in terms of goodwill and trust. In particular, the most demanding norms—like the shared reasons constraint, or as Rawls calls it, ‘civility’—may well be required only of public officials taking part in highly formal deliberation. Hence, even if Parvin were right that some deliberative theorists hold ordinary citizens to standards of civility,22 there is nothing necessary about this: we can and should recognize the appeal of civility without demanding it of all citizens. Nevertheless, this first response can only take us so far. There are reasons to think that, even in arenas that do not involve strong constraints—such as civility constraints requiring that speakers appeal to shared reasons—fruitful public discussion still requires excessive goodwill. If so, the problem of goodwill also applies to accounts of public speech or to arenas of public speech that involve less constrained forms of communication. Why might this be? Inclusive public discourse, I argued previously, matters partly as a means of holding political decision makers accountable, by making them responsive to one’s concerns.23 But for this to be true, one’s interlocutors must be willing to listen to one’s complaints. Furthermore, if one’s interlocutors listen but are unwilling to change their minds or try to answer honestly, it might seem unclear in what sense they are being responsive to one’s concerns. Call these attitudinal dispositions—to listen, to be open to changing one’s mind, and to respond earnestly—the disposition to engage sincerely. Sincere engagement does intuitively seem required for fruitful public engagement. Accordingly, Iris Marion Young, despite largely

21  Chapter 1, Section 4.1. 22  See, however, Chapter 1, Section 4.2, for evidence contradicting Parvin’s interpretation of Rawls on civility. 23  Introduction; Chapter 1, Section 2.2.

The Problem of Goodwill  137 rejecting the aforementioned civility constraint, claims that participants must have ‘a disposition to listen to others, [. . .] [to] make an effort to understand them by asking questions’ and ‘to be willing to change [their] opinions’.24 This is problematic from the standpoint of trust. Intuitively, I am unlikely to listen, change my mind, and respond earnestly unless I trust my interlocutors to reciprocate.25 So, in divided societies where citizens distrust each other, even the weaker attitudinal requirement of sincere engagement may be too demanding. Premise (1) still looms. One might reply by challenging the value of sincere engagement.26 But for now I want to accept the prima facie plausible assumption that sincere engagement is needed for fruitful public discourse. Instead, my second point is that the systemic approach contributes to weakening premise (1) in another way, which Chapter 2 introduced. The systemic approach underscores how discursive arenas with overlapping constituencies can interact with each other. Most significantly for our purposes, a system of public discourse makes networked communication possible. Suppose A and C distrust each other too much to engage sincerely, but that A and B trust each other sufficiently, as do B and C. In such circumstances, A can sincerely engage with B in one sphere. In turn, B can sincerely engage with C, carrying over the information, narratives, and arguments gleaned from her ­discussion with A.27 Consider how this second point differs from the first. The first was that adopting the systemic approach to public discourse can lower how much goodwill is required in immediate communicative interactions, by waiving the most stringent discursive norms in some spheres. The second is that even when immediate communication remains too demanding in terms of goodwill, we can circumvent this problem by having third parties mediate exchanges between affectively polarized participants. Contra premise (1), then, high levels of goodwill and trust do not seem necessary for fruitful exchanges, at least not in the long run. Nevertheless, this last clause (‘at least not in the long run’) highlights the ­principal limitation of my second point. In contexts of deep distrust, there might not be a mediator B who is trusted by both A and C. If so, the network of mediators connecting A and C may have to be extremely long, so that the process of  exchanging narratives and arguments becomes excruciatingly slow. In the immediate future, then, distrust and the attending lack of sincere engagement may reduce the effectiveness of democratic public speech at countering social problems.

24  Young (2000, 24–5). See also Mansbridge et al. (2010, 66), Schwartzman (2011, 384–7), and Bächtiger et al. (2010, 49). 25  Parvin (2015, 417). 26  E.g., Markovits (2006), Runciman (2008). 27  Chapter 2, Section 4.3.

138  Democratic Speech in Divided Times To overcome this problem, we must resort to substantial trust- and goodwill-­ building mechanisms. Part of this work will, to reiterate, be external to public discourse. In particular, as we will see in Chapter 7, we may need to promote the creation of more socially integrated arenas. Such integrated arenas, I will suggest, can foster greater trust between their members, and can bolster the network of mediators that connects affectively polarized citizens. But we cannot rest content with an approach that is purely external to public speech. If, as premise (3) asserts, inclusive public speech can play no role in rebuilding trust or goodwill, then its overall usefulness in the near future remains somewhat limited. Thus, although the systemic approach does blunt the force of the problem of goodwill, we must go further and challenge premise (3). Accordingly, in what follows, I will explore how democratic public discourse supplies tools for rebuilding the trust and attending goodwill needed for sincere engagement.

4.  Can Non-­ideal Public Discourse Produce Goodwill? Why deny, as premise (3) does, that democratic discursive processes can generate trust and the attending goodwill? For some critics, the problem is that very little evidence supports this claim. Shapiro, for example, states that ‘there is no particular reason to think deliberation will bring people together’.28 Parvin goes further, asserting that empirical evidence largely suggests that having diverse groups talk about politics is ineffective at bringing them together.29 In fact, however, the evidence is more mixed than critics suggest. While some of it is indeed negative,30 much of it is not.31 For example, as part of a deliberative opinion poll taking place in Northern Ireland, Robert Luskin et al. invited Protestants and Catholics to deliberate together on the future of local schools. Though the issue was divisive, they found that deliberating caused members of both groups to regard each other as more trustworthy.32 Perhaps, then, the problem is less that talking together cannot build trust and goodwill, and more that the kind of talk that can do so is unlikely to arise in non-­ideal circumstances.33 This refined criticism strikes home against many existing arguments for using inclusive discourse as a trust-­building tool. Consider Amandine Catala’s defence of deliberative trust-­ building. To guarantee that deliberation will generate trust, she recommends ‘implementing strict rules of 28  Shapiro (2003, 27). 29  Parvin (2015, 418). 30  E.g., Conover and Searing (2005), Searing et al. (2007, 610–11), Sunstein (2002). 31 E.g., Dryzek and List (2003, 10–12), List and Koenig-­Archibugi (2010, 103–9), Fishkin and Luskin (2005), Anderson (2010, 123–7). Even Searing et al., who produce some of the evidence Parvin refers to, actually acknowledge that deliberation can produce fellow feeling (2007, 612). 32  Luskin et al. (2014, 132). 33  Lenard (2005, 375–6).

The Problem of Goodwill  139 deliberation [. . .]: for example, listening carefully, speaking respectfully, being responsive to others’ contributions’, etc.34 This, however, is unsatisfactory. We are investigating how public discourse can generate the trust that is necessary for sincere engagement. But in explaining how it might do so, Catala presupposes that participants have been made to sincerely engage with each other. So, she presupposes the very dispositions we are trying to generate. A related concern might apply to Luskin et al.’s deliberative poll in Northern Ireland. Although the participants did come from divided social groups, deliberative polls involve highly atypical settings: the discussions they host occur in small groups; they typically focus on well-­defined, sometimes relatively technical, policy problems; and they are highly structured and facilitated by trained mod­er­ators.35 Consequently, one might worry that such deliberative experiments do not tell us much about the more typical and larger-­scale public discourse that pervades non-­ideal societies. But this refined worry too is somewhat overstated. Some of the evidence in favour of discursive trust-­building does look at interpersonal discussion that occurs in non-­ideal conditions. Notably, Christian List and Matthias Koenig-­ Archibugi argue that the public debates that culminated in the creation of the International Criminal Court (ICC) played a crucial part in promoting trust and cohesion between participants. Significantly, this case involved wide-­ranging discussion, rather than discussion taking place in highly controlled and restricted settings: a complex network of formal and informal dialogue occurred between diverse participants, including not just government officials, but also NGOs and activists.36 Moreover, the discussions took place in a transnational setting, where it is typically assumed that, other things being equal, bonds of trust are weaker than in national settings. Consequently, their case study supplies prima facie grounds for thinking that even large-­scale, non-­ideal discursive processes can play an important trust-­building function.37 In citing this countervailing positive evidence, my point is simply that the empirical question concerning public discourse’s efficacy at generating trust is more open and nuanced than critics of discursive trust-­building often claim. Therefore, it is worth exploring further, both empirically and philosophically. Now, what is lacking in the above positive evidence is a detailed analysis of precisely what features of the non-­ideal situation were instrumental in generating trust and goodwill. In the ICC case, was trust-­building made possible despite non-­ideal features of the situation—perhaps by the fact that participants had similar socio-­ economic backgrounds? Or did non-­ ideal features also play a 34  Catala (2015, 436). 35  Luskin et al. (2014, 118). 36  List and Koenig-­Archibugi (2010, 103–9). 37  For more positive evidence pertaining to non-­ideal conditions, see Searing et al. (2007, 612), Maddison (2015, 1015), and Steiner (2012, ch. 2).

140  Democratic Speech in Divided Times contributory role? And if they did, which ones contributed, and how? While some of the evidence regarding discursive trust-­building provides grounds for optimism, this lack of clarity makes it hard to derive much normative guidance regarding non-­ideal public discourse. To remedy these shortcomings, I wish to investigate what specific features of the public discourse that characterizes non-­ideal conditions might play a role in enhancing trust and goodwill. My aim, though, remains modest. Drawing on philosophical work concerning topics such as trust, hypocrisy, and anger, I  will develop hypotheses concerning what properties of non-­ideal public ­discourse may be useful, and how. Along the way, I will strive to indicate how these the­ or­ et­ ic­ al hypotheses square with existing empirical evidence, and where they call for more empirical research aimed at testing and refining these hypotheses. So, my intention is not to say that non-­ideal political discourse always helps build trust overall. Clearly, such discourse is capable of damaging as well as improving trust and the attending goodwill. But it is to enhance our understanding of which features of non-­ideal discourse have potential for trust-­building. Armed with such an understanding, political philosophers and social scientists will be in a better position to explore under what conditions public speech can be harnessed to play an all-­things-­considered positive role in building trust and goodwill. More specifically, I will explore how three features that characterize non-­ideal public discourse—hypocrisy (Section  4.1), anger (Section  4.2), and (counter-­ intuitively) the occurrence of public hate speech (Section  4.3)—offer resources for promoting trust and sincere engagement.

4.1  Public Hypocrisy and Trust Consider a conversation taking place amidst very low levels of mutual goodwill and trust. A and B are publicly debating a policy issue, but A is unwilling to listen, respond, or change her mind in response to B’s claims. In part, this may be because A simply wants to pursue her self-­interest, rather than the public interest. But it may also be partly because, even insofar as A believes the right thing to do is to engage with B in a sincere and public-­spirited way, she does not trust B to reciprocate. Even then, provided that there is a widespread expectation that participants should be public-­spirited and sincere, publicity creates incentives for A to appeal to public or shared values. As Jon Elster famously argues, ‘there are certain arguments that simply cannot be stated publicly. In a political debate it is prag­mat­ic­ al­ly impossible to argue that a given solution should be chosen just because it is

The Problem of Goodwill  141 good for oneself ’.38 Now, Elster is almost certainly overstating his argument when he claims that brute appeals to self-­interest are impossible. Such appeals do occasionally occur.39 But the point remains that public scrutiny creates pressure to frame one’s arguments in public-­spirited terms. Therefore, even if A is solely motivated to pursue her self-­interest, the need to gain public support for her favoured policies strongly encourages her to appeal to public or shared values. Once A does so, it becomes possible to challenge her on these grounds. Suppose a politician publicly defends tax cuts for the wealthy by appealing to the value of freedom. Her opponents can then observe that if she truly cares about freedom, she should support redistributive taxation instead, since severe poverty makes people unfree. This puts the official under pressure. If she disregards the public challenge that she is being inconsistent, or answers this challenge by appealing to her own self-­interest, she risks appearing incompetent or hypo­crit­ ical, and therefore losing support. Hence, the pressures of publicity incentivize her to answer the challenge, and to do so in ethical terms. The idea, then, is that ill-­motivated participants in public debate can be induced to listen to ethical challenges, respond to them, and eventually change their stance, by what Elster calls the ‘civilizing force of hypocrisy’.40 Public attention creates pressure for speakers to hypocritically appeal to public ideals that they may not believe in. In turn, the desire not to be exposed as hypocrites constrains them into being more responsive interlocutors. This effect of public hypocrisy is not mere theoretical speculation. On the contrary, Frank Schimmelfennig argues that it is crucial to explaining why the European Union expanded to Central and Eastern Europe in the early 2000s. Although many powerful EU member states felt that enlargement was against their interest, they were rhetorically trapped into accepting the proposal. European integration had long been publicly justified by appeal to the ideology of a pan-­European community of liberal-­democratic states. Advocates of expansion exposed inconsistencies between their opponents’ past rhetoric of pan-­ Europeanism and their later opposition to Eastern expansion. To avoid being shamed as hypocritical—a prospect that might have impaired future cooperation with other member states—recalcitrant member states publicly changed their positions.41 38  Elster (1997, 12). 39  Gutmann and Thompson (1996, 126–7). 40  Elster (1998, 111). See also Chambers (2004, 405) and Dryzek (2000, 46–7). 41  Schimmelfennig (2001, 48). Naurin (2007) offers contrasting evidence, which he believes warrants scepticism regarding the civilizing force of hypocrisy. However, his evidence seems inconclusive. Naurin’s research focuses on business lobbyists. But unlike with public officials, people do not expect lobbyists to act according to public interest: if anything, the opposite is expected. Accordingly, Naurin (2007, 32) himself concedes that ‘lobbyists presumably have less to lose from being publicly revealed than, for instance, political parties’. Naurin might nonetheless persist as follows: he finds not just that there is no civilizing effect of publicity on lobbyists’ behaviour, but that lobbyists are less public-­ spirited in public than in closed debates with public officials. Again, however, it seems possible to

142  Democratic Speech in Divided Times One might object that public hypocrisy nevertheless does not generate the kind of goodwill we wanted. The question has been how to get participants to trust one another, to acknowledge one another as trustworthy, so that they might then be disposed to engage sincerely with each other—that is, to listen, respond earnestly, and eventually change their minds. But the pressure of publicity on self-­ interested actors seemingly does not achieve this. After all, A is hypocritical, not sincere, in her dialogical engagement. Even if she listens to and answers B’s challenges, nothing guarantees that she actually believes the answers she gives. And A can change her publicly held position without genuinely changing her mind.42 Similarly, A is not made trustworthy with respect to discursive engagement. While she might reliably engage in dialogue and answer challenges, she does so purely out of self-­interest, not out of a genuine commitment to sincere engagement on these issues. Mere reliability, most philosophical analyses of trust agree, is insufficient for trustworthiness.43 So, public hypocrisy is insufficient to build lasting bonds of trust, and relatedly, insufficient to make participants sincerely engage with each other. The first thing to say here is that this shortcoming is relatively unproblematic. This is because A remains functionally equivalent to someone committed to sincere engagement. She is forced, by fear of reputational damage, to listen to others, respond, and eventually change her stated position. Furthermore, unless her response is persuasive, she risks losing the support of listeners. So it may seem inconsequential whether or not she believes what she says: the fact that she is forced to act as if she were sincerely engaging in a trustworthy manner is enough to make her responsive and accountable to public criticism. Indeed, it might not matter to Eastern European states seeking EU membership whether member states that eventually assented really had grown to prefer enlargement. What matters is that public hypocrisy forced them to assent. We can go further than this initial reply. Besides increasing A’s accountability to B, hypocritically engaging in reason-­giving might actually, over time, make A committed to sincerely engaging with B. The first reason has to do with cognitive dissonance. According to Elster, hypocritical public speakers persistently have to defend values they do not believe in, which is a powerful source of psychological discomfort. To reduce this painful cognitive dissonance, individuals may actually come to believe the ideals they defend.44 Hence, however hypocritical and self-­ interested to start with, publicly upholding the ideal of sincere engagement can eventually make A genuinely embrace this ideal. explain this observation without abandoning the hypothesis that publicity usually induces a kind of civilizing hypocrisy. In private settings, lobbyists try to win politicians over. And since politicians face pressures to be public-­spirited, lobbyists presumably need to show them that their proposals can be defended in public-­spirited terms. 42  Mansbridge et al. (2010, 73–4n28). 43  See Hawley (2014), and note 13, this chapter. 44  Elster (1997, 12).

The Problem of Goodwill  143 Some might have doubts about this first possibility, which depends on ­ sychological theories of cognitive dissonance whose precise explanatory scope p remains contested.45 But Robert Goodin suggests an alternative and perhaps less controversial possibility. Sometimes, the problem is not that individuals are purely self-­interested, but that they are only weakly committed to certain ideals. If so, Goodin suggests, public hypocrisy helps remind them of the ideals they already embrace, and underscores the implications of those ideals.46 This second process requires that participants have some measure of pre-­existing commitment to the ideal in question. However, this is not exceedingly worrying in our context: there is good evidence that citizens are often at least weakly committed to the ideals of listening to, trying to understand, and answering the claims of other ­participants.47 In such conditions, publicly induced hypocrisy can reinvigorate the commitment that very many participants have to sincere engagement, and thereby make them more trustworthy interlocutors. Now, the present strategy for building goodwill and mutual trust, which proceeds through publicly induced hypocrisy, does presuppose that some background conditions hold. First, as mentioned earlier, there must be a widespread expectation that public speakers should engage in a sincere and public-­spirited manner. But this, Elster suggests, is not necessarily unrealistic. It is consistent with many citizens not embracing this ideal, and with some individuals endorsing it but being too weak-­willed to comply with it.48 Second, the speaker’s discourse must be made public via monitoring, so that other participants can impose sanctions when speakers fail to act in a sincere and public-­spirited manner.49 This, too, does not seem out of reach. For one thing, the sanctions in question need not be very heavy-­handed. They could simply consist of reputational damage, lack of future cooperation, or, for elected officials, the threat of being voted out of office. As for monitoring mechanisms, Patti Lenard observes that they already ‘are a regular part of democratic life’.50 For instance, media outlets habitually report speakers’ present claims and past statements, thereby enabling listeners to hold them accountable when the two are inconsistent. Still, one problem with monitoring speakers is political ignorance. Even if reliable media sources monitor public agents, participants may fail to avail themselves of this information. Worse, as we saw in Chapter  4, they might instead defer to the unreliable sources that spread misinformation about political matters. For now, I assume that we can achieve significant levels of reliable

45  For discussion, see Festinger (1957) and Cooper and Fazio (1984). 46  Goodin (1992, 137). 47  Conover and Searing (2005, 276). 48  Elster (1995, 248–9). 49  Although I emphasize publicity’s value, this is compatible with thinking that secrecy can also play a valuable role. On how to combine public and secret deliberation, see Gutmann and Thompson (1996, 114–17) and Chambers (2004, 405–9). 50  Lenard (2012, 143–5).

144  Democratic Speech in Divided Times monitoring by the public. Chapter 6, however, will consider the more pessimistic scenario, by directly examining contexts where citizens are highly ignorant about politics. Let us conclude. Public discourse in non-­ideal conditions is rife with hypocrisy. But this hypocrisy can actually serve a positive purpose in conditions where trust and goodwill are low. That is, it can dispose participants to listen to one another, offer responses, and eventually change their stated positions. Admittedly, this mechanism presupposes some felicitous background conditions, including the existence of mechanisms for monitoring and sanctioning public speakers. However, these are often already part of democratic politics. Additionally, if we make further—albeit modest—empirical assumptions (for instance, about cognitive dissonance), there is reason to think that public hypocrisy may over time generate a genuinely trustworthy commitment to sincere engagement. Importantly, however, what follows will not depend on these further empirical assumptions. Instead, I want to examine how, once we grant that participants are at least disposed to listen to each other, other typical features of non-­ideal public discourse can contribute to building trust and goodwill.

4.2  Anger and Trust In divided contexts, an important aspect of public discourse is the exchange of personal narratives, or testimony, including emotionally charged narratives. Chapters 1 and 2 recommended welcoming such narratives on epistemic grounds, to counteract the ignorance of some social groups regarding the experiences of other groups. But philosophers concerned with transitional justice frequently emphasize their usefulness for rebuilding trust in divided contexts. For example, Cécile Fabre argues that sharing difficult stories regarding what one has endured, often in a way that is suffused with negative emotions, can contribute importantly to transforming damaged social relations.51 In particular, she suggests that South Africa’s post-­apartheid Truth and Reconciliation Commission (TRC), where victims and perpetrators were invited to publicly tell their stories, helped to rebuild relations of trust. This suggestion—that emotionally charged narrative can play an important trust-­building role during transitional periods—is promising for the purposes of challenging premise (3) of the problem of goodwill. Transitional periods typically involve extremely low levels of mutual trust and goodwill. Therefore, if

51  Fabre (2016, 280). See also Chakravarti (2014), Maddison (2015, 1022–3), Mihai (2016), Steiner (2012, 84). Importantly, the success of TRCs at generating trust is mixed. While some evidence is positive (e.g., Gibson (2004)), some is not (e.g., Shaw (2005)). My aim is to develop hypotheses regarding what might explain the positive side.

The Problem of Goodwill  145 emotionally charged narrative can indeed heal damaged social relations during transitional periods, this means that, even in highly non-­ideal contexts, public discourse has the potential to help generate trust and the attending goodwill. Let us consider, then, what kinds of emotionally charged personal narratives might be useful in the process of rebuilding trust, and how. One might think that some kinds of emotionally charged testimony can contribute to rebuilding trust, while denying that testimony charged with anger can do so. This seems to be Nussbaum’s position. As Chapter 2 discussed, while Nussbaum acknowledges the importance of emotional appeals in divided settings, she explicitly asserts that angry narratives are deleterious for trust.52 However, excluding anger in this way would weaken my reply to the problem of goodwill, for two reasons. First, narratives of anger are pervasive in non-­ideal conditions, which characteristically involve widespread injustices. Thus, if we want to suggest—as I do—that typical features of non-­ideal public speech can be made to contribute to discursive trust-­building, we should try and find out whether angry narratives in particular can do so. Furthermore, discouraging angry public speech because it allegedly inhibits trust-­building procedures jars with my recommendation, in Chapter  2, that we give a substantial place to such speech. For these two reasons, the rest of this section will focus largely on angry personal narratives. Chapter 2 defended angry narratives by arguing that they have distinctive epistemic value, while accepting for the sake of argument Nussbaum’s claim that their impact on trust and goodwill is at best inert and at worst negative. I then argued that even conceding this, the systemic quality of public discourse helps to circumvent anger’s alleged bad consequences on trust.53 Here, by contrast, I wish to push back against the assumption that the influence on trust of angry narratives is categorically negative. This is of course not to say that their influence on trust is always positive overall. Establishing when exactly their influence is overall positive remains partly an empirical task. But that task, as I explained earlier, depends on theoretically informed hypotheses regarding what specific positive trust-­building effects angry testimony might have. My aim here is to develop such hypotheses. So let us ask: assuming that discursive participants listen to one another, how might voicing angry personal narratives contribute to building trust between them? It will help to have a case of angry testimony before us. In Sing the Rage: Listening to Anger After Mass Violence, Sonali Chakravarti offers the poignant example of Lendiso Richard Ndumo Galela, an anti-­apartheid activist who was tortured while in detention. Galela supplied the following testimony at the South African TRC:

52  Nussbaum (2016, 213, 233).

53  Chapter 2, Sections 4.3–5.

146  Democratic Speech in Divided Times There is no amount of money that can pay for a person’s dignity. That is a fact. I  am the equivalent of a corpse, I do not have any dignity left. I have been stripped of my dignity and that is what I wanted to say and I would just like to say that even though my physical appearance seems like that of a man I have been stripped of my manhood and a specialist told me that it would give, it would take three months for me to be treated and if nothing happens, I should know that I would be condemned sexually. [. . .] What I am saying is that I do not see what the Commission can do for me.54

Why might listening to this resentful narrative induce the listener to trust Galela? A recurrent suggestion among theorists of transitional justice, which Fabre not­ ably articulates, is that hearing such a narrative can have a humanizing influence: it helps the listener see the speaker as a human being, whose humanity is worthy of concern and respect. Anger, recall, represents its target as involving a moral violation or injustice.55 So, an angry personal narrative reveals that the speaker is grappling with a powerful sense of injustice (‘I am the equivalent of a corpse [. . .] I have been stripped of my dignity’), and discloses the difficult experiences that prompted this feeling (how the torture Galela sustained resulted in his impotence). Being confronted with the speaker’s psychological agitation and discovering the painful experiences he has endured helps us understand who he is and where he is coming from. Accordingly, it helps the listener perceive the speaker as an individual in their own right—as a human being with particular experiences and concerns.56 This humanizing influence is important. As Katherine Cramer and Elizabeth Anderson highlight in their recent examinations of urban/rural divisions and racial divisions in the US, distrust is often driven by the fact that individuals from estranged social groups perceive each other solely as members of a resented out-­ group, in a way that erases or obscures their shared humanity.57 Insofar as this is the case, angry narratives’ humanizing function seems promising with respect to trust-­building. But there is another, less examined, way in which listening to anger might be positively associated with trust: listening to angry testimony might enhance the listener’s disposition to act trustworthily because, in some cases, the speaker’s angry narrative itself constitutes a manifestation of trust. Telling one’s story can require disclosing deeply sensitive aspects of one’s life, particularly in conflict-­ ridden circumstances. This seems likely, not just when expressing grief or despair, but also anger. Since the appropriate target of anger is a moral violation, angry

54  Cited in Chakravarti (2014, 138). 55  Chapter 2, Section 2. 56 Fabre (2016, 264). See also Chakravarti (2014, 153), Maddison (2015, 1021), Halpern and Weinstein (2004, 571–9). 57  Cramer (2016, esp. 8–9); Anderson (2010, 44–50).

The Problem of Goodwill  147 testimony may sometimes involve revealing how one has sustained violating or degrading treatment. For Galela, giving testimony means expressing anger at a deeply humiliating injustice (‘I have been stripped of my dignity [. . .] I have been stripped of my manhood’). In such cases—where the injustice that the speaker’s anger targets is degrading or humiliating—delivering angry testimony implies baring oneself, and hence making oneself vulnerable to one’s listeners. Now, according to philosophical analyses of trust, voluntarily making oneself vul­ner­ able to others just is placing trust in them.58 Thus, in making himself vulnerable to his listeners, the speaker whose anger targets humiliating wrongs he has sustained is signalling trust in his listeners—in particular, trust that they will not take advantage of that vulnerability by, say, mocking him. Why does this matter? Philip Pettit has argued that when A places trust in B, this generates a new reason for B to act in a trustworthy way and eventually to reciprocate by extending trust. To place trust in another, Pettit suggests, is to show esteem for them. Since individuals typically seek the esteem of others, they desire to maintain this trust. Therefore, placing trust in someone gives the trustee a new reason, based in their desire for esteem, to act trustworthily. This is what Pettit calls the ‘cunning of trust’.59 In making himself vulnerable to his listeners through his anger-­infused narrative, then, Galela gives them a new reason to act in a trustworthy way. One might worry that the ‘cunning of trust’ does not apply to non-­ideal circumstances. This property of trust requires that the trustee desires the truster’s esteem. But we often recoil at being esteemed by people whom we deeply dislike.60 So, if the speaker and listener come from estranged social groups, one might think that the intended trust-­building effect will not occur. This worry derives much of its force from an ambiguity surrounding what kind of esteem the speaker is signalling to the listener. What makes us recoil from the esteem of our enemies is typically the worry that they extend this esteem for the wrong reasons. I worry about being appreciated by racists because of the thought that they appreciate me because I am not black. But, importantly, the kind of esteem being communicated here does not necessarily seem problematic in this way. At least part of what the speaker’s trusting reliance communicates, here, is the expectation that the listener will not take unfair advantage of their vulnerability. Intuitively, being respected as the kind of person who does not take unfair advantage of others’ vulnerability is something we have good moral reason to seek, even from those whom we dislike.61 To illustrate this last point, consider an example. During the 1963 Birmingham protests in Alabama, Commissioner for Public Safety ‘Bull’ Connor commanded 58  E.g., Baier (1986, 235), Pettit (1995, 208), Jones (1996, 12), Lenard (2012, 18–20). 59  Pettit (1995, 214). 60  Pettit (1995, 221). 61  This reason may be overridden. But it is a reason nonetheless.

148  Democratic Speech in Divided Times his men to target civil rights protesters with fire hoses. The protesters responded by getting to their knees and praying. In doing so, they made themselves vul­ner­ able to police forces, and signalled their expectation that the police would not exploit that vulnerability. Struck by this gesture, Connor’s men let the marchers through.62 On one interpretation, the protestors’ act of trust communicated a kind of esteem that Connor’s men had reason to value, even though it came from a group that they otherwise (we can plausibly assume) strongly resented. So far, I have emphasized how hearing testimony charged with anger might dispose the listener to engage in trusting relations. But delivering the narrative might also help. One might be unwilling to trust others because one feels resentful at being ignored or misrecognized. If so, publicly relating what one has endured seems to be part and parcel of what is needed to restore one’s trust.63 Again, Galela’s case is instructive. What keeps Galela from participating in social life is partly the perception that his ordeal has degraded him, and that this deg­ rad­ation is not properly acknowledged (‘no amount of money [. . .] can pay for a person’s dignity’). Accordingly, having his terrible experiences and his resulting anger acknowledged seems to play a constitutive role in restoring his willingness to engage in trusting relations. Giving impassioned testimony and receiving uptake, then, can be part of what makes speakers trust their listeners. A closely related point, which Chakravarti emphasizes throughout her analysis of anger in TRCs, is that giving angry testimony may also be instrumental to making the speaker trust others. Because angry testimony delivers facts about the world and the speaker’s psychological state—and more specifically, about the wrongs they see themselves as having sustained—it yields insights into what citizens need to restore trust.64 As we have just seen, Galela’s narrative indicates that financial reparations are the wrong kind of response to his anger. Instead, expressive acts aimed at making him feel dignified and respected seem a more fitting remedy. We have, then, several theoretical hypotheses regarding how testimonies charged with negative emotions such as anger might contribute to rebuilding trust and goodwill. They can help the listener perceive the speaker’s humanity; signal trust that incentivizes trustworthiness; make the speaker feel recognized as a member of society; and yield information regarding what measures are needed to restore the speaker’s trust. Importantly, although Galela’s narrative arguably exemplifies all of these trust-­ building functions, not all types of angry speech should be expected to perform these functions equally. For instance, the hypothesis that angry speech signals trust in a trust-­incentivizing way applies especially to cases where the injustice targeted by the anger is degrading or humiliating. To reiterate, this is because 62  Anderson (2010, 100). 63  E.g., Fabre (2016, 265), Tirrell (2015, 247). 64  Chakravarti (2014, 3, 20–1, 167).

The Problem of Goodwill  149 these are the cases where it is clearest that revealing the injustice suffered makes one vulnerable, and therefore constitutes an act of trust. This restriction need not apply to the other trust-­building functions of angry speech I have outlined. Delivering her testimony might conceivably help the angry speaker feel ac­know­ ledged, or yield information regarding how to restore her trust, or make listeners appreciate her humanity, irrespective of whether the injustices she is reporting were humiliating.65 So, far from suggesting that all instances of angry narrative will perform these positive trust-­building roles equally, the theoretical hypotheses I have outlined help explain why we can expect different kinds of anger to be effective at different functions. This matters, because existing empirical evidence regarding how angry communications affect interpersonal relations often does not distinguish between different kinds of anger. And when it does, it rarely makes discriminations based on the intentional objects of the anger (e.g., what kind of injustice the anger is directed at, whether it targets the injustice only or also its perpetrators, etc.).66 As illustrated above, the philosophical hypotheses I have outlined give us reason to expect that such differences may well influence angry speech’s impact on trust. Consequently, these hypotheses can and should guide the empirical task of ascertaining precisely when angry narratives can have a positive influence on trust, all things considered. Even with this qualification, objections remain. First, one might worry that the foregoing discussion does not show much about the trust-­building effectiveness 65  Another dimension concerns less the kind of injustice that the anger is directed at, and more whether the angry speech targets the perpetrators of the injustice. Galela’s testimony hardly refers to the perpetrators. For instance, he uses the passive tense when saying ‘I have been stripped of my dignity’. Would the trust-­building effects of angry testimony decline if Galela’s anger targeted the perpetrators (‘Look what you’ve done to me! You have stripped me of my dignity!’)? This depends on what particular trust-­building function we are examining. With respect to how delivering a piece of angry testimony helps the speaker trust the audience, it is unclear why it would make a difference. However, things seem more problematic when considering how angry testimony helps the listener trust the speaker. Being targeted by anger might make the listener feel defensive, not trusting. Some evidence seemingly supports this worry. Steinel et al. (2008) find that negotiators are more concessive when confronted with anger that targets their behaviour than when facing anger directed at their person. Nevertheless, their study remains inconclusive for our purposes. First, it focuses on what concessions were made, rather than directly measuring perceptions of the speaker’s trustworthiness. Second, it considers cases where the expressed anger solely targets the behaviour (‘This offer makes me really angry’), and where it solely targets the person (‘This person makes me really angry’). This leaves it unclear what to make of ‘mixed’ cases where the speaker’s anger explicitly targets the listener because of their behaviour (‘I’m angry that you made me this offer’). So it remains inconclusive for angry statements such as ‘You have stripped me of my dignity!’. 66  For studies that do not distinguish different kinds of anger, see, e.g., Tiedens (2001), Sinaceur and Tiedens (2006), Van Beest et al. (2008), and Wubben et al. (2009). Some studies do discriminate between different kinds of anger, but they typically draw distinctions that cross-­cut the ones I suggest are philosophically interesting. Indeed, they mainly distinguish between threatening and non-­ threatening expressions of anger, or direct and indirect expressions of anger. E.g., Guerrero (1994), Sereno et al. (1987). Neither of those distinctions concerns what the anger’s intentional object is. One exception is Steinel et al. (2008), who distinguish between person-­oriented and behaviour-­oriented anger. As note 65 discusses, however, their research has limitations for our purposes.

150  Democratic Speech in Divided Times of angry testimony itself. After all, my running example is taken from the South African TRC, whose discursive (or speech-­based) processes relied upon heavy institutional support. In particular, they were facilitated by moderators who, among other things, reminded participants of what was expected of them—sincerity, openness, listening, and so on. But even if institutional support is a necessary ingredient in trust-­building, this does not undermine the more fundamental point: that given this support, negative personal narratives offer important resources for building trust in deeply divided contexts, and (as the example of the TRC suggests) can seemingly do so in large-­scale settings. More generally, this suggests that, when paired with other kinds of measures, public speech can (a) contribute to trust-­building within fragmented and unjust societies, (b) in virtue of some of its non-­ideal properties (here, anger). Limited though this point may be, it still casts doubt on premise (3), the categorical claim that public discourse cannot contribute to trust-­building in non-­ideal conditions. Nevertheless, a second worry arises. In exploring anger’s trust-­building potential, I have focused on fitting instances of anger, where the angry testimony denounces genuine injustices. But what about unfitting anger? As discussed in previous chapters, many occurrences of angry speech in non-­ideal conditions express unfitting or misdirected anger. For some, publicly revealing their impassioned viewpoint means expressing rage at marginalized groups. So making space for angry storytelling also risks letting in public expressions of xenophobic or ra­cist rage. Allowing such hate speech might seem disastrous for trust-­building. The worry is not just that such speech cannot perform the positive trust-­building roles I have just listed. It is that, in addition, it risks having a highly negative impact on trust. Chakravarti, who generally defends angry narratives as a trust-­building tool, recognizes this worry and therefore stipulates that angry speakers should have the freedom to tell their story ‘as long as it does not become hate speech’.67 However, this response seems institutionally problematic. We are supposed to make space for personal storytelling, where individuals voice the grievances they harbour deep down. Simultaneously, deeply misguided grievances must be excluded. But often, in divided social settings, whether or not a particular grievance is misguided is precisely what is in dispute. Therefore, it seems difficult unambiguously to stipulate which views are ‘out’ before hearing people’s testimony. And given this uncertainty, excluding hateful views risks leading to self-­censorship that defeats the aim of having people say what is really troubling them. Perhaps there is an institutional way of avoiding this problem. But I do not want to rely excessively on this possibility. Moreover, Chapter  3 introduced

67  Chakravarti (2014, 156).

The Problem of Goodwill  151 defeasible reasons against forcibly or legally excluding public hate speech.68 Hence, considerations of consistency militate against taking this route. Finally, even if we could suppress hate speech without inhibiting the expression of fitting angry testimonies, this would constitute a significant concession to the problem of goodwill: it concedes that we need heavy-­handed sanitization of public discourse before it can be used for trust-­building. Instead, then, the final section will introduce reasons for doubting the view that failing to forcibly exclude hateful public narratives would necessarily undermine the process of building trust and goodwill.

4.3  Hate Speech and Trust Briefly put, the concern is that public hate speech seems likely to alienate ­listeners, particularly those whom it targets. Hate speech denies its targets’ basic social standing. Therefore, if A utters hate speech that targets B, this will presumably drastically diminish the trust and goodwill that B has towards A.  Now, failing to forcibly exclude hate speech increases opportunities for ­publicly expressing such speech. From these premises, one might conclude that failing to exclude hate speech is bound to have a destructive effect on trust and goodwill. But this argument is too quick. To see why, consider that, despite accepting its premises, Marcus Schulzke has argued for precisely the opposite conclusion: he suggests that refraining from excluding public hate speech is instrumental to rebuilding social trust. The crux of his argument is that, because allowing hate speech increases opportunities for publicly expressing hateful views, doing so helps us identify who does and does not hold hateful views.69 And this identifiability, in turn, is helpful when it comes to rebuilding trust. There are two reasons for this. Schulzke’s first suggestion is that the identifiability of hate speakers makes them and their beliefs easier targets of counterspeech: once hateful views are in the open, we can publicly oppose them and expose their baselessness.70 Although this might conceivably get hate speakers to abandon their abhorrent views, Schulzke concedes that this is unlikely. More significant is the potential effect of counterspeech on third parties. ‘Bystanders, especially those who might be susceptible to becoming prejudiced, can profit from seeing public debate [. . .] that demonstrates how little evidence there is to support hateful belief systems.’71 The idea, then, is that allowing public hate speech creates opportunities for counterspeakers to change the minds of those who might otherwise have been tempted 68  Chapter 3, Section 4. 70  Schulzke (2016, 236).

69  Schulzke (2016, 233). 71  Schulzke (2016, 236).

152  Democratic Speech in Divided Times by degrading views. This seems good for trust: repudiating hateful views makes one more capable of forming trusting relations with the targets of those views. This first trust-­based argument for allowing hate speech needs refinement. Schulzke assumes that because hateful views are unsupported by good reasons, counterspeech will generally succeed in persuading listeners to reject them. However, there are several problems with this assumption. In the first place, how persuasive a view appears is not wholly determined by the logic supporting it. The perceived authority of the person voicing it also matters. What is needed, then, is not just for private citizens to respond to hate speakers. Instead, as I argued in Chapter  3, authoritative public officials, and ideally state-­backed officials, must actively oppose public hate speech with their speech. Second, in the above quote, Schulzke seems to be recommending that we publicly debate specific hateful claims in order to negate and discredit them. But this seems quite similar to the idea of negative counterspeech, which Chapter 3 showed to be ill-­advised: ex­pli­ cit­ly negating and debating hateful propositions, recall, risks reinforcing their salience and thereby exacerbating their negative effect on the public assurance of dignity. To avoid this outcome, we should instead oppose hateful propositions via ‘positive’ counterspeech, which puts forward correct and compelling visions of the world that implicate (but do not reiterate) the wrongness of the hateful pro­ posi­tions at hand. Even with these refinements, there are limits to what counterspeech can achieve by way of persuasion.72 Some hateful citizens will simply be unwilling to change their minds. Worse, feeling publicly challenged might further entrench their commitment to hateful views. So even if Schulzke is right that the identifiability of hateful doctrines creates new opportunities for steering some citizens away from hateful views—and, in doing so, helps to improve social trust—it nonetheless seems to have some countervailing bad effects on trust. In particular, the speech of hateful citizens who do not abandon their views will continue to alienate targets of hate speech. This is where the second and more important trust-­related function of making hate speakers identifiable kicks in. To appreciate this function, consider that even if being on the receiving end of hate speech makes trusting the hate speaker impossible, it does not necessarily follow that things would be better with respect to trust if hateful citizens were forced not to speak their minds. There are two possibilities. Either hate speech bans make the existence of hate speakers wholly invisible, or they do not. If hate speech bans entirely conceal the existence of hateful citizens, they put targets of hateful views at risk of engaging in unwarranted relations of trust—that 72  This is compatible with Chapter 3’s argument. There, my emphasis was less on how state counterspeech can persuade, and more on its ability to block hate speech’s assault on the dignity of vul­ner­ able groups. See Chapter 3, Section 5.2.

The Problem of Goodwill  153 is, of trusting citizens who are untrustworthy. Suppose that B does not know that A is a hateful citizen, and, out of ignorance, relies on A. The trust B places in A seems unwarranted: A’s hateful attitudes towards B seem inherently in tension with a commitment to behaving trustworthily towards B. Now, as Russell Hardin explains, engaging in such an unwarranted relation of trust is undesirable: because unwarranted trust is by definition likely to be betrayed, it is both unstable and dangerous for the truster.73 The second (and more empirically plausible) possibility is that hate speech bans do not make hateful citizens wholly invisible. Even if hateful citizens do not publicly express hateful views, it is still known that some people hold hateful views. But what becomes less well known is who exactly holds these views. This lack of identifiability is again problematic. Insofar as they do not know who holds hateful views, targets of hateful views must either run the risk of trusting untrustworthy hateful citizens, or refrain from trusting anyone from groups that are known to contain some hateful citizens.74 To illustrate, consider Zimasile Bota’s testimony to the South African TRC. In her testimony, Bota asserted: One of my requests is that the perpetrators, the Government must bring them forward. We cannot work together with the police. We know that the police are protecting us, but we cannot work together with the police who have blood on their hands. [. . .] My request is that the police must come forward.75

Bota is claiming that because many police officers were complicit in racist apartheid practices, they cannot be trusted. Until those who were complicit are identified, working with the police seems impossible, as it is unclear which members are trustworthy and which are not. Restoring trust, she suggests, requires making it more visible who is and who is not trustworthy. This is precisely what legally permitting public hate speech, while advocating (as I have done) robust counterspeech against it, would in principle achieve.76 Allowing A to voice his hateful views towards B, while having non-­hateful third parties (C) verbally counter these views, helps B see that while she cannot trust A, she can trust C. This seems intuitively preferable to the two previous scenarios: (1) the situation where B does not know that anyone has hateful views, and therefore misguidedly trusts A; and (2) the situation where B only knows that one of A and C has degrading views, and is therefore unsure which of the two she can 73  Hardin (2002, 29–32). This last argument is compatible with my claim, in Section 4.2, that pla­ cing trust produces a new incentive to be trustworthy. The difference is that, here, B places trust out of ignorance: the description of A that motivates B to extend trust is one that A does not actually satisfy. Consequently, A cannot justifiably think that B is signalling esteem for A. 74  Schulzke (2016, 230–6). 75  Cited in Chakravarti (2014, 69). 76  Schulzke (2016, 230–5).

154  Democratic Speech in Divided Times safely trust. Indeed, making hateful citizens visible helps targets of hateful views trust trustworthy citizens and only trustworthy citizens. In sum, the objection that not excluding hate speech is bound to undermine trust is too hasty. By making hateful citizens more visible, the practice of allowing hate speech while authoritatively denouncing it performs two trust-­building functions: it creates new opportunities for persuading hateful citizens; and when persuasion fails (as will often happen), it at least makes it clearer who can and cannot safely be trusted. This second point concedes that hate speakers and their targets will not trust each other. Even so, I have suggested that the fundamental trust-­related problem here is the existence of hateful citizens. Their un­trust­worthi­ ness with respect to their targets is what destabilizes trusting relations, rather than the policy of allowing those hateful citizens to air their views. Thus, the more crucial point against premise (3) remains: given that hateful citizens exist, the policy of allowing public hate speech (while authoritatively and positively opposing it) arguably seems to be part of the solution, rather than the source of the problem.

5. Conclusion According to the problem of goodwill, fruitful public discussion between members of diverse groups presupposes unrealistic levels of mutual goodwill and (as a presupposition thereof) trust. Moreover, it is said that public discussion cannot itself contribute to generating trust and the attending goodwill in non-­ideal conditions. The present chapter has articulated reasons for pushing back against this challenge. To begin, I showed that in light of systemic character of public discourse, discursive processes presuppose less goodwill than one might think: highly demanding discursive norms need not apply to most arenas of public speech; and when the norms that do apply to most people—for example, norms of sincere engagement—remain too demanding, the networked interaction between different spheres helps circumvent the communicative barriers raised by distrust. These qualifications notwithstanding, the problem of goodwill remains problematic. Accordingly, most of my argument tackled another claim: that the public discourse that characterizes non-­ideal contexts cannot play a part in generating greater trust and goodwill. Specifically, I explored respects in which three features of non-­ideal public discourse—hypocrisy, anger, and the occurrence of hate speech—offer important resources for rebuilding trust and the attending goodwill. Even if we start from pessimistic assumptions regarding existing goodwill and  trust, the practice of hypocritically pretending to be public-­spirited and ­sincere may, as Elster suggests, make participants more accountable to their

The Problem of Goodwill  155 interlocutors. Public hypocrisy induces participants to listen, and eventually to answer challenges and change their stated positions. Given a disposition to listen, other characteristic features of non-­ideal public discourse have promising trust-­building properties. Drawing on philosophical approaches to trust, anger, and transitional justice, I theorized several trust-­building functions of personal narratives charged with anger. Telling one’s anger-­infused story can have a humanizing influence on the listener’s perception of the speaker, signal trust in a way that invites the listener’s trustworthiness, help the speaker feel acknowledged, and provide information that is instrumental to rebuilding social bonds. But one problem with making space for angry testimony in divided contexts is that testimonies expressing unfitting anger might conceivably veer into hate speech. Nevertheless, given a situation where some individuals have hateful views, it is unclear that allowing public hate speech while authoritatively denouncing it makes matters worse for trust. Instead, as we have seen, there are reasons to think that doing so improves social trust by making it clearer who can and cannot be trusted. Notice, finally, how this response to the problem of goodwill accords with the normative recommendations of previous chapters. Emotionally charged narratives, including angry testimony, are forms of speech I defended in Chapters  1 and  2. And Chapter  3 recommended legally allowing public hate speech while having authoritative public figures denounce it. A similar observation holds with public hypocrisy. The occurrence of fruitful hypocrisy results, in part, from the expectation that participants should appeal to considerations that are public or shared—a norm defended in Chapter 1. I originally defended these norms on the grounds that they play a key role in identifying social problems—such as group-­based harms and injustices—and motivating responses to these problems. But the fact that these very same norms offer prima facie promising resources for generating trust and goodwill in divided contexts should bolster this initial defence. It suggests that the normative account of inclusive democratic speech I have offered is well suited to withstanding the shortage of goodwill that characterizes divided political settings.

6

The Problem of Political Ignorance 1. Introduction 1.1  Political Ignorance Survey after survey suggests that citizens of contemporary democracies tend to know relatively little about political affairs.1 For one thing, voters are often uninformed about which political actors support which policies, and what they have done about it. In 2010, for example, only a third of Americans knew that the controversial bailout bill (TARP) had been enacted under George W. Bush rather than Barack Obama.2 Furthermore, voters commonly do not know how the political decision-­making system works. Consider, for instance, that in 2006, most Americans were ignorant of the respective functions of the three branches of government.3 Finally, voters tend to be misinformed about facts that are relevant to assessing the costs and benefits of policies. Notably, they typically vastly overestimate the costs of foreign aid, and underestimate defence spending.4 The fact ‘that the public is overwhelmingly ignorant when it comes to politics,’ Jeffrey Friedman concludes, ‘is one of the strongest findings that have been produced by any social science—possibly the strongest’.5 There are several reasons for this political ignorance. One is that gathering reliable and balanced political information can be challenging. This is particularly true in contemporary divided democracies where (as we saw in Chapter 4) public discourse is rife with political misinformation, and where (as we saw in Chapter 5) individuals with different perspectives may be reluctant to engage with one another. At the same time, in large-­scale democracies, the probability of making a difference with one’s vote is minimal. This suggests that the potential payoff of being an informed voter is low. If being knowledgeable about politics comes at a high cost and has a low reward, it might seem rational to remain politically ignorant.6 This political ignorance, many have argued, constitutes a significant threat to the value of democratic public discourse.7 Even if people are willing to engage 1  For an overview of this evidence, see Somin (2016, ch. 1). 2  Somin (2016, 31). 3  Somin (2016, 20). 4  Somin (2016, 18). 5  Friedman (1998, 397). 6  E.g., Somin (1998, 435–8). 7  For this objection, see, e.g., Brennan (2016, chs. 1, 2, 4), Somin (2016, chs. 1–2), Achen and Bartels (2016, chs. 1–7), Guerrero (2014, 140), Anderson (1998, 490–1), Posner (2003, 163).

Democratic Speech in Divided Times. Maxime Lepoutre, Oxford University Press (2021). © Maxime Lepoutre. DOI: 10.1093/oso/9780198869757.003.0007

The Problem of Political Ignorance  157 with each other, one might worry that they remain insufficiently knowledgeable to do so competently. After all, if people do not how the political system works, what policies different politicians support, or what the likely outcomes of those policies might be, they may not be able to distinguish sound public reasoning from unsound public reasoning, let alone engage in sound public reasoning of their own. If, for example, I mistakenly believe that Obama passed TARP, I might be swayed by the fallacious claim that Obama should be voted out for office for bailing out the banks. Or, if I vastly overestimate how much the state spends on foreign aid, I am more likely to be persuaded by spurious arguments suggesting that foreign aid is why the state cannot afford more generous domestic welfare policies. In sum, the general worry is that if voters are deeply ignorant about politics, then engaging in public debate may not always help them identify what policies or politicians would best serve their interests and concerns. Often, it might have the opposite effect, inducing voters to form further false beliefs about which policies and politicians they should prefer. And, if engaging in public debate risks instilling false beliefs about political matters, it may not help ignorant voters hold policymaking accountable. Hence, political ignorance risks undermining both the epistemic function and the accountability (or anti-­domination) function of public discourse.

1.2  Political Ignorance and Group Cognition Democratic theorists have a ready response to the problem of political ignorance. This response holds that, even if people are generally uninformed about politics, they can readily remedy this problem by taking ‘information shortcuts’. For example, people can turn to more informed acquaintances, opinion leaders, experts, and pollsters for testimony or advice regarding which politicians and policies they should support.8 From this perspective, the fact that many people are unable to answer political questions when surveyed is of little concern, because they can find the answers they need, when they need them, by seeking out and listening to reliable sources. The appeal to information shortcuts aims to salvage the value of democratic public discourse in two ways. First, it suggests that people can readily circumvent their political ignorance. In other words, even if a lack of information would prevent us from competently exchanging reasons with others in public debate, this is not a deep or intractable problem: we can readily use information shortcuts 8  For this response, see, e.g., Page and Shapiro (1992), Lupia and McCubbins (1998), Christiano (2012), Lupia (2016, ch. 5), Goodin and Spiekermann (2018, ch. 12).

158  Democratic Speech in Divided Times to get the information we need to then engage fruitfully with others. Second, public discourse itself plays a role in circumventing this political ignorance. Information shortcuts often operate by means of public speech. As mentioned above, the public testimony or advice given by experts and opinion leaders are paradigmatic examples of information shortcuts. So, not only can people circumvent the political ignorance that threatens to weaken the value of democratic public discourse, but, in addition, public discourse can actually play a role in achieving this. There is, however, a problem with this common response. The response assumes that when people seek out information shortcuts in the public sphere, their aim is to form accurate or reliable beliefs. But there is ample evidence suggesting that, when it comes to politics, people’s social group identities strongly condition what information and whose advice they end up believing. Indeed, the way people go about forming political judgments is deeply shaped by group affiliations such as their race, gender, ethnicity, or party. This, critics argue, indicates that people are motivated not so much by the desire to acquire accurate political information, as by the desire to satisfy their emotionally charged group loyalties. In politics, it would appear, people seek out and process information less like scientists—who seek the truth—and more like sports fans—who embrace whatever views will show their team in the best light. In turn, if people acquire information in a way that is unrelated to the truth, their resulting political beliefs will probably be epistemically dubious, ill-­ suited to holding policymaking accountable, and (relatedly) unlikely to help them engage in fruitful public discussion.9 Thus, the most familiar way of countering the problem of political ignorance has increasingly been criticized for being out of touch with human psychology. Democratic theorists suggest that politically uninformed people could form reliable beliefs about politics by deferring to cues or testimony from appropriate sources. But the phenomenon of group cognition—the pervasive influence of group loyalties on people’s judgments about politics—suggests, in the eyes of many, that these hopes are misplaced. For Jason Brennan, the democratic ideal, with its emphasis on promoting political knowledge and accountability via open-­ minded public discussion, disregards ‘the overwhelming consensus in political psychology [. . .] that most citizens process information in deeply biased, partisan, motivated ways’.10 Similarly, Christopher Achen and Larry Bartels, who have most influentially articulated this worry about group cognition, conclude that

9  See, in particular, Kahan (2015), Achen and Bartels (2016, chs. 8–10), Somin (2016, ch. 3), Brennan (2016, ch. 2), and Bagg (2018a). 10  Brennan (2016, 12).

The Problem of Political Ignorance  159 normative democratic theory is ‘marked by an addiction to romantic theories’ of human nature.11 By way of response to the problem of political ignorance, the present chapter will therefore closely scrutinize the phenomenon of group cognition. I will argue that, in fact, group cognition need not be a problem for democratic public discourse. This is because, contrary to what critics suggest, it need not constitute an impediment to political learning. On the contrary: under the right conditions, group cognition seems an epistemically valuable practice for otherwise uninformed people to go about forming beliefs about politics within the discursive system. This, to be clear, remains a qualified response to the problem at hand. I am not arguing that group cognition is never a problem for public discourse, nor even that it is not currently a problem in contemporary democracies. Instead, as I will emphasize throughout, specific background social conditions must obtain for group cognition to play a positive function. When they do not, group cognition can impair, rather than advance, the value of inclusive public speech. Yet, qualified though it may be, this conclusion matters. First, it gives us a clearer diagnosis of the problem that political ignorance poses for public discourse. It locates the problem, less in inalterable facts about the importance of group identity to human beings, and more in social conditions that are ultimately contingent. Accordingly, the problem is not that human psychology is inconsistent with the possibility of valuable democratic public speech. This diagnosis, in turn, is action-­guiding. By clarifying the conditions under which group cognition can be a key ingredient in democracy, it highlights tractable strategies for reconciling group cognition and the value of public discourse in real-­world conditions. My argument will proceed as follows. After outlining why the influence of social group membership on political judgment may seem to be in tension with the value of democratic public discourse (Section 2), I contend that this tension can be dissolved. The reason, at bottom, is that group membership defines epistemically significant social perspectives. As a result, the testimony, advice, and positions of members of one’s group constitute a useful heuristic for determining what one should think about political issues (Section 3). I then consider and rebut two difficulties with this reconciliation. First, in a particular domain where group cognition occurs—the domain of politically relevant technical facts—group perspectives do not seem epistemically relevant (Section 4). Second, many social perspectives, because they contain blindspots, risk leading to political dogmatism (Section 5). Neither of these difficulties ends up being decisive: even though ‘technical’ and ‘dogmatic’ forms of group cognition 11  Achen and Bartels (2016, 20).

160  Democratic Speech in Divided Times do sometimes impair the value of democratic public discourse, I will show that this need not be the case. Both, under the right conditions, can instead play ­valuable epistemic functions.

2.  The Problem of Group Cognition 2.1  Group Cognition People form political judgments—judgments that bear on what politicians and policies they should support—in a way that is deeply influenced by their social group attachments. In recent years, this observation has most influentially been articulated by Achen and Bartels, who assert that ‘for most people, most of the time, party and group loyalties are the primary drivers of vote choice’.12 Typically, social group attachments shape people’s political judgments indirectly: they first shape people’s party affiliation, which later influences their judgments concerning policies and politicians.13 For instance, Achen and Bartels find that, in the New Deal era, the main reason why Boston Jews joined the Democratic Party had little to do with domestic or foreign policy. Instead, it was fundamentally ethnic: it was a response to the fact that Jews had begun acquiring greater status within the Democratic Party.14 Similarly, they argue that ‘racial and regional identity was the more important factor’ underpinning Southern realignment, whereby Southern whites massively migrated to the Republican Party from the 1950s onwards. ‘Southern whites viewed the parties primarily as collections of social groups, not as packages of policy positions’: in light of increasing black American mobilization within the Democratic Party, Southern whites left because they no longer perceived it as the party for ‘people like them’. Only later did their policy views start to conform to the Republican Party’s.15 A final illustrative example concerns the Tea Party. Drawing on extensive interview data, Katherine Cramer suggests that Tea Party support is driven, less by antecedent policy preferences—indeed, Tea Party policies sometimes seem paradoxically opposed to the policies favoured by rural white Americans—and more by Tea Party politicians’ ability to present themselves as identifying with, and coming from, rural white America.16 That being said, the influence of social groups on political judgment is not always mediated by partisanship, and it can sometimes run against existing party affiliations. For example, Achen and Bartels observe that gender identity had a 12  Achen and Bartels (2016, 272; see also 4, 218). See also Somin (2016, ch. 3), Kahan (2015), Iyengar et al. (2012), Huddy et al. (2015). 13  Achen and Bartels (2016, 266–8, 299, 307). 14  Achen and Bartels (2016, 40). 15  Achen and Bartels (2016, 253; see also 246–64). 16  Cramer (2016, 184–203).

The Problem of Political Ignorance  161 deep influence on Americans’ views on abortion in the 1980s, and specifically on whether those views conformed to party lines. While men altered their views on abortion to adhere to party lines, women did not. Instead, women predominantly ‘gravitate[d] to the party sharing their view about abortion’.17 The main point is this: either via their influence on party affiliation, or more directly, social group attachments profoundly shape people’s views on political matters. Why might this be problematic?

2.2  The Epistemic Irrelevance of Groups The objection, briefly put, is that social group attachments are not an epistemically reliable indicator of which policies, parties, or political representatives might be desirable. In particular, the fact that someone belongs to our social group does not give us a reason to believe that their judgments about parties, policies, or other political matters are trustworthy. So, shared group membership does not give us an epistemic reason to defer to someone’s testimony or advice about political matters. Nor, relatedly, does it give us a reason to believe that they would make a good political representative. What drives this objection is an account of what social groups are. Often, critics of group cognition characterize social groups extensionally, by offering examples of such groups.18 As shown above, the examples that come up most often are race, gender, ethnicity, regional attachments, and party affiliation. However, this extensional characterization only pushes back the key question: what is it about these groups that makes them an inadequate epistemic indicator of whose political judgment to trust and of what policies to support? Although Achen and Bartels do not commit to a specific and systematic intensional account of social groups—that is, an account of what it means for a set of people to constitute a social group—they do offer some indications that shed light on the above objection. First, they observe that social group memberships tend to originate in arbitrary ways. If group memberships resulted from reasoned choices that were based on one’s prior policy preferences or ideological commitments, then their influence on judgments about politics might seem epistemically respectable. Indeed, if this were the case, the fact that someone belongs to one’s group would be a sign that they concur with judgments that one antecedently takes to be sound. This, in turn, would constitute a pro tanto reason to believe that their judgment is trustworthy. However, Achen and Bartels insist that this is not the case. Social group memberships such as race or ethnicity are unchosen. And even insofar as some social group memberships can be chosen, 17  Achen and Bartels (2016, 17). 18  Achen and Bartels (2016, chs. 8–10; Kahan (2015, 11–12).

162  Democratic Speech in Divided Times they tend to be chosen in ways that are in an important sense arbitrary: specifically, people’s grounds for joining groups are generally not their prior reasoned judgments about policy or ideology.19 Because of their arbitrary origin, then, group attachments do not give us epistemic reasons in the way that groups formed on the basis of reasoned policy or ideology judgments would. Second, not only do group attachments originate in arbitrary ways, but they are also emotional in nature. Alongside other critics of group cognition, Achen and Bartels repeatedly emphasize ‘the powerful role of emotion rather than reason in directing group activity’.20 The idea is that group attachments invite people to think about politics in a way that is charged with emotion, rather than based on rational thought or logic. And since rational thought and logic are the paradigmatic epistemically reliable ways of forming judgments, this suggests that group cognition is not epistemically reliable. It aims not at producing accurate beliefs about political matters, but rather at satisfying the emotional dispositions that define one’s group identity. One might respond that this account, however plausible it may be for racial, gender, or ethnic groups, is implausible as a characterization of another kind of judgment-­shaping group: party affiliation. According to Regina Rini, political parties reflect people’s moral and political values, or ideology. If people choose their party based on reasoned ideological judgments, and if party affiliation later shapes their beliefs about political matters (via, for instance, the testimony or advice of party leaders), then one of the main forms of group cognition—namely, party-­based cognition—might be epistemically useful after all.21 However, this line of response neglects the empirical reality of how party affiliations actually form. Political elites do often choose parties based on prior reasoned judgments about ideology and policy. But Achen and Bartels are adamant that most people do not: ideological commitments are ‘more often an effect of partisanship than its cause’.22 As we have already seen in Section  2.1, partisan affiliation instead habitually results from people’s prior social group attachments, such as their race, gender, ethnicity, or regional identity. Thus, ‘partisanship is both a form of social identity and, in significant part, a product of social identity’.23 Hence, party affiliation too seems vulnerable to the above concerns. If partisanship derives from group loyalties that arise arbitrarily, then its origin also seems, at bottom, to be arbitrary. Moreover, if the social group attachments that undergird party loyalty are emotional in nature, then party loyalty too is presumably emotional in nature. This is indeed how Brennan characterizes 19  Achen and Bartels (2016, 220–3). 20  Achen and Bartels (2016, 215); see also Achen and Bartels (2016, 228, 255, 293), Kahan (2015, 2–3), Huddy et al. (2015, 3–4), Brennan (2016, 12), and Somin (2016, 101–2). 21  Rini (2017, 51). 22  Achen and Bartels (2016, 234; see also 233, 268). 23  Achen and Bartels (2016, 234).

The Problem of Political Ignorance  163 partisanship: people support political parties in an impassioned way, he suggests, much as hooligans support sports teams.24 So, like other kinds of social group attachments, party affiliation appears to be an epistemically inadequate basis for forming judgments about political matters.

2.3  The Depth of the Problem In sum, the problem of group cognition runs as follows. People use social group membership as a basis for forming judgments relating to political matters. But social group membership is not an epistemically reliable basis for forming political judgments. So, people form political judgments by relying on cues and testimony that are epistemically unreliable. This, in turn, seems troubling for the ideal of democracy and democratic public discourse. As we saw earlier, the standard response to concerns about political ignorance is to suggest that politically uninformed people can rely on information shortcuts, such as the testimony or cues provided by more informed sources. In this way, even when people start out knowing little about politics, democratic public discourse can give them the epistemic resources they need to reliably identify which policies or politicians would serve their interests, and to subsequently engage in fruitful discussions of their own on these topics. Therefore, if people are capable of deferring to reliable cues and testimony, their initial political ignorance need not undermine the epistemic and accountability-­ promoting functions of inclusive public discourse. But if, instead, politically uninformed people base their judgments on epistemically arbitrary cues and testimony, then this solution is no longer available.25 Prior to criticizing this argument, it is worth emphasizing why, if successful, it would constitute a distinctively powerful challenge to democratic public speech. First, not only is group cognition widespread, but it also seems to be deeply rooted in human psychology. As Achen and Bartels put it, ‘human nature makes group attachments powerful forces in political thinking’.26 This matters because, if group cognition is part of human nature, and if this way of forming political judgments threatens the epistemic and accountability-­promoting functions of democratic public discourse, then it is unclear whether there is any way—short of  altering human nature itself—of salvaging the value of democratic public discourse. Second, the problem of group cognition is unlikely to be solved by appealing to aggregative mechanisms. A popular response to the fact that individual voters sometimes make poor political judgments points to aggregative mechanisms: 24  Brennan (2016, ch. 2). See also Huddy et al. (2015, 3). 25  Achen and Bartels (2016, 176). 26  Achen and Bartels (2016, 232, 306), emphasis added.

164  Democratic Speech in Divided Times aggregating many individual judgments through majority rule can, under the right conditions, yield decisions that are far more reliable than any individual citizen is. However, the most influential such argument, which derives from the Condorcet Jury Theorem, requires that people vote independently of one another. This condition is clearly violated if people’s views about political matters are deeply shaped by their group attachments. Thus, if sound, the problem of group cognition is extremely forceful. One might challenge its empirical premise—that social group identities deeply shape political judgments.27 However, I will take a different route. I will argue that group cognition is typically not an epistemically misguided way of forming judgments about political matters. So, using group identity as a heuristic when deciding which information shortcuts to rely on is, often, an effective way to circumvent or indeed overcome one’s political ignorance.

3.  The Epistemic Value of Group Cognition 3.1  Groups and Social Perspectives In suggesting that group cognition is epistemically misguided (in ways that, for the reasons cited previously, risk jeopardizing the value of democratic public discourse) critics overlook an influential tradition of democratic theory, which gives social groups a central role in the democratic process. Specifically, according to some democratic theorists, there are strong reasons for voters to be represented by members of their own social groups. Jane Mansbridge, for instance, replies ‘Yes’ to the question ‘Should blacks represent blacks and women represent women?’.28 Similarly, both Iris Marion Young and Melissa Williams advocate ‘group representation’ as a way of ensuring that distinctive perspectives and marginalized voices are heard in public debate.29 Public discourse, these theorists suggest, is significantly enhanced when, in formal deliberative arenas, people are represented by members of their own social groups. This tradition, I will argue, reveals that group cognition does in fact have epistemic value, in virtue of which it is an effective way of circumventing or overcoming political ignorance. Our question, then, is this: in what respects might social group attachments be epistemically relevant to forming judgments about political matters? Why, in particular, does the fact that someone comes from one’s social group give one reason to believe that their judgment is reliable or trustworthy?

27  See, e.g., Huddy (2018). 28  Mansbridge (1999b). 29  Young (2000, ch. 4); Williams (2000, 124).

The Problem of Political Ignorance  165 By articulating the notion of a social perspective, Young’s discussion of group representation helps address these questions. Part of what it means to belong to a social group defined along axes such as race, gender, class, or ethnicity, according to Young, is that one occupies a distinctive position in the social structure. This, in turn, means that one is subjected to a distinctive set of social constraints and enablements by the laws, norms, and physical infrastructure that constitute the social context. So, a social group differs from a random set of people in that its members ‘experience similar constraints or enablements’.30 The social group ‘women’, on this view, is partly defined by exposure to a distinctive set of shared constraints, such as the expectation that one shoulder the lion’s share of childcare, and disproportionate subjection to harassment in the workplace.31 Likewise, as Ann Cudd explains, to be a middle-­class black man is, in part, persistently to encounter different constraints and enablements than a middle-­class white man might: ‘the middle-­class black is likely to be followed by store workers, he is likely to be harassed by police in white neighbourhoods or if seen driving an expensive car, women cross the street when they seem him coming down the sidewalk’, and so on.32 Because they experience group-­specific constraints and enablements, members of a social group have distinctive knowledge that members of other groups may lack. Young calls ‘social perspective’ the particular point of view or ‘way of looking at things’ which results from having this group-­specific knowledge.33 A group’s social perspective involves, first, a descriptive component: knowledge of particular facts about the workings of society that one gleans from one’s group-­ specific experiences of constraint and enablement. But social perspectives also involve a normative component, which falls out of the descriptive component. Because certain societal constraints are more salient to members of a particular social group, they are more likely to be objects of concern to these members. Put differently, these constraints are more likely to be in the foreground of their thinking, as problems to be addressed.34 For example, because micro-­aggressions are typically more visible to a middle-­class black man than to a middle-­class white man, they are more likely to be at the forefront of the former’s concerns. Similarly, in virtue of having greater personal experience of sexual harassment, women might not only know more about this issue, but also take it more seriously as a problem to be tackled. The upshot, for Young, is that it is epistemically reasonable to give preference to representatives who come from one’s own social group. The fact that someone comes from one’s group is an indicator that they share a relevant social perspective: they have knowledge of social constraints that one experiences, and they are more likely than others to share one’s normative concerns relating to 30  Young (2000, 95–100). See also Cudd (2007, 44). 32  Cudd (2007, 44–5). 33  Young (2000, 136–7).

31  Young (2011, ch. 2). 34  Young (2000, 102, 114–17, 136).

166  Democratic Speech in Divided Times those constraints. Consequently, shared group membership gives one a pro tanto epistemic reason to think that their political judgment is trustworthy—and, by extension, that they would make a good representative. Indeed, one has reason to believe that they will reason about policy issues in light of descriptive and normative considerations one deems important; that they will raise these descriptive and normative considerations when engaging in public debate; and that they will give political advice or testimony that is informed by these descriptive and normative considerations. For example, Young notes, the fact that a given candidate is a woman is a reason to believe that she is more likely to appreciate the pervasiveness and seriousness of sexual harassment.35 Accordingly, Mansbridge observes, ‘it is not surprising [. . .] that women legislators have usually been the ones to bring these issues [of harassment] to the legislative table’.36 The point, however, is not restricted to judgments about political representatives. If shared group membership is a pro tanto reason to trust someone’s political judgment, then the party affiliation of members of one’s group constitutes a valuable heuristic for deciding which party to support. In other words, the fact that members of one’s social group disproportionately belong to or support party X might well constitute a pro tanto epistemic reason to support X. What this fact signals, in short, is that X, together with the positions it advocates, may be distinctively attuned to group-­specific constraints that one judges to be politically important. In this light, Achen and Bartels’ observation that social group attachments drive party affiliation may not be so troubling after all. When Boston Jews saw that Jews were well represented in the Democratic Party, they learned that members of the Democratic Party shared many of their experiences of social constraint, and were more likely to attend to associated concerns. Likewise with the Tea Party. When its leaders assert that they come from rural communities, they thereby signal an affinity with the perspective of rural inhabitants—what Cramer famously calls ‘rural consciousness’. This, Cramer suggests, might involve sharing experiences of rising rural joblessness, of the physical toll that rural jobs take, or of having one’s lifestyle derided or vilified.37 This avowed affinity with the social constraints that constitute ‘rural consciousness’ gives rural inhabitants a reason to trust them. Social group membership can also more directly inform judgments about policy. This is because the descriptive and normative concerns contained in group-­ specific social perspectives can sometimes bear directly on policy matters. This makes sense of another case Achen and Bartels worry about: the influence of gender on judgments about abortion. If women have relevant experiences that men lack (e.g., experiences of childbearing) and if the proposed legislation would remove constraints that men do not face (e.g., constraints on what women do with their bodies) then it is neither surprising nor epistemically troubling that 35  Young (2000, 140).

36  Mansbridge (1999b, 647).

37  Cramer (2016, 66–84).

The Problem of Political Ignorance  167 women formed their judgments about abortion differently than men. They had relevant experiential knowledge, which men lacked, that bore directly on the issue. Seen in this light, group cognition seems an epistemically promising way for politically uninformed voters—the majority of voters—to circumvent their ignorance. For one thing, using social group membership as a heuristic allows voters to identify and select a party or representative that is sensitive to important group-­specific constraints. In turn, these representatives or party elites can bring their greater political knowledge to bear on these group-­specific issues, both when advocating policies in formal public debates, and when providing advice, feedback, or testimony to their constituents. In some cases (such as the case of abortion) voters can even directly rely on the experiences involved in their group identities to get the information they need effectively to assess policies. This epistemic function of group cognition contributes to salvaging the value of democratic public discourse in the two ways advertised in the introduction. In the first place, it helps voters access the information that they will need competently to engage in public discussions of their own. And, often, it achieves this effect via public discourse, by helping voters identify sources of advice or testimony that they have good reason to trust. There is, however, a possible concern here. The political issues that we have been mentioning so far (e.g., abortion, sexual harassment, racial discrimination, rural vilification) are issues to which the normative and descriptive considerations contained in social group perspectives are clearly relevant. For instance, the distinctive experiences involved in being a member of a racial minority are often directly relevant to understanding and assessing racial discrimination. So it makes sense, when looking for information shortcuts relating to this issue, to use race as a heuristic. The concern is that, for some politically important questions, the experiences contained in group perspectives might seem significantly less relevant. In particular, highly technical or scientific questions (e.g., what are the expected consequences of complex environmental processes; what are the expected health risks of a given vaccine) often have important implications for policy. But day-to-­day experiences of social constraint or opportunity are, on the face of it, not relevant to answering these highly technical questions—nor, relatedly, to determining who is a reliable source on these matters. So, if people’s group identities nevertheless influence their beliefs on these technical matters, this might seem prima facie epistemically troubling. In Section 4, I will consider an objection which holds that this is in fact the case: sometimes, the objection claims, group cognition affects beliefs about politically relevant technical matters; but group cognition is not epistemically appropriate in this domain. Before we turn to address this concern, though, a final clarification relating to the idea of a social group is needed.

168  Democratic Speech in Divided Times

3.2  A Different Conception of Groups? My defence of the epistemic value of group cognition appeals to Young’s account of social groups. But one might wonder if the kinds of social groups Young discusses are the same as the kinds of social groups involved in group cognition. If not, my argument fails: even if groups in Young’s sense are epistemically important, people form their beliefs about political matters on the basis of a different kind of group membership. In fact, there are good reasons to think that Young’s social groups are broadly equivalent to the social groups involved in group cognition. To begin, Young’s conception of social groups is extensionally similar to Achen and Bartels’. Indeed, the core examples of social groups she gives—for example, race, gender, class, or region—are also examples of groups that shape their members’ political judgment.38 Moreover, Young’s conception of social groups seems compatible with Achen and Bartels’ intensional characterization of social groups. The first intensional characteristic, recall, is that group memberships typically have an arbitrary origin: they are habitually not chosen; and even when they are, this is generally not on the basis of prior reasoned judgments about ideology or policy. This fits easily with Young’s account of social groups. To belong to a social group, for Young, is to be exposed to a particular set of social constraints and enablements. Crucially, Young emphasizes that what particular constraints and enablements one is subjected to is something ‘over which [one] ha[s] little control’,39 and often a mere ‘accident of birth’.40 In this light, we can now appreciate that the problem of group cognition commits the genetic fallacy. From the fact that the origin of social group attachments is arbitrary—that is, not chosen on the basis of sound political judgments—they infer that their nature is arbitrary from the perspective of sound political judgments. But Young’s account shows that this is incorrect. Regardless of how randomly one ended up in a particular social position, being in that position remains epistemically relevant to political judgment. This, once more, is because one’s social position gives one privileged epistemic access to politically relevant constraints and enablements.

38  The extensional equivalence might seem imperfect. Achen and Bartels (2016, ch. 8) mention ethnicity and religion (e.g., Judaism, Catholicism) as examples of judgment-­shaping group attachments. These might seem to be cultural groups—that is, groups united by shared values—rather than social groups in Young’s sense—that is, groups united by shared experiences of social constraint and enablement. But this apparent mismatch is unproblematic. As Young (2000, 98) observes, cultural groups are often also social groups. When a specific cultural group, because it is identified as such, is subjected to common social constraints, it thereby becomes a social group. This might happen, say, when Jews or Catholics experience group-­specific acts of discrimination. 39  Young (2000, 92). 40  Young (2000, 96).

The Problem of Political Ignorance  169 The second intensional characteristic of social groups that drive political j­udgment—that they are emotionally charged—might seem more problematic. ‘Emotional attachments’, according to Achen and Bartels, ‘transcend thinking’.41 But social group attachments in Young’s sense do not transcend thinking. On the contrary, Young emphasizes how group attachments provide descriptive knowledge of social constraints, which in turn enriches normative reasoning about politics. So, one might conclude, social groups in Young’s sense cannot be emotionally charged. This worry depends on the assumption that emotions are divorced from rational thought. However, we saw in Chapter 2 that this is a grave misconception: emotions in fact have an important cognitive dimension, in virtue of which they contribute fruitfully to reasoning. In particular, recall that emotions are sources of salience: they ‘establish focus on certain aspects of a situation, they act as “spotlights” ’, which ‘pu[t] some properties of a situation into the foreground’.42 To be more specific, different emotions put a spotlight on different kinds of properties. Fear highlights features of our environment that are potentially dangerous; anger focuses our attention on actual or potential sources of injustice; grief makes loss salient to us; and so on. Because they play this varied salience role, emotions have epistemic value: they help us navigate complex environments by drawing our attention to important facts that we may otherwise overlook. In doing so, they enrich the information on the basis of which we reason. Once emotions are understood to be sources of salience, the idea that group attachments are emotionally charged seems intimately connected to Young’s idea that groups are characterized by a distinctive social perspective. As we have seen, a group’s social perspective is a way of looking at things that is informed by that group’s specific experiences of social constraint. It typically involves having these constraints in the foreground of one’s thinking, as problems to be addressed. In short, then, to have a social perspective is partly to experience certain constraints as salient. We should therefore not be surprised if group attachments involve both a distinctive social perspective and certain emotional dispositions: since emotions are one important way in which we might experience certain social constraints as salient, and since perceiving certain social constraints as salient is part of what it means to have a particular social perspective, a shared social perspective may well manifest itself in shared emotional dispositions. This point is vividly illustrated in James Baldwin’s reflections, introduced in Chapter 2, on what it means to be a black American in the 1960s: To be a Negro in [America] and to be relatively conscious is to be in a rage almost all of the time [. . .] Part of that rage is this: it isn’t only what is happening 41  Achen and Bartels (2016, 228). 42  Brun et al. (2008, 18). See also Chapter 2, Section 4, for more extensive discussion.

170  Democratic Speech in Divided Times to you, but it’s what is happening all around you all of the time, in the face of the most extraordinary and criminal indifference.43

Baldwin posits a tight connection between what it means to be a black American and an emotional experience, the experience of profound anger. At the same time, he identifies this anger with the state of being alert to the injustices black Americans are subjected to (being ‘conscious’ of ‘what is happening all around you’). For Baldwin, then, there is no real contrast between the Youngian social perspective of black Americans—the perspective that arises out of their distinctive experiences of social constraint—and black Americans’ experience of anger. Rather, black Americans’ social perspective is partly realized in their shared experience of anger. Thus, the kinds of social groups at play in my Youngian defence of group cognition seem importantly similar to those involved in evidence of group cognition: they often originate arbitrarily; they may well be emotionally charged; and they are drawn along similar axes, such as race, gender, class, region, and so on. Still, one might worry that my argument only applies to a subset of the groups involved in group cognition. Young’s examples of epistemically fruitful social perspectives generally concern groups whose shared experiences involve unfair disadvantage—for example, women, people of colour, and working-­class groups. But group cognition affects relatively privileged groups as well. As discussed earlier, feelings of white Southern identity led white Southerners to flock to the Republican Party from the 1950s onwards. It may seem far less obvious that membership in comparatively privileged groups such as this one facilitates access to distinctive politically relevant knowledge. And if these groups do not yield distinctive knowledge, these instances of group cognition may seem devoid of epistemic value. Privileged perspectives do in fact have some epistemic value, for at least two reasons. First, some genuine social constraints are flipsides of a comparatively high status. For instance, because higher education is a marker of middle- and upper-­middle-­class status, student debt is something disproportionately experienced by these groups. Another reason relates to segregation. Contemporary democracies often involve pervasive segregation along the lines of class, race, and ethnicity. Advantaged and disadvantaged social groups tend to live in ­different residential areas, where they attend different schools and work different jobs.44 These different areas frequently face different challenges relating, say, to transportation, housing, or education. Consequently, privileged groups may have distinctive knowledge of the constraints that affect the areas in which they disproportionately reside. 43  Baldwin et al. (1961, 205).

44  Young (2000, ch. 6); Anderson (2010).

The Problem of Political Ignorance  171 Consider how this last point might apply to the case of Southern whites. Achen and Bartels report that their political realignment was not explained by negative attitudes towards black Americans, nor by policy judgments relating to racial integration (surprisingly, Southern whites who left the Democratic Party favoured and opposed racial integration in equal measure).45 What explained it, they suggest, is rather a sense of affinity with (predominantly white) Southern identity.46 And just as the (predominantly white) rural inhabitants interviewed by Cramer faced distinctive challenges relating to the specific place in which they lived and the specific jobs that they occupied, one might surmise that so, too, did Southern whites. Yet a problem remains: even if privileged perspectives facilitate distinctive knowledge of some important social constraints, these perspectives typically also obscure other, likely more important, constraints. While privileged groups may experience problems relating to their relatively high status (such as student debt), they have comparatively little exposure to the greater social problems that affect disadvantaged groups. Relatedly, while segregation may give them distinctive knowledge of their own neighbourhoods, it simultaneously conceals other neighbourhoods and the out-­groups that inhabit them. So, even if privileged perspectives have some epistemic value, they may well also involve important distortions. This last concern is obvious in the case of racial segregation. One consequence of pervasive racial segregation, which we examined in Chapter 1,47 is that white Americans are often ignorant of the constraints black Americans face. Elizabeth Anderson has powerfully argued that this group-­based ignorance, together with  the psychological tendency for in-­group favouritism, facilitates further perspectival distortions. For example, it makes white Americans more likely to mistakenly explain black disadvantage in terms of black Americans’ inherent dispositions, rather than adverse social circumstances.48 The problem, then, is this. Although privileged group perspectives do involve  valuable group-­specific knowledge, this epistemic contribution may be counterbalanced or even outweighed by epistemic distortions. We can put the point slightly differently. Although privileged social perspectives yield valid pro tanto epistemic reasons for forming judgments about political matters, those reasons may not be sufficiently strong, all things considered, to warrant influencing political judgments so significantly. Insofar as members of privileged groups allow their group perspective to determine their political judgments, they are therefore granting too much weight to some politically relevant evidence (that which comes from their group-­specific experiences) relative to other evidence (that which their group-­specific experiences obscure). Therefore, group cognition 45  Achen and Bartels (2016, 251–8). 46  Achen and Bartels (2016, 255–7). 47  Chapter 1, Section 2.3. 48  Anderson (2010, 44–50).

172  Democratic Speech in Divided Times makes them behave dogmatically. And, unless it is somehow tempered, this dogmatism leads them to form many inaccurate judgments (e.g., distorted perceptions of out-­groups). Note that this problem, though especially acute with privileged perspectives, is actually not confined to them. For Young, the epistemic benefits of disadvantaged perspectives too may come at an epistemic cost: ‘they too are liable to bias [. . .] in overstating the nature of situations, misunderstanding their causes, or laying blame in the wrong place’.49 While the perspectives of disadvantaged groups are distinctively attuned to very important social problems, they nevertheless experience only part of the social context, and are therefore bound to have blindspots. They may, for instance, overlook urgent problems that other disadvantaged groups experience. For example, although the rural perspective Cramer explores is rightly sensitive to the serious economic difficulties that rural communities face, it also tends to underestimate the economic difficulties encountered in urban areas. Partly because of this blindspot, ‘rural consciousness’ typically involves resentment towards urban dwellers whom it commonly depicts as privileged, undeserving, and arrogant.50 So, although the problem is arguably more severe with privileged groups (since these tend to overlook more important social constraints) it is not unique to them. Disadvantaged group perspectives, too, may give excessive weight to some politically relevant information, to the exclusion of other information. This examination of the groups involved in group cognition leaves us with a challenge. Even if they possess epistemic value, social group perspectives also typically involve a countervailing epistemic cost, which is especially severe in the case of privileged perspectives. Though such perspectives reveal politically important group-­ specific constraints, they are liable to neglect even greater constraints relating to out-­groups. When this happens, group cognition gives excessive weight to information stemming from one’s group membership, relative to countervailing pieces of information. It risks, in short, leading to a dogmatic form of cognition. More generally, the defence of group cognition I have offered so far points to  two possible limitations. In Section  3.1, we noted that group-­ specific ­experiences and concerns may be epistemically irrelevant to some domains of  political judgment (particularly ones that involve significant technical ­information). And we have just seen that even when they are relevant, they may dogmatically be given too much weight. The rest of this chapter considers these two concerns in turn.

49  Young (2000, 117).

50  Cramer (2016, 85–8, 140–4).

The Problem of Political Ignorance  173

4.  Technical Group Cognition 4.1  The Problem of Technical Group Cognition In some cases—cases of technical group cognition—group attachments profoundly influence people’s judgments about technical descriptive matters that are relevant to political decisions. In 1988, for instance, Democrats were far less likely than Republicans to know that inflation had dropped since 1980.51 As for Republicans, they were far less likely to know, in 1996, that the deficit had decreased under the Clinton administration.52 The most striking examples, which I will focus on for the remainder of this section, concern scientific testimony about climate change. As Dan Kahan has shown, how far people accept testimony about climate science from a well-­credentialled scientist depends significantly on their group attachments. Specifically, Republicans were less likely to judge that scientists’ testimony was reliable if scientists said that anthropogenic climate change presented a high societal risk than if they said the opposite. The reverse was true for Democrats.53 These cases are distinctively troubling because group attachments, and the perspectives they define, seem irrelevant to assessing the truth of technical facts. According to Kahan, in letting their partisan attachments influence their judgments about technical facts, people are letting normative commitments affect their assessments of descriptive matters. So, technical group cognition seems to involve running afoul of the is/ought distinction: people are fallaciously inferring an ‘is’ from an ‘ought’.54 One might respond that this objection is too quick. Recall that, for Achen and Bartels, partisanship is driven by attachments to broader social groups, such as racial, gender, or regional groups. Furthermore, as discussed in Section 3.1, the perspective defined by social group attachments contains both descriptive and normative components: in particular, descriptive knowledge of certain social constraints, and normative commitments that are intimately associated with these constraints. Since group perspectives do involve descriptive commitments, relying on group attachments to assess technical descriptive matters need not involve deriving an ‘is’ from an ‘ought’. This initial response is unsuccessful. Although group perspectives involve descriptive components, these components typically seem irrelevant to the technical matters at hand. For instance, the experiences that characterize rural life in America—for example, rising joblessness—do not give rural Americans special 51  Achen and Bartels (2016, 277–8). 52  Achen and Bartels (2016, 280–4). 53  Kahan (2015, 4). 54  Kahan (2015, 1). See also Lynch (2019, 68–9).

174  Democratic Speech in Divided Times insight into the reality of anthropogenic climate change. This stands in contrast to the cases examined in Section  3.1. The kinds of social experiences women disproportionately have, for example, do often yield descriptive knowledge that is relevant to judgments about the reality of sexual harassment. So, we are left with our original problem. In cases involving technical descriptive facts, group cognition purportedly lets the wrong kinds of reasons influence our judgments. Often, as Kahan maintains, it seems as though the normative concerns embedded in our group perspectives influence our views about politically relevant descriptive matters. Yet these normative concerns are irrelevant to assessing the truth or epistemic correctness of descriptive claims.

4.2  The Relevance of Values to Descriptive Judgments In fact, normative concerns (e.g., ‘joblessness is a pressing problem that needs addressing’) can be relevant to assessing the truth or epistemic correctness of technical descriptive claims (e.g., ‘anthropogenic climate change is real’). This, as we will see, does not mean that real-­world cases of technical group cognition are always epistemically rational, all things considered. But it does mean that the above objection is far too pessimistic: group-­based normative concerns can, under specific conditions, yield good epistemic grounds for accepting or rejecting testimony about technical matters. To see why, consider that scientific hypotheses are typically underdetermined by evidence: as philosophers of science have long observed, the available evidence hardly ever suffices to fully confirm scientific hypotheses.55 The upshot of this uncertainty is that when scientists decide whether to assert that a scientific hypothesis d is true, they need an epistemic standard for assertion: an account of how strong their evidence for d must be to warrant asserting d.56 How should this standard be determined? Epistemic values should of course play a major role. That is, when deciding whether d is sufficiently supported to warrant being asserted in public, scientists should consider how far it satisfies epistemic considerations of empirical adequacy, predictive precision, internal consistency, explanatory power, and so on. Nevertheless, non-­ epistemic considerations (such as moral and political concerns) also have an important role to play in setting scientists’ epistemic standards for assertion.57 Testimony about politically relevant technical matters can have a significant social impact, for example by influencing which policies are eventually adopted. It would be morally irresponsible for scientists not to consider this potential impact when deciding whether or not to say something. That is, 55  E.g., Lakatos (1970). 56  John (2017, 159–60); Douglas (2009, ch. 5). 57  For a comprehensive overview, see Resnik and Elliott (2019).

The Problem of Political Ignorance  175 when determining whether their evidence for d is sufficient to warrant asserting d, scientists should weigh the social costs of false positives (asserting d when it is false) against the social costs of false negatives (not asserting d when it is true). This weighing, in turn, necessarily relies on normative judgments about the value or disvalue of different outcomes.58 To illustrate, suppose scientist A is deciding whether to assert d (‘anthropogenic climate change is real’). When determining whether her evidence for d is strong enough, A must clearly give significant weight to d’s empirical adequacy, explanatory power, predictive precision, etc. Yet A should also consider the potential environmental damage that would result if climate change is real yet we fail to act against it, as well as the potential harmful effects of environmental regulation on certain jobs if climate change is false and we do act against it. How she weighs these respective considerations depends on how much she values environmental protection compared to job protection. If she values the former more strongly, she might judge that a false positive would be preferable to a false negative, and therefore adopt a lower epistemic standard for asserting d than if she held opposite values. Thus, while A’s epistemic standard for asserting d is constrained by epistemic values, it also depends partly on moral and political judgments. As we will see shortly, the constraints set by epistemic values can sometimes be quite restrictive. But even if moral and political considerations exercise a weaker influence than epistemic values, the relevant point for now is that such considerations exercise some influence. This point, moreover, is not merely theoretical: scientists’ judgments about technical matters (including climate change) do vary somewhat according to their moral and political orientations.59 Now, if A’s epistemic standard for asserting d is influenced—among other things—by A’s normative judgments, then normative considerations are epistemically relevant to assessing A’s testimony. Indeed, if A has incorrect ­normative commitments, this yields a pro tanto reason to believe that A is using an incorrect epistemic standard. Hence, if a listener B disagrees with A’s normative judgments, this gives B a defeasible reason for doubting the epistemic reliability of A’s testimony.60 Suppose, for example, that B comes from a rural community where jobs are scarce and, consequently, believes that one of the state’s normative priorities should be creating jobs in economically deprived areas. Suppose, moreover, that environmental regulation would threaten key job-­providing industries in rural areas. Since the stakes of environmental regulation are very high for B (given her normative concerns) it makes sense for her to adopt a relatively high epistemic standard for accepting claims asserting the reality of climate change. Suppose, finally, 58  Douglas (2009, ch. 4). 60  John (2017, 160).

59  E.g., Slovic et al. (1995, 674); Bolsen et al. (2015, 286).

176  Democratic Speech in Divided Times that scientist A is insensitive to the economic threat rural areas face—perhaps because this threat is less visible to A’s social group than to B’s. If so, then B has a reason to believe that, because of this normative blindspot, A may have an insufficiently high epistemic standard for asserting d. What this shows is that the above objection to technical group cognition is overly hasty: normative considerations (such as those grounded in group-­based perspectives) are, in principle, relevant to whether, epistemically speaking, one should accept claims about technical descriptive matters.61 Technical group cognition need not involve a conceptual fallacy. Of course, there are also epistemically irrelevant ways in which group attachments might influence assessments of technical information. As we saw in Chapter  5, members of different social groups sometimes strongly dislike each other. This group-­based animus can induce people to reject technical descriptive evidence simply because it comes from, or benefits, a resented out-­group. Such a process lets epistemically irrelevant considerations (namely, intergroup dislike) affect assessments of technical evidence. My point here is not that this cannot happen. Rather, in highlighting the epistemic relevance of group-­based normative commitments to technical evidence, I am suggesting that technical group cognition is nonetheless not necessarily misguided. Accordingly, from the mere observation that people let their group attachments influence assessments of technical matters, one cannot immediately infer that they are engaging in epistemically wrong-­headed conduct.62 This conclusion should not be overstated. I have noted that the epistemic reasons supplied by group-­based normative commitments are defeasible. Although group-­based normative commitments can yield some valid grounds not to defer to testimony about technical matters, listeners may nonetheless sometimes be epistemically required, all things considered, to defer. For group-­based resistance to technical testimony to be epistemically apt, overall, at least three conditions must typically be satisfied: the influence of normative considerations on the testifier’s epistemic standard must not be trivial relative to that of epistemic values; the testifier’s normative commitments should differ substantially from the listener’s; and the listener’s group-­based normative concerns must not be deeply misguided. To clarify these three conditions, consider three scenarios where it would arguably 61  John (2017, 158). 62  Some observations relating to technical group cognition in fact fit more easily with the optimistic interpretation I have offered than with the group-­antipathy interpretation. For instance, Berinsky (2017, 258) finds that Republicans and Democrats were more willing to believe a descriptive claim that supported a Democratic policy when it was voiced by a Republican. This fits naturally with the account I have offered. When the speaker reports descriptive information that threatens her own party-­based normative commitments, listeners have reason to believe that the speaker holds this information to a high epistemic standard. By contrast, from the perspective of the group-­antipathy interpretation, it is puzzling why Democrats trust descriptive testimony that supports their cause more when it comes from a Republican than from a Democrat.

The Problem of Political Ignorance  177 be wrong, all things considered, to reject technical descriptive claims because of one’s group-­based normative concerns. The first scenario concerns the comparative influence of epistemic and non-epistemic considerations on the testifier’s standard for assertion. As discussed above, epistemic values constrain the influence of moral and political considerations. In some contexts, those constraints are highly restrictive. Indeed, as Stephen John has argued, the epistemic values of some scientific communities—and specifically, the International Panel on Climate Change (IPCC)—mandate epistemic standards that are extremely high.63 The IPCC excludes relatively untested ‘frontier science’ and non-­peer-­reviewed evidence from its reports.64 So, although moral and political values may still exercise some influence over the IPCC’s ­epistemic standards, those standards cannot fall below a very high threshold. The upshot, for John, is that the IPCC’s epistemic standards for assertion generally exceed those of non-­ scientists, even taking into account the high levels of evidence required by some non-­scientists’ normative commitments. The broader point is this: when the testifier’s epistemic values are so demanding that they severely limit the influence of normative judgments on their standards, listeners would usually be epistemically wrong, all things considered, not to defer to the testimony. Even though a listener’s normative disagreement with the testifier can give her a reason to think that the testifier has an excessively low epistemic standard—as in the earlier example of the rural inhabitant—that reason is likely to be overridden by the following countervailing consideration: that, given her stringent epistemic values, the testifier’s standard must be very high. Accordingly, the testifier’s epistemic standard is very likely to satisfy the listener’s standard, notwithstanding their normative disagreement. The second scenario relates to the content of the testifier’s normative concerns. For the listener’s group-­based normative judgments to give her a reason to doubt the testifier’s technical claims, the testifier’s normative commitments must actually diverge from her own. After all, if the testifier and listener have identical normative judgments instead, then, from the listener’s perspective, the fact that the testifier’s normative concerns influence her epistemic standard provides a reason to think that the testifier has a correct standard. This reveals another set of cases where listeners might be wrong not to defer to testimony about technical matters. Sometimes, listeners mistakenly believe that the testifier’s group-­based normative commitments diverge from their own. For instance, conservative voters might believe that all climate scientists are extremely liberal, when in fact the scientist currently giving testimony has conservative leanings. In such a scenario, the listeners are liable to infer, erroneously, that this

63  John (2017, 167–8).

64  John (2017, 166–7).

178  Democratic Speech in Divided Times scientist’s epistemic standard is lower than their own. And, consequently, they are liable to misguidedly reject the scientist’s claims about climate change. The final scenario concerns the listener’s own normative commitments. In some cases, a listener’s group-­based normative judgments may be somewhat misguided or inaccurate. Now, if normative judgments are inaccurate, they constitute a misleading signal of what the right epistemic standard is: misguided normative judgments risk recommending a standard for accepting technical information that is either too high or too low.65 For example, if a group gives too much normative weight to job creation relative to environmental preservation, they risk adopting an epistemic standard for accepting evidence of climate change that is too high. Put differently, this group will be excessively resistant to evidence about climate change, given how good the evidence is and what the normative stakes at hand actually are. So, we have a further kind of case where listeners would arguably be wrong to reject a scientist’s testimony because of their group-­based normative commitments. The problem here is not necessarily that listeners fail to appreciate how stringent the testifier’s epistemic values are. Indeed, John acknowledges that unusually extreme normative priorities can sometimes lead listeners to adopt epistemic standards that are even higher than the IPCC’s very stringent standards.66 Nor is the problem that listeners misjudge the content of the testifier’s normative commitments. Rather, the problem here is that listeners’ own epistemic standards are driven by misguided group-­based commitments. Hence, they risk being overly resistant to certain forms of technical evidence, and insufficiently resistant to others. The lessons for technical group cognition are therefore nuanced. My core conclusion is that, contrary to what is often thought, technical group cognition is not necessarily epistemically misguided. Given that moral and political commitments exercise some influence over the epistemic standards that guide testimony about technical descriptive matters, group-­ based normative judgments are epistemically relevant to assessing testifiers’ epistemic standards. Having said that, technical group cognition may lead to overall epistemically bad behaviour if group members start from incorrect judgments: for instance, if they are mistaken about what the testifier’s normative concerns are, about the relative influence of those normative concerns on the testifier’s epistemic standards, or about what normative concerns should be prioritized. When this happens, technical group cognition may lead people to be overly or insufficiently resistant, all things considered, to politically important technical information. Importantly, though, this problematic group-­based dogmatism is in fact not specific to the technical domain. It is a problem that we have already encountered

65  John (2017, 161).

66  John (2017, 167).

The Problem of Political Ignorance  179 when discussing the blindspots of social perspectives, particularly privileged ­perspectives, in Section  3.2. As we saw then, it arises more broadly because social  perspectives typically yield partial or incomplete knowledge of society. Consequently, group cognition risks giving excessive weight to some descriptive and normative considerations (those that stem from people’s group-­based experiences) and insufficient weight to others (those that run against people’s group-­ based experiences). It is to this problem that I turn, finally, in the next section.

5.  Dogmatic Group Cognition 5.1  The Problem of Dogmatic Group Cognition So far, I have shown that social group perspectives are epistemically relevant to political judgments, technical as well as non-­technical. Nevertheless, we have also  seen in previous sections that group perspectives often involve epistemic blindspots. While a group’s perspective yields important descriptive and normative information about the constraints encountered by that group, it characteristically involves far less information about out-­ groups and the constraints they face. Because of this evidential bias, group cognition risks giving rise to a form of group-­based dogmatism. Relying on a skewed group perspective when assessing new evidence and claims may make people overly suspicious of, and resistant to, some information and claims (namely, those that jar with their group-­based descriptive and normative commitments) and insufficiently critical of other information and claims (those that accord with their group-­based descriptive and normative commitments). To make this clearer, recall an example offered in the previous section. If rural communities witness job losses resulting from the regulation of polluting industries, but have little exposure to the environmental consequences of those industries, they may overprioritize job creation, relative to environmental preservation. This, in turn, might lead them to embrace excessively demanding standards for accepting reports that climate change is real. The reverse might be said for urban communities. If, because of their partial experiences of social constraint, they give insufficient weight to job creation relative to environmental protection, they may become overly resistant to evidence that some areas depend on polluting industries for their economic survival. As discussed in Section 3.2, this problem is particularly salient with privileged groups, because the constraints these groups are likely to overlook may be even more pressing than the ones they have distinctive access to. If a middle-­class couple have daily experiences of struggling to finance their children’s college education, but little experience of the poverty faced by more disadvantaged

180  Democratic Speech in Divided Times groups, they may end up being more concerned about the former challenge than the latter. Accordingly, they may adopt an excessively demanding standard for accepting the claim that redistributive taxation from middle-­class to working-­ class households is needed. To summarize, when group perspectives involve significant blindspots as well as significant epistemic insights, group cognition may lead to dogmatism. If, as Section 3.2 suggested, such cases of dogmatic group cognition are common, or if they affect pressing political issues (such as climate change), then group cognition might still significantly undermine the epistemic value of democratic public discourse—and, by extension, its ability to promote accountable policymaking.

5.2  The Systemic Role of Dogmatic Group Cognition This worry is nevertheless misplaced. As I will now show, dogma can in fact be epistemically fruitful when considered from a broader, systemic, perspective. By  implication, even dogmatic forms of group cognition can, under the right ­conditions, contribute positively to the epistemic value of democratic public discourse. The insight that dogmatism can be epistemically fruitful is most familiar from, and has been most extensively developed within, philosophical examinations of scientific progress. The starting point for this insight, Imre Lakatos has influentially argued, is that the balance of existing evidence is not always a good indicator of  which scientific theories are best.67 Many scientific theories that eventually proved superior to their competitors previously faced numerous counterexamples and were comparatively poorly supported by the evidence. Take Prout’s hypothesis that the atomic weights of all pure chemical elements are whole numbers. When it first appeared, it was widely inconsistent with existing scientific evidence. Proutians nevertheless refused to reject it. Instead, they hypothesized that the abundant anomalies were due to faulty experimental techniques. Only a century later, when Proutians had tirelessly revolutionized chemistry’s experimental techniques, was Prout’s theory finally widely acknowledged as superior.68 Crucially, according to Lakatos, this instance of scientific progress was enabled by Proutians’ dogmatism: they stubbornly kept their theory despite its glaring inconsistency with the balance of existing evidence; they assumed that these anomalies indicated a problem with experimental methods, not with their theory; and they did so even though they did not yet have the analytical tools needed to support this assumption.69 67  Lakatos (1970). 68  Lakatos (1970, 138–54). 69  Lakatos (1970, 138–54).

The Problem of Political Ignorance  181 Generalizing from historical episodes such as this one, Lakatos concludes that scientific investigation is most productive when many rival research programmes dogmatically defend their core hypotheses. Scientists within a research programme, Lakatos insists, should follow a ‘negative heuristic’: a rule that forbids them from taking countervailing evidence as a sign that the basic tenets (or ‘hard core’) of their programme are false, and instead requires them to refine their auxiliary hypotheses (say, hypotheses about experimental techniques) to explain this evidence away. Thus, the negative heuristic requires scientists to commit dogmatically to the hard core. They strongly discount countervailing data until they can account for it in a way that preserves their core hypotheses. Lakatos is explicit about this: he overtly praises ‘the dogmatic attitude of sticking to a theory as long as possible’.70 While this dogmatism seems epistemically irrational at an individual level— indeed, it requires scientists to strongly discount countervailing evidence— Lakatos’s point is that it is fruitful at a broader, systemic level. When many scientists dogmatically pursue competing theories, this allows the scientific community to refine and test many theories at once. In so doing, the community minimizes the risk of prematurely abandoning superior yet under-­supported theories. Using formal modelling techniques, Kevin Zollman has recently refined Lakatos’s thesis. Like Lakatos, Zollman finds that ‘endowing individuals with dogmatic priors has a good effect when the overall behavior of the community is in focus’.71 An epistemic system with dogmatic agents is less likely to prematurely discard superior theories, and more likely to eventually widely embrace superior theories. But Zollman also qualifies Lakatos’s thesis. He finds that the systemic benefit of dogmatism occurs only if the information yielded by different scientists’ investigations circulates widely between them. By contrast, when dogmatism is paired with low circulation of information between scientists, it impairs the system’s epistemic productivity.72 In these cases, scientists fail to arrive at a consensus on superior theories, even after the community has had ample opportunity to test multiple theories’ promise. The scientific case underscores the following conclusion. Provided that certain background conditions are satisfied—namely, that information circulates widely across the community—dogmatism needn’t impede epistemically fruitful inquiry.

70  Lakatos (1970, 177). 71  Zollman (2010, 33). 72  Zollman (2010, 31–3). A related condition is that the dogmatic agents in the epistemic system must not be absolutely dogmatic. They must give some non-­trivial epistemic weight to countervailing evidence, such that a sufficiently large quantity of it could potentially sway them. Otherwise, theories that prove superior over time could never gain the assent of dogmatic agents. This success condition is related to the ‘circulation of information’ condition: if agents dismiss countervailing information offhand, giving it no consideration or weight whatsoever, then the information has not genuinely circulated between them in any meaningful sense.

182  Democratic Speech in Divided Times Even if dogmatism is irrational at an individual level, the competition of multiple dogmatic groups can stimulate epistemic progress at a systemic level. We can carry this conclusion over to the case of dogmatic group cognition— and, in so doing, partly alleviate concerns about its impact on political discourse. Even if dogmatically upholding the descriptive and normative components of one’s group perspective seems an epistemically defective way for individuals to form political judgments, it does not necessarily follow that this impairs the epistemic fruitfulness of the broader democratic process of public discourse. The competition of multiple dogmatic social groups may, like the competition of dogmatic scientific groups, be epistemically beneficial. More specifically, a democratic community composed of many social groups, each of which defends its starting commitments with some (non-­absolute) degree of dogmatism, is liable to generate more varied information that is relevant to forming sound political judgments. Rural inhabitants who overprioritize job creation will be more disposed to uncover evidence of how environmental regulation harms rural jobs.73 And because, as discussed earlier, they hold evidence of climate change to excessively high epistemic standards, their scrutiny may contribute, say, to revealing genuine limitations with the techniques used in climate science.74 Conversely, committed environmentalists may, because of their contrasting epistemic standards, be more effective at finding flaws in evidence that environmental regulation harms jobs75 and more likely to generate new evidence of anthropogenic climate change. A similar observation extends to privileged groups. Consider again middle-­ class adults who overprioritize the economic difficulties that their group faces. Because of their attending resistance to the idea of redistributive taxation towards working-­class groups, they are more disposed to uncover evidence of problems posed by such policies. They may, for instance, unearth new evidence of the burdens such policies impose on the middle class, or of the impact (if there is one) such policies might have on economic growth. While the countervailing benefits of redistributive taxation highlighted by the perspectives of disadvantaged groups may well outweigh these problems, evidence of such problems is clearly relevant to identifying the best possible policy. So, from a systemic perspective, even privileged group perspectives that suffer from serious blindspots may play a fruitful complementary role. To reiterate, this systemic role of dogmatic group cognition parallels the scientific case. Scientists who dogmatically pursue different research programmes stimulate scientific progress by unearthing different and conflicting findings.

73  Hochschild (2016, 71–2). 74  See, e.g., the emergence of the distinction between ‘high proof ’ and ‘frontier’ evidence, discussed in Edwards (1999, 455–61). 75  Hochschild (2016, 74).

The Problem of Political Ignorance  183 Analogously, different dogmatic social groups can fruitfully complement one another in the democratic community by unearthing different pieces of politically relevant information. However, this conclusion requires qualification. Individual-­level dogmatism improves the epistemic system’s productivity only when the epistemic system involves meaningful networks of communication between dogmatic agents, so that the information each agent generates circulates readily to others. In this light, dogmatic group cognition can gradually contribute to better political judgments, but only when different groups widely share the different information they generate. When they do not, the benefits of dogmatic group cognition—namely, the production of a more varied body of evidence—fails adequately to impact people’s judgments. In these conditions, a wide range of politically relevant information is produced, but never pooled. This qualified conclusion resonates with the systemic approach to public discourse, which I have relied upon in previous chapters.76 Remember that, according to the systemic interpretation of public discourse, inclusive political talk does not occur in a single all-­encompassing arena. Instead, it occurs in a system composed of many smaller parts. Put differently, democratic public discourse should be distributed across very many arenas that have different constituencies.77 Because they have different constituencies, these different discursive arenas typically operate under different working assumptions. An all-­ women consciousness-­raising group, a community meeting in central Detroit, a town-­ hall meeting in rural Wisconsin, and a political party conference take different experiences and concerns for granted. On the whole, these divergent assumptions are not critically scrutinized by the relevant group. Instead, they serve as a premise for discussion. Although different discursive spheres have different constituencies and different working assumptions, proponents of the systemic approach emphasize that they must nevertheless remain sufficiently closely connected to one another. Indeed, Jane Mansbridge warns against letting ‘parts of the deliberative system become decoupled in the sense that good reasons arising from one part fail to penetrate the others’.78 The point is that, once discussion within an arena yields new insights, this information should be shared with other arenas. Only then can the discursive system be truly epistemically fruitful. So, the systemic ideal of democratic public discourse advocates discussion among different groups, where these groups uncritically accept different assumptions, provided that they communicate the results of their discussions to one another. This closely resembles our conclusion: that dogmatically exploring 76  See, e.g., Chapter 1, Section 4; Chapter 2, Section 5; Chapter 5, Section 3. 77  Mansbridge et al. (2012, ch. 1). 78  Mansbridge et al. (2012, 23).

184  Democratic Speech in Divided Times one’s group perspective can play a valuable role in the broader system, provided that the information which emerges from this dogmatic exploration circulates widely between different groups. This consonance matters, because it shows that critics of group cognition are wrong on two counts: not only do they mistakenly hold that dogmatic group cognition necessarily impairs the public process whereby people form political judgments, but they also misleadingly suggest that existing democratic theory is insensitive to this phenomenon. Far from being insensitive to the reality of dogmatic group cognition, normative democratic theory helps to specify the conditions under which such cognition can be epistemically productive. The point so far has been that dogmatic group cognition can be reconciled with the systemic ideal of democratic public discourse. But this does not mean that this phenomenon poses no problem for actual democracies. On the contrary: by suggesting that dogmatic group cognition is acceptable provided that information circulates freely between different groups, the systemic ideal helps identify more precisely what goes wrong in actual democratic public discourse. The problem is that contemporary democracies often involve dogmatic group cognition and a lack of intergroup communication. There are at least two related reasons for this lack of intergroup communication, both of which we have encountered previously. First, different groups often live segregated lives: they inhabit different regions or neighbourhoods, occupy different jobs, belong to different social media enclaves, and consult different news outlets. Because of this fragmentation, different groups tend to have little contact with one another. Second, as we saw in Chapter  5, contemporary democracies involve substantial intergroup dislike and distrust (or ‘affective polarization’). Therefore, people from different groups may refuse to speak to each other even when their paths do meet. These two conditions are causally linked. Thoroughgoing segregation, Anderson has argued, plays a central role in explaining why different groups come to dislike each other. Segregation leads to infrequent contact with out-­ groups, which induces people to underestimate the challenges out-­groups face, to misattribute instances of bad behaviour to their essential dispositions, and to regard deviant members of out-­groups as representative of the whole group. Segregation thus crucially facilitates intergroup dislike. In turn, mutual dislike induces different groups to avoid each other and thereby self-­segregate.79 The result is a problematic fragmentation of the public sphere. Because different groups inhabit separate spheres, and because they are reluctant to speak and  engage with one another, they fail to pool the insights and arguments that they independently generate. We end up with the unhealthy combination that ­philosophers of science discourage: dogmatism and low circulation of information. 79  Anderson (2010, 67–75).

The Problem of Political Ignorance  185 This diagnosis has a crucial upshot. It renders more tractable the problem that group cognition poses for actual democracies. What makes the problem of group cognition so daunting, as noted in Section  2, is that it stems from dispositions that are deeply embedded in human psychology. Accordingly, it seems as though saving democratic public discourse would require, as Achen and Bartels claim, ‘a radical change in human nature’.80 But once we appreciate that the problem is not (dogmatic) group cognition per se, but rather its conjunction with a fragmented public sphere, alternative remedies come into view. We can strive to make public discourse less fragmented, by tackling the segregation and mutual antipathy that obstruct intergroup communication. Doing so is difficult, to be sure. But it does not involve altering human nature itself. For one thing, we saw in Chapter  5 that there are strategies internal to democratic public discourse for rebuilding goodwill between estranged social groups. And even when these discursive strategies are not readily available, or when they are limited in efficacy, we can also tackle the intergroup segregation that keeps groups apart and (as a result) facilitates their mutual antipathy. Anti-­ segregation strategies might include redrawing the boundaries of political districts to make them less homogeneous; incentivizing the creation of more diverse neighbourhoods; promoting greater intergroup integration within schools; encouraging news stories to include links to opposing coverage of the same stories, and so on.81 The next chapter will be devoted to examining and assessing such ‘integrative’ policies, which aim to make the public sphere less fragmented by combatting intergroup segregation. In sum, situating dogmatic group cognition in a broader system of public discourse, by analogy with scientific practice, is helpful in several ways: it reveals that dogmatic group cognition is not inherently epistemically problematic for political judgment; it highlights the auxiliary conditions that can nevertheless make it so; and, consequently, it identifies tractable strategies for addressing the real-­world difficulties created by dogmatic group cognition.

6. Conclusion We started with the following problem. When surveyed, people often seem to know very little about the details of policymaking. This ignorance, in turn, raises the worry that they may be incapable of engaging in fruitful public discussion about politics. Now, in principle, people could overcome this problematic ignorance by employing information shortcuts. In particular, they could circumvent gaps in their political knowledge by relying on the political cues and

80  Achen and Bartels (2016, 10).

81  Anderson (2010, chs. 6–7; 2011, 158).

186  Democratic Speech in Divided Times testimony offered by more informed public speakers. But it is often held that the phenomenon of ‘group cognition’ precludes this possibility: people’s political judgments are in fact deeply shaped—critics would say, ‘distorted’—by their group loyalties. Since this process of group cognition is taken to be epistemically unreliable, the alleged upshot is that people fail to learn effectively from information shortcuts in the public sphere. Yet this objection, I have argued, is far too quick. It wrongly dismisses group cognition as an impediment to political learning within the system of public discourse. But in fact, social group attachments involve an epistemically rich social perspective, which is informed by group-­specific experiences of social constraint and opportunity. This experience-­based knowledge and the attending normative concerns seem a relevant basis for deciding whose policy views—and, by extension, whose political testimony—to trust. Still, group cognition might seem distinctively problematic in two types of cases: when group perspectives are brought to bear on testimony about politically relevant technical matters; and when the judgments or concerns contained in group perspectives are incomplete or partly distorted, so that they lead to dogmatic resistance to new evidence and claims. Fortunately, developments in philosophy of science help alleviate both problems. Because normative commitments inevitably influence scientists’ epistemic standards, group-­based normative concerns are epistemically relevant to assessing scientific testimony on technical matters. Reflecting on the nature of scientific investigation is also instructive with respect to dogmatic group cognition. Having many research programmes dogmatically investigate their core hypotheses can help produce greater scientific knowledge. Analogously, having many social groups engage in dogmatic group cognition can help generate more varied politically relevant information than would otherwise emerge. If pooled appropriately, this information can in turn contribute to enriching people’s political judgments. This insight, moreover, seems fully congruent with the systemic ideal of public discourse that I have been recommending. So, group cognition need not jeopardize the ideal of democratic public discourse. On the contrary, within an integrated system of public discourse, the practice of using group membership to select information shortcuts can help otherwise uninformed citizens identify policies, parties, and politicians that are responsive or accountable to their interests. This, to emphasize, does not mean that group cognition is never problematic in actual democracies. As we have seen, when the public sphere is not integrated, but rather deeply fragmented, dogmatic group cognition may impede the formation of sound political judgments. What this last conclusion reveals is that, to the extent that group cognition poses a problem for democratic public discourse, it does so not on its own, but  only in conjunction with particular social conditions. This insight has an

The Problem of Political Ignorance  187 important normative payoff. If saving democratic public discourse required ­altering human nature, we might despair of ever doing so. Instead, however, we can target the social conditions that divert group cognition from its positive potential. Arduous though this task may be, it is more tractable than the alternative: it targets, not built-­in features of human cognition, but social arrangements that are ultimately contingent. Accordingly, in the next chapter, I will investigate how we might pursue the de-­fragmentation of the democratic public sphere in a morally permissible way.

7

The Problem of Fragmentation 1.  Introduction: A Fragmented Public Sphere Public discourse should be inclusive, or ‘cross-­cutting’. It is important, in other words, that public discourse bring together people from diverse cultures and diverse social positions. In previous chapters, we have encountered at least two reasons why this is so. The first is epistemic. Having citizens from different walks of life trade stories and exchange reasons for and against practical proposals helps to pool socially dispersed information that is relevant to good policymaking.1 The second concerns domination. By publicly challenging decision makers, citizens can make them more accountable or responsive to their concerns and interests. Accordingly, when none are excluded or insulated from public debate, each has the opportunity to hold others to account, and thereby prevent them from wielding power arbitrarily.2 Yet we have also seen, notably in the last chapter, that actual public discourse tends to be far removed from this ideal. The public sphere of contemporary democracies is typically deeply fragmented, such that people predominantly talk to members of their own social group.3 As Diana Mutz famously reports, ‘like talks to like [. . .] network survey after network survey has shown that people talk more to those who are like them than to those who are not’.4 Indeed, people talk far more to those who are like them than to those who are not: ‘in some of the evidence, under one quarter of the US population reported exposure to even one person of oppositional political views through their personal networks’.5 Thus, instead of people from different social groups coming together to share their insights and to hold one another accountable, people talk politics within homogeneous clusters. Talking across difference—be it partisan, racial, ethnic, regional, or class-­based—is the exception, not the rule. This absence of inclusive political discourse is of course no accident. As Chapter 6 noted, it is rooted, in significant part, in deep and multifaceted patterns of group-­based segregation. For one thing, social groups tend to be spatially

1  Introduction; Chapter 1, Section 5; Chapter 2, Section 4; Chapter 6, Section 5. 2  Introduction; Chapter 1, Section 2; Chapter 3, Section 4. 3  Chapter 6, Section 5. See also Chapter 5, Section 3. 4  Mutz (2006, 9). 5  Mutz (2006, 41). For similar observations in offline and online contexts, see also Talisse (2019, 73) and Settle (2018, 65–6).

Democratic Speech in Divided Times. Maxime Lepoutre, Oxford University Press (2021). © Maxime Lepoutre. DOI: 10.1093/oso/9780198869757.003.0008

The Problem of Fragmentation  189 segregated. Different racial, partisan, ethnic, and class-­based groups overwhelmingly live in different neighbourhoods or regional areas. In addition, these groups are commonly socially segregated.6 They attend separate schools, work different jobs, visit different places of worship, and have different networks of friends and acquaintances.7 This social segregation extends to online spaces, such as social networking sites, where members of different groups often cluster in differentiated networks of friends or contacts.8 Later, we will see that this multifaceted segregation, or fragmentation,9 has ­plural causes. It results partly, and to varying degrees, from people’s choices regarding where to live and work, what activities to practice, and with whom to associate. But it also results partly, and to varying degrees, from structural forces that define and constrain people’s residential, professional, and social options. For now, what matters is this: whatever its specific causes, thoroughgoing intergroup fragmentation systematically threatens the possibility for inclusive public discourse. The purpose of the present chapter is to investigate what measures are mandated by this problem. If widespread spatial and social fragmentation jeopardizes inclusive public discourse, and if (as I have been arguing) inclusive public discourse is normatively central to democratic life, it seems that we ought to reverse this fragmentation. Yet policies aimed at doing so—integrative policies—remain deeply controversial. These normative controversies have received surprisingly little attention within general discussions of democratic public discourse. Although deliberative the­or­ ists widely lament the threat posed by intergroup fragmentation, their arguments remain primarily diagnostic: they offer relatively little examination of the policies and of the costs involved in reversing this fragmentation.10 By contrast, this issue has received ample attention within the specific context of racial integration. A heated debate has ignited surrounding the normative costs of policies aimed at bringing segregated racial groups into closer physical contact with one another. Given that this constitutes the most sophisticated existing discussion of the normative costs of integration, it will constitute a helpful starting point for thinking about the costs involved in de-­fragmenting public discourse.

6  See, e.g., Massey and Denton (1993), Young (2000, ch. 6), Fiss (2003), Anderson (2010, chs. 1–3), Sampson (2012, ch. 5), Cramer (2016), Dottle (2019), and Talisse (2019, ch. 3). 7  See, e.g., Young (2000, ch. 6), Anderson (2010, chs. 1–3), and Talisse (2019, ch. 3). 8  For an overview of evidence of online segregation, see Sunstein (2017, 114–26). 9  Here, I use the terms ‘segregation’ and ‘fragmentation’ interchangeably. 10  See, e.g., Mutz (2006), Sunstein (2017, ch. 1), and Talisse (2019). Though Talisse and Sunstein do offer some prescriptions, most of their argument is diagnostic. Moreover, the prescriptive part of Talisse’s argument is more an articulation of the condition that should be cultivated—namely, civic friendship—than a normative assessment of attempts at realizing this condition. A partial exception here is Fishkin (2018), who articulates and defends proposals aimed at fostering political talk across difference. In Section 4, I will argue that his approach, though valuable, is insufficient for our purposes.

190  Democratic Speech in Divided Times Having said that, I will depart from existing defences of racial integration in an important respect. Although advocates of racial integration do sometimes care about improving public discourse, their emphasis is typically on integration as a way of achieving social equality.11 My aim is somewhat different. In line with my broader inquiry, I aim to offer a defence of integrative policies—of policies that combat segregation in its spatial and social dimensions—that focuses on their value for inclusive public discourse. There are two reasons why this approach to defending integrative policies seems distinctively promising. First, some of the ways in which fragmentation undermines public discourse are unrelated to social equality. Indeed, as the case of partisan segregation suggests, segregation can impede political discourse between groups that are not particularly unequal. This matters, because critics of racial integration have persistently argued that integrative policies are not strictly necessary to achieving social equality. For example, both Tommie Shelby and Iris Marion Young insist that redistributive policies could be just as effective.12 A discourse-­oriented approach to justifying integrative policies can partially sidestep this empirical challenge. If segregation undermines public discourse, even in the absence of social inequality, then we have strong reasons to implement integrative policies even if there are alternative ways of remediating social inequality. Second, a focus on integration as a way of improving public discourse goes some way towards alleviating the costs that have traditionally been associated with integrative policies. Here, the systemic nature of public discourse once more proves important: I will argue that, from a systemic perspective, the integrative policies needed to restore fruitful political communication are less demanding— and so, less morally costly—than they initially appear. Thus, my main contention will be the following: we need integrative policies to salvage inclusive public discourse; and upon closer examination, we have good reason to think that these policies do not come at an intolerable moral cost. My argument will proceed as below. After threading together the various ways in which spatial and social fragmentation impairs inclusive public discourse— some of which have to do with social equality, others not—I suggest that this problem generates a strong pro tanto justification for integrative policies

11  Fiss (2003); Anderson (2010). Although Anderson does discuss how segregation undermines democracy (ch. 5), most of her investigation focuses squarely on social equality (chs. 2–4). Moreover, Anderson’s democracy-­based argument for integration is itself partly about social equality. That is, when articulating how segregation undermines democracy, only part of Anderson’s argument concerns public discourse; the other part rather emphasizes the fact that segregation threatens the egalitarian social relations that constitute the democratic way of life. 12  Shelby (2016, 68–9, 77); Young (2000, 277). See also Valls (2018, ch. 6). Anderson (2010, 188; 2013, 7), Fiss (2003, 25–7), and Poama (2020, 133) argue that redistributive measures are highly imperfect strategies for realizing equality between segregated groups. But even if they are right, this is not a decisive response. After all, Shelby (2014, 275–6) contends that residential integration too is imperfectly effective at achieving social equality.

The Problem of Fragmentation  191 (Section 2). Next, I consider a serious objection: that integrative policies severely curtail people’s associative freedom, and that this curtailment is particularly unacceptable in the case of vulnerable minorities. This objection, I argue, overlooks the meaningful ways in which integrative policies also advance freedom of association. So, even if such policies do significantly curtail freedom of as­so­ci­ ation in some respects, it does not necessarily follow that they are bad, overall, for associative freedom (Section  3). But we should not stop there. Even if the potential associative costs of integration do not yield a decisive case against it, they remain very serious. Accordingly, the rest of the chapter goes further and demonstrates that, with a systemic understanding of public discourse in hand, we can substantially alleviate the gravest costs involved in integrating public discourse (Section 4).

2.  The Public Discourse Argument for Integration 2.1  How Fragmentation Impairs Inclusive Public Discourse The spatial and social fragmentation of contemporary democracies—along axes such as party, race, ethnicity, and class—threatens inclusive public discourse in direct and indirect ways. First, and most directly, multifaceted segregation makes it less likely that different groups will spontaneously enter into contact with one another. If members of different groups live in different neighbourhoods or work different jobs, they are less likely to encounter each other in the course of their everyday lives. And if they are less likely to encounter each other, they are also less likely to talk to each other.13 A second and more indirect problem concerns material inequality. Spatial and  social segregation plays an important role in maintaining and amplifying intergroup economic inequality. This phenomenon, as Young has notably shown, is particularly vivid in the case of spatial segregation.14 When poorer groups are  residentially segregated from wealthier groups, this tends to exacerbate the in­equal­ities between them. Residents of segregated lower-­income neighbourhoods are often physically distant from, and therefore unable to take and keep, better jobs. They find themselves relatively isolated from valuable networks that  disseminate information regarding professional opportunities. They often pay higher taxes in return for fewer and worse public services, such as transportation, schools, and law enforcement. Meanwhile, businesses struggle to maintain

13  Mutz (2006, 84). See also Young (2000, 210) and Welch et al. (2001, 62). 14  Young (2000, 205–6). See also Anderson (2010, ch. 2).

192  Democratic Speech in Divided Times themselves in this environment, which contributes to a further decline in local property values, and to further poverty. Severe material inequality, in turn, constitutes an obstacle to inclusive public discourse. Wealth readily amplifies the speech of those who have it. Affluent groups can use their economic power, say, to buy political advertising space or to finance well-­drilled political advocacy organizations. When this happens, the concerns and insights of disadvantaged groups risk being drowned out by others’ unfairly loud voices. The other indirect problem is that spatial and social segregation facilitates the emergence of deeply negative attitudes, such as distrust and dislike, between groups (or ‘affective polarization’). Specifically, because group segregation limits intergroup contact, people tend to overlook the challenges that out-­groups face, to falsely attribute instances of bad behaviour by members of out-­groups to their essential dispositions, and to regard deviant members of out-­groups as representative of the whole group. Thus, the problem of fragmentation fuels the problem of goodwill diagnosed in Chapter 5.15 This last difficulty has perhaps most extensively been illustrated in the context of interracial stigma in the United States. The fact that white Americans have little exposure to the constraints experienced by black Americans makes them more likely mistakenly to explain black disadvantage in terms of stigmatizing racial stereotypes.16 But the difficulty at hand is of course not specific to the racial case. There is evidence that segregation, and the mutual ignorance it supports, also facilitates the development of vilifying attitudes between rural and urban communities, between Republicans and Democrats, and between working-­class and upper-­class groups.17 For reasons that we have already encountered in Chapter  5, the intergroup antipathy facilitated by segregation constitutes yet another obstacle to inclusive public discourse. Even when different groups do come into contact with one another, and even when they do have the material resources needed to make themselves heard, they may nonetheless be unwilling—because of their deeply negative views of one another—to talk together. In sum, whether directly or indirectly, spatial and social fragmentation makes discussion across difference far less likely. This constitutes an urgent problem for public discourse, both as a device for preventing domination and as an epistemic tool. If some groups remain insulated from the challenges of other groups, then they can, to that extent, avoid being publicly held to account for their actions. And if different groups are unable or unwilling to talk together, they are less likely to pool their group-­specific insights. 15  On this point, see also Chapter 6, Section 5. 16  Chapter 6, Section 3.2. See also Anderson (2010, ch. 3). 17  See, respectively, Cramer (2016), Talisse (2019, 86–8), and Isenberg (2016).

The Problem of Fragmentation  193 This conclusion, in itself troubling for democratic speech, has two notable and negative ramifications. First, it makes public discourse vulnerable to group-­based dogmatism. As we saw in Chapter  6, group-­based dogmatism can work for democracy only if different social groups readily pool their information and insights. If social and spatial fragmentation undermines this pooling, group-­ based dogmatism will instead further erode democratic processes of learning and accountability. Second, in conditions such as these, political discussion risks leading to belief polarization. When a homogeneous group is insulated from out-­group challenges and social pressure, and when it does not pool its insights with those of out-­ groups, its members tend to adopt more extreme versions of their initial beliefs, irrespective of what those initial beliefs were. When Republicans speak exclusively to Republicans, for instance, they are likely to adopt increasingly conservative views. Likewise, when Democrats speak exclusively to Democrats, they are likely to adopt increasingly liberal views.18 The fact that homogeneous discussion leads to belief polarization is doubly worrying. For one thing, belief polarization is epistemically dubious: shifting to more extreme views may just as well bring people further from the truth as closer to it. What is more, belief polarization may breed further social and spatial fragmentation. As Robert Talisse explains, the more belief polarization occurs, the more opposing groups’ perspectives will grow apart from each other. As a result, each group will increasingly tend to see the other as outrageous, benighted or unintelligible. And the more this intergroup dislike and alienation grows, the more different groups will be disposed to stay away from each other.19 The mutually reinforcing dynamic that Talisse identifies between fragmentation and belief polarization has an important upshot. The problem is not simply that fragmentation, by discouraging inclusive public discussion about politics, jeopardizes the epistemic and anti-­domination functions of such discussion. It is also that fragmentation’s negative effect on public discourse risks giving rise to still more fragmentation, in a self-­perpetuating cycle.

2.2  The Integrative Solution Spatial and social fragmentation constitutes a severe and self-­ perpetuating impediment to inclusive public discourse. Accordingly, insofar as we value inclusive public discourse, we have strong reasons to adopt integrative policies. We have strong reasons, in other words, to adopt policies aimed at re-­establishing

18  Sunstein (2002). The phrase ‘belief polarization’ is Talisse’s (2019). 19  Talisse (2019, 118–20).

194  Democratic Speech in Divided Times substantive contact and interaction between social groups, across the various ­arenas where political talk might conceivably occur. We should pause, briefly, to emphasize why integration is needed. As advertised in the introduction, some critics of integration contend that, even if social and spatial fragmentation amplifies intergroup inequality, it does not follow that we should adopt integrative policies. This is because redistributing resources may be a more efficient way of combating inequality.20 But the present argument for integration sidesteps this challenge. As we have just seen, some of the ways in which fragmentation undermines inclusive public discourse are unrelated to social equality. Residential and social fragmentation impairs political discussion across groups not just by amplifying intergroup in­equal­ities, but also by limiting the opportunities for spontaneous intergroup contact and by facilitating intergroup antipathy and distrust. So, even if we conceded, controversially, that redistributive policies are the best way to remediate intergroup inequality, such policies would remain insufficient as a response to our problem. Integrative policies are needed as well. What might this involve, concretely? Elizabeth Anderson’s defence of racial integration remains the most systematic attempt at addressing this question to date. For Anderson, the integrative solution should tackle fragmentation in its various manifestations. In the first instance, we should achieve spatial integration by creating socially mixed neighbourhoods. But this, Anderson insists, is not enough. We should also integrate other social arenas, such as the workplace, juries, schools, universities, as well as online spaces. To promote this objective, Anderson recommends a battery of policies, including: eliminating class-­segregative zoning regulations; distributing housing vou­ chers that can only be used in certain districts; adopting aggressive affirmative action programmes in universities and the workplace; redefining the boundaries of school districts; abolishing group-­based—particularly race-­based—academic ‘tracking’ within schools; selecting for juries that have an integrated composition; and restructuring online spaces to ensure that different groups are exposed to one another.21 Now, Anderson offers these policies with a view to racial integration. And different axes of fragmentation may of course call for different specific integrative policies. Nevertheless, taken as a starting point, Anderson’s picture of integration remains highly informative. First, her account brings into view some of the most significant normative challenges involved in realizing an integrated society. Anderson’s integrative ideal requires integration across multiple dimensions, including the residential dimension. And, as we will see in Section 3, the greatest normative challenges of integration arguably have to do with racial residential 20  See note 12.

21  Anderson (2010, chs. 6, 7, 9; 2011, 158–9).

The Problem of Fragmentation  195 integration. Starting from Anderson’s demanding account will therefore be useful, since my purpose in the rest of the chapter is to assess the potential challenges involved in achieving intergroup integration. Second, we should not separate the different axes of integration too neatly. Many of these axes are reinforcing. For example, because black Americans are on  average poorer than white Americans, predominantly committed to the Democratic Party, and disproportionately concentrated in urban areas, there is significant overlap between racial integration, on the one hand, and class, partisan, and geographic integration, on the other. So, starting from policies aimed at achieving racial integration sheds some light on what might be involved in achieving class or partisan integration.22 Let us take stock. To salvage public political discourse, we have strong reasons to adopt integrative policies aimed at bringing fragmented social groups into closer contact with one another. But these reasons could in principle be overridden by countervailing reasons. In what follows, I will consider what I take to be the strongest source of such countervailing reasons: namely, the worry that integration violates the associative freedom of the groups it brings together.

3.  Integration and Freedom of Association 3.1  Integration against Freedom of Association? People have a strong interest in freedom of association. They have a strong interest, that is, in having the opportunity to choose with whom they do and do not associate—and as a precondition thereof, with whom they do and do not engage and interact.23 The integrative project arguably restricts this freedom. In particular, integration sets constraints on people’s options to dissociate from others. Anderson, for example, explicitly states that integration is inconsistent with people engaging in group-­based self-­segregation.24 Moreover, integrative policies often enforce these constraints by pressuring members of different groups into sustained contact with one another. This restriction is visible across different potential arenas of integration. If social media platforms are restructured so as to be less fragmented, people may

22  This is particularly apparent within Fiss’s (2003) defence of integration, in which class and race are inextricably bound up with one another. 23  For a helpful discussion, see Brownlee (2015, 2019). Brownlee (2015, 269; 2019) emphasizes that interactions are prerequisites for association, and that interactions and associations bleed into each other. So, insofar as people are made to interact with some people but not others, this also affects their associative opportunities. 24  Anderson (2010, 183–4).

196  Democratic Speech in Divided Times have less control over whom they see online, and whom they do not see online. Similarly, if schools are forcibly integrated, then children and their parents will regularly have to come into contact with children and parents whom they may have wanted to avoid. That being said, the concern that integration restricts associative freedom appears most often in the context of residential integration. When a policy induces group A to move into a neighbourhood primarily inhabited by group B, B are forced (at the cost of moving) to live alongside a group that they may not have wanted to live with. The associative freedom of group A is often also curtailed. When, as Anderson recommends, the state gives group A housing vouchers that can only be used in B’s neighbourhood, it thereby restricts recipients’ associative options in order to promote an integrative goal. Although members of A retain the opportunity not to move, availing themselves of this opportunity comes at the cost of foregoing subsidized housing—a cost that is generally very high, not least because housing vouchers are generally distributed to communities living in ­significant poverty. Hence, by making access to a basic social good (affordable housing, typically in a safer and wealthier neighbourhood) conditional on living alongside B, such a residential policy exercises strong pressure on members of A to do so. This pressure is so great that Andrew Valls, in his critique of racial ­integration, labels such vouchers ‘coercive’.25 What this suggests, for many critics, is that the integrative solution is insufficiently respectful of people’s freedom of association. On this line of thought, which Young and Shelby notably articulate, people should of course have the opportunity to integrate—and so, to interact and live with members of other social groups—if they so desire. However, they should not be forced or strongly pressured to do so. In denying this, Young observes, the integrative solution deprives people of the opportunity not to associate with other social groups. As a result, it leaves little room for many people’s desire, which Young considers under­stand­ able, predominantly ‘to live and associate with others for whom they feel particular affinity’.26 Neither Anderson nor other advocates of integration have much to say in response to this critique. When Anderson considers freedom of association, her aim is simply to show that racial segregation in the United States historically did not result from people’s uncoerced decisions to self-­segregate.27 But this does not address the present concern. Even if historical patterns of spatial and social fragmentation resulted from unjust acts of coercion and discrimination, the point that critics of integration are making is that some forms of thoroughgoing fragmentation should nevertheless remain permissible: namely, those that result from 25  Valls (2018, 146). 26  Young (2000, 216–17; see also 224). See, similarly, Shelby (2016, 67) and Hochschild (2003, 219). 27  Anderson (2010, 69–73).

The Problem of Fragmentation  197 people’s voluntary decisions to dissociate from others.28 The problem is that the integrative solution seems to rule these out. This silence constitutes an important oversight. Freedom of association is widely regarded as one of the most fundamental individual freedoms.29 Accordingly, if integration strongly curtails people’s freedom of association, then it may come at an extremely high—and perhaps, intolerably high—moral cost. There are, in fact, several things to say in response to this concern. A pre­lim­in­ ary response is that the value of associative freedom, though very important, is not absolute. Kimberley Brownlee, for instance, stresses that the interest in dis­ soci­at­ing or disengaging from others can be overridden when dissociation would jeopardize basic needs.30 This is relevant to the issue of integration. If, as I have suggested, epistemically sound and accountable policymaking depends im­port­ ant­ly on the existence of intergroup communication, then we arguably have a basic need to re-­establish such inclusive discourse. And, by implication, we arguably have a basic need for integration. So, even if integration strongly curtails associative freedom, our deep interest in inclusive political communication could still outweigh this cost. But we should not stop at this initial response. Baldly asserting that the value of inclusive public speech overrides the value of freedom of association would make the case for integration too controversial. So we should go further, and challenge the alleged inconsistency between integration and associative freedom. Here, two important observations are in order. The first is that restrictions on freedom of association are especially problematic when they affect intimate or close associations. Very few grounds could justify forcing or strongly pressuring someone to be best friends with, or married to, a complete stranger.31 But the social integration needed to re-­establish intergroup political communication does not require forcing or pressuring anyone into such intimate associations. What is needed is that people regularly communicate about politically relevant matters with people from different walks of life. People do not need to be intimate friends or family to do this. Mutz is unequivocal on this last point. In her seminal examination of political discussion networks, she finds that people are in fact more likely to be exposed to countervailing or out-­group perspectives by their ‘weak ties’ than by their close or intimate ties.32 Accordingly, when examining how we might promote a greater volume of political discussion across difference, Mutz singles out weak relations over intimate relations: ‘in order to increase levels of exposure to oppositional

28  See, e.g., Shelby (2016, 50–7). 29  On the value of freedom of association, see White (1997, 385), Quong (2011, 15), and Rawls (2001, 45). 30  Brownlee (2015, 273; 2019, section 3.2). See also White (1997, 382ff.). 31  White (1997, 386); Brownlee (2019, section 3.1; 2015, 273). 32  Mutz (2006, ch. 2).

198  Democratic Speech in Divided Times views in the population,’ she asserts, ‘people will need to have a greater number of weak ties and probably less intimacy on average within their [discussion] networks’.33 In line with this recommendation, the integrative policies I have put forward typically do not aim to pressure members of different groups into close associations, such as tight friendships or loving relationships. As we have seen in Section 2.2, they rather aim to multiply the situations in which members of different groups are, say, colleagues, schoolmates, or neighbours. So far, I have simply argued that the cost integration imposes on associative freedom is lower than it initially appears. But there is a second and more fundamental problem with the freedom of association critique. It overlooks the fact that, although integration constrains freedom of association in some respects, it also increases freedom of association in others. Freedom of association, as defined above, consists in having the opportunity to choose with whom one does and does not associate. Hence, the extent of one’s associative freedom is a function of two factors: negatively, the extent to which one can dissociate from others; and positively, the extent to which one can form associations with others. The positive dimension of associative freedom has a key implication. It means that one’s freedom of association depends in part on the range and quality of one’s options to form associations with others. Indeed, as Brownlee notes, the op­por­ tun­ity to choose with whom one does and does not associate would be largely devoid of value, if there were no people one could potentially associate with.34 Crucially, moreover, the range and quality of one’s associative options is defined, in significant part, by how others exercise their freedom of association. If everyone decides to dissociate from me, I will be left without associative options.35 The upshot of this last point is that restricting associative freedom in some respects can increase this freedom in others. By preventing A from dissociating with B, I decrease A’s negative opportunities to dissociate from others. But in doing so, I also increase B’s positive opportunity to form associations with others. This theoretical upshot bears directly on the issue of integration. Critics of integration, as we saw earlier, typically accept that people should have the op­por­ tun­ity voluntarily to integrate and associate with other groups. What they insist is that people should nevertheless also have the opportunity not to integrate. There is, however, an unacknowledged tension here. For the reasons just outlined, indiscriminately protecting social groups’ opportunity not to integrate can jeopardize others’ opportunity voluntarily to integrate. When one social group A actively dissociates from group B, members of B lack the positive opportunity to interact or associate with members of A.

33  Mutz (2006, 86; see also 2, 26, 54). 35  Brownlee (2019, section 3.2).

34  Brownlee (2015, 268, 275–6).

The Problem of Fragmentation  199 For example, if—as has historically been the case—white Americans flee to the suburbs when black Americans move into their neighbourhoods, this deprives black Americans of the opportunity to live alongside and interact meaningfully with white Americans.36 A similar observation holds in circumstances where most members of two social groups dissociate from one another, as may increasingly be the case with Republicans and Democrats.37 When this happens, members of either group who would rather live in a mixed community lack the opportunity to do so, because the aggregate effects of others’ associative decisions result in a dearth of mixed neighbourhoods. What cases such as these reveal is that integrative policies are not categorically bad from the standpoint of associative freedom. By restricting some op­por­tun­ ities to dissociate, they can also protect some positive opportunities to form associations. In sum, even if integration does restrict freedom of association, and even if this freedom is extremely valuable, it does not follow that we should forego integrative solutions. First, integrative solutions generally do not restrict the most valuable form of associative freedom—namely, freedom of intimate association. Second, and more fundamentally, integrative policies also benefit associative freedom. In principle, these associative benefits could outweigh the associative costs of integration. So, absent further considerations, the objection at hand fails to establish that integration is bad, all things considered, for freedom of association.

3.2  The Freedom of Association of Vulnerable Groups Still, even if it is unclear why in general the associative costs of integration must outweigh its associative benefits, this concern may be more compelling when applied to specific forms of integration. Put differently, one might think that some particular integrative policies impose distinctively high associative costs. The most powerful argument to this effect considers residential integration involving vulnerable or historically disadvantaged minority groups. Specifically, Shelby has influentially argued that the push for residential integration between white and black Americans neglects the strong interest that black Americans have in retaining the opportunity to reside away from whites and, relatedly, to live in predominantly black neighbourhoods. Shelby articulates two reasons for thinking that this associative opportunity is especially valuable. The first is that there may be significant costs, for vulnerable minorities, to residing in neighbourhoods predominantly populated by members of the dominant majority. For example, black Americans may want ‘to avoid 36  See, e.g., Sampson (2012, 304–8) and Welch et al. (2001, 69).

37  Talisse (2019, ch. 3).

200  Democratic Speech in Divided Times residing in white neighborhoods to limit unpleasant experiences with whites’, such as experiences of racist discrimination or abuse.38 From this perspective, residential self-­segregation provides black Americans with refuge from hostile or unwelcoming out-­groups. Second, there may also be an important positive appeal, for vulnerable mi­nor­ ities, to residing in a neighbourhood predominantly populated by members of their own group. According to Shelby, black Americans may prefer living in a black neighbourhood because they value living alongside people who share important aspects of their heritage, culture, and life experiences; because such neighbourhoods are likely to contain more businesses and associations that cater to black Americans’ tastes or preferences; and, crucially, because spatial concentration may allow black Americans to exercise greater influence as a group in local politics.39 What Shelby’s argument suggests is that black Americans may have a very strong associative interest in retaining the opportunity to live in a segregated neighbourhood. Accordingly, the associative cost that enforced racial residential integration imposes on black Americans seems extremely high. It is important to emphasize, moreover, that this problem is fairly widespread. While Shelby focuses exclusively on racial residential integration, Young points out that similar considerations apply to the residential integration of other vulnerable groups, such as disadvantaged ethnic or class groups.40 Furthermore, given its entanglement with these other axes of social fragmentation,41 even partisan residential integration may interfere somewhat with vulnerable groups’ strong associative interest in residential self-­segregation. These associative costs to vulnerable groups constitute a very serious—indeed, arguably the most serious— challenge to integration. The first thing to say in reply is that, perhaps counter-­intuitively, these costs apply more readily to token forms of integration than to the thoroughgoing integration I am advocating. Shelby acknowledges that many of the positive benefits he attaches to black self-­segregation do not strictly speaking require living in an overwhelmingly or even predominantly black neighbourhood. Rather, what they really require is living in a neighbourhood that possesses a ‘critical mass’ (20 to 50 per cent) of black Americans. The existence of a critical mass would, he suggests, allow black Americans to live alongside many who share their cultural practices or group-­specific experiences, to access businesses or associations that cater to their group-­specific tastes and preferences, and to exercise empowering political

38  Shelby (2016, 59). See also Young (2000, 216–17) and Brownlee (2019, section 6). 39  Shelby (2016, 55–6, 61–2). 40  Young (2000, ch. 6). 41  See Section 2.2, this chapter, for discussion. As discussed in Chapter 6, Achen and Bartels (2016, ch. 9) argue that partisan attachments are often driven by other social group identities, such as race, class, and ethnicity.

The Problem of Fragmentation  201 solidarity.42 We might add that the existence of such a critical mass could also give black Americans some measure of refuge from, and support in the face of, white hostility and discrimination. There is tentative evidence for this last point. Douglas Massey and Nancy Denton find that, while black Americans tend to be very reluctant to enter all-­white neighbourhoods, they are generally comfortable entering majority-­white neighbourhoods provided there is a substantial minority of black Americans.43 In this light, thoroughgoing forms of integration may avoid the associative costs Shelby cites to a significant degree. The reason Shelby ends up defending self-­segregation is that, partly because of white flight, black Americans often lack the opportunity to live in mixed neighbourhoods containing a critical mass of black people. So they are left with a dilemma: either they integrate in a token or isolated way, which leaves them as a very small minority in an overwhelmingly white neighbourhood; or they live in an overwhelmingly black neighbourhood.44 Of these two options, only the latter provides the associative benefits Shelby highlights. But the point of robust or thoroughgoing integration just is to overcome this dilemma by providing a new option. As we saw in Section 3.1, by curtailing social groups’ opportunity to dissociate from one another, integrative policies can create a new associative opportunity. In this case, the opportunity is to live in a racially mixed neighbourhood, containing (provided the integration is robust enough) a large minority, or critical mass, of black Americans.45 However, this first response only partly defuses the objection. Even if robust integration creates mixed neighbourhoods containing a critical mass of the vulnerable group, this is unlikely fully to obviate the associative costs Shelby worries about. At least in the early stages of integration, disadvantaged minorities are likely to be exposed to some hostility and discrimination from the dominant out-­group. Far from denying this, Anderson suggests that this is in fact a prediction of her diagnosis of the wrongs of segregation: since segregation is a major cause of intergroup dislike, we should expect initial interactions in newly ­integrated settings to ‘engage racially stigmatizing ideas, manifested in various forms of discrimination and unhappy interaction’.46 What is more, minority group members who have only recently settled in a mixed neighbourhood may not yet have had the chance to develop the in-­group support network and the in-­group political solidarity that would otherwise provide refuge or protection from this conflict.

42  Shelby (2016, 61). For related observations, see also Anderson (2010, 133–4) and Fiss (2003, 33). 43  Massey and Denton (1993, 89–90). 44  Shelby (2016, 61). 45 This does raise the question of how creating neighbourhoods with a critical mass of black Americans would be possible in the United States as a whole, since black Americans only represent 13 per cent of the American population. The answer, which I develop in Section 4.1, is that not all neighbourhoods need to be integrated. 46  Anderson (2010, 182).

202  Democratic Speech in Divided Times An important cost therefore remains. Creating robustly mixed neighbourhoods generates a valuable opportunity for intergroup association while making room for significant forms of in-­group interaction and in-­group solidarity. But in joining such neighbourhoods, vulnerable groups are nevertheless sacrificing at least some of the benefits that they would usually derive from residential self-­segregation. It is impossible to say in the abstract whether this cost is worth undertaking. Yet there are at least prima facie reasons to think that, in the case of racial integration in the United States, it may potentially be. In American Apartheid, Massey and Denton report that black Americans generally prefer living in racially mixed neighbourhoods, where black people constitute a critical mass rather than an overwhelming majority. Building on their work, Esther Havekes and her colleagues have since provided further evidence for this result. They notably find that, on average, black Americans’ ideal preference is for living in a neighbourhood where slightly over a third of residents are black. However, because black Americans typically lack the opportunity to live in such mixed neighbourhoods, they frequently end up living in far more segregated neighbourhoods.47 These results are significant for our purposes. They suggest that, despite the above-­ mentioned cost to racial minorities of residing in mixed neighbourhoods, black Americans on average value the opportunity to live in such neighbourhoods more highly than the opportunity to live in a segregated neighbourhood. In this case, then, we have a prima facie reason to believe that the associative benefits of residential integration for vulnerable groups may outweigh its associative costs to these groups. Hence, even in its stronger and more specific form, the freedom of association critique of integration does not seem decisive. But we should not rest content with this observation. For all that I have said, the associative costs to vulnerable minorities involved in residential integration remain serious. Even if (some) vulnerable groups do in the aggregate think that these costs are worth undertaking, it seems deeply regrettable for the rehabilitation of inclusive political discourse to come at such a great moral price. Moreover, some members or sub-­communities of these vulnerable groups surely disagree that these costs are worth undertaking, all things considered. It would be troubling, to say the least, for them nonetheless to be forced to engage in residential integration. Consequently, unlike most defences of integration,48 we should go beyond the claim that the benefits of residential integration outweigh its costs. Even if (as I have argued) that may well be true, we must also investigate as far as possible how 47  See, e.g., Massey and Denton (1993, 89–91), Havekes et al. (2016, 116–21), and Welch et al. (2001, 69). 48  Anderson (2010, 380), in particular, responds to the concerns outlined in this section by insisting that ‘all great ideals of justice require enduring struggle even in the face of repeated disappointment’.

The Problem of Fragmentation  203 the costs involved in integrating public discourse could be lessened. The systemic understanding of public discourse can be of assistance here.

4.  Integrating a System It is tempting to think of public discourse as one immense conversation. Yet, as we have seen throughout this book, this is a misleading picture. Inclusive public discourse does not occur in one immense arena that includes everyone. Rather, it takes the form of a system: a network composed of very many arenas or ‘nodes’ of discussion. These arenas vary significantly in size, in function, and in formality. Moreover, depending on how much their constituencies overlap and on where they are located, they also vary in how closely connected they are to each other.49 Thinking of public discourse in terms of a system, rather than of a unitary arena, has already proven helpful in numerous ways. It highlights how group-­ based dogmatism can play a fruitful epistemic role in democratic public ­discourse.50 It shows that some discursive arenas are governed by less stringent norms than others, such that they do not demand too much of distrustful or antagonistic participants.51 And it reveals how political discourse, including angry political  discourse, might circulate between people who are unwilling to talk to each other.52 For related reasons, the systemic perspective on public discourse is also useful when examining the costs of integration. We have seen that the costs of residential integration can be great, even if they may be outweighed by integration’s bene­fits. The systemic view reveals that it is in fact possible to meaningfully alleviate these costs—though not, unfortunately, altogether avoid them. It does so in two main respects.

4.1  Networked Integration First, the systemic approach brings into view the possibility for networked interaction across different residential arenas. When examining the implications of mutual ill will for public speech in Chapter 5, we introduced the notion of networked discourse. Even if A and C are unwilling to talk to each other, there may nonetheless be an intermediate node in the discursive network, B, to whom they are both willing to talk. If so, then A can speak to B, who then relays A’s insights and arguments to C, and vice versa. By 49  Chapter 1, Section 4; Chapter 2, Section 5; Chapter 5, Section 3; Chapter 6, Section 5. 50  Chapter 6, Section 5. 51  Chapter 2, Section 5; Chapter 5, Section 3. 52  Chapter 2, Section 4.3; Chapter 5, Section 3.

204  Democratic Speech in Divided Times means of this networked or mediated engagement, insights and arguments generated by A and C can readily circulate between them. This point is actually not specific to cases where A and C are unwilling to talk. Networked discourse also helps in circumstances where they are unable to talk, because they lack opportunities to do so. Here too, provided there is an agent B who has opportunities to talk to both A and C, B can mediate between A and C. We can apply this last insight to the problem of residential fragmentation. Suppose two neighbourhoods A and C are spatially distant from one another, such that inhabitants of A and inhabitants of C have few opportunities to interact. From a systemic perspective, this need not be overly problematic. Provided there exists a neighbourhood B that is close to A and close to C, inhabitants of B can mediate between inhabitants of A and C. Neighbourhood B, one might say, serves as an indirect point of contact, or interface, between the two neighbourhoods. Though straightforward, this insight has an important upshot: re-­establishing meaningful political exchanges between spatially fragmented groups (e.g., between black and white Americans) does not necessarily require integrating all neighbourhoods. The possibility for networked discussion makes this unnecessary. Rather than integrating all neighbourhoods, we can instead focus on creating a  substantial number of integrated neighbourhoods, where these integrated neighbourhoods then serve as an interface between communities that remain spatially segregated. The idea is roughly the following. Within integrated neighbourhoods, different social groups can interact and expose one another to diverging insights and challenges to a greater extent than in non-­integrated neighbourhoods. At the same time, inhabitants of integrated neighbourhoods can also connect to members of their respective social groups who live in nearby, non-­integrated, neighbourhoods. Thus, integrated neighbourhoods can promote intergroup exchanges not only within their boundaries, but also beyond them: if integrated neighbourhoods are arenas of intergroup exchange, and if they are in turn connected to more segregated neighbourhoods, then they can ferry insights and challenges between segregated neighbourhoods. This networked approach to residential integration is promising for our purposes. It highlights a way of promoting political exchanges across different social groups that nevertheless makes room for some degree of residential self-­ segregation. If only some neighbourhoods need to be integrated—albeit far more neighbourhoods than currently tends to be the case—this leaves space for members of vulnerable minorities who prefer residential self-­ segregation to stay among their own. So, networked residential integration alleviates the costs diagnosed in Section 3.2. This proposal is not merely a theoretical possibility. There is evidence that integrated neighbourhoods can in fact play the mediating role I have outlined. For one thing, in his groundbreaking investigation of neighbourhood effects in

The Problem of Fragmentation  205 Chicago, Robert Sampson finds abundant evidence of ‘spillover’ effects, whereby the properties of one neighbourhood influence spatially proximate neighbourhoods. ‘Opportunities for contact increase with physical propinquity,’ Sampson observes, ‘and social interactions may in turn concatenate along chains of contact, with their influences ultimately felt away from the geographic point of origin.’53 What this general finding suggests, for our purposes, is that the intergroup engagement taking place within integrated neighbourhoods may be felt beyond those neighbourhoods. In particular, ‘chains of contact’ spreading out from integrated neighbourhoods may expose non-­ integrated neighbourhoods to the insights or concerns generated via intergroup exchanges. Sampson’s examination of informational networks between elites across different neighbourhoods of Chicago lends tentative support to this suggestion. Spatial proximity, Sampson finds, is a strong determinant of inter-­neighbourhood informational networks. Accordingly, while there tend to be very few elite information ties between spatially segregated black and white neighbourhoods, racially mixed neighbourhoods seem partly to alleviate this problem. Perhaps because they constitute a spatial point of contact between otherwise segregated neighbourhoods, many of Chicago’s racially mixed neighbourhoods (for instance, the Near West Side, the Near North Side, and the Loop54) also constitute some of the most vibrant inter-­neighbourhood hubs of information.55 This accords with the picture I have been putting forward: in virtue of their informational ties to other, less integrated, neighbourhoods, integrated neighbourhoods can serve as mediating nodes between them. Yet my proposal for networked integration might raise the following concern. If only some neighbourhoods are integrated and they serve as intermediaries between otherwise segregated neighbourhoods, there will presumably be a lower density of intergroup communication than if all neighbourhoods were integrated and involved intergroup communication. Instead of every residential node in the system being directly involved in intergroup communication, only some nodes will be. Thus, the strategy I am proposing for mitigating the costs of residential integration might seem to reduce the value—for instance, the epistemic value—of public discourse. There are, in fact, reasons for optimism here. In his pioneering work on the epistemology of networks, which we encountered in Chapter 6, Kevin Zollman finds that limiting the density of communication between groups or agents need not be epistemically harmful for the system as a whole. In conditions where agents or groups are not dogmatic, limiting the circulation of information 53  Sampson (2012, 239; see also 243–6). 54  The Loop and the Near North Side have grown less integrated in recent years. But they were relatively integrated in the late 1990s and early 2000s, which is where Sampson’s data about elite information networks is drawn from. 55  Sampson (2012, esp. 342–50).

206  Democratic Speech in Divided Times

Figure 7.1  The cycle, the wheel, and the complete graph

between them can actually be a good thing. That is, systems involving relatively few or moderate connections between non-­dogmatic agents tend to perform better, epistemically speaking, than systems where everyone speaks to everyone. We can put this more schematically. Where agents or groups are not dogmatic, systems structured like a cycle and like a wheel (where nodes represent agents or groups, and segments represent flows of information) significantly outperform the complete graph, where all nodes in the system are connected to all others (see Figure 7.1).56 So, while it is crucial that there be meaningful circulation of information between key agents or groups in the system—such that severe intergroup fragmentation remains deeply problematic for political discourse—this does not mean that more circulation is always better. In some cases, it might even be better for there to be meaningful yet restricted communication. The case where people are dogmatic might seem more problematic. We saw in Chapter  6 that groups often process information dogmatically, and that—on Zollman’s analysis—the more dogmatic people are, the more important it is to have dense networks of communication between them. But even here, introducing some limits to intergroup communication need not come at a very high cost. Zollman does indeed show that systems with low levels of communication (such as the cycle) perform much worse than other systems when they are populated by dogmatic agents. This is the problem we diagnosed in Chapter 6. Yet, even at significant levels of dogmatism, epistemic systems involving moderate levels of communication (such as the wheel) perform nearly as well as systems with very dense networks of communication (such as the complete graph). It is only at the very highest levels of dogmatism on Zollman’s scale that a substantial gap appears between the two.57 So, Zollman’s analysis suggests that we can sacrifice some density of intergroup communication without sacrificing much in the way of epistemic performance. Amidst low levels of dogmatism, constrained intergroup communication might 56  Zollman (2010, 27–8).

57  Zollman (2010, 30–1).

The Problem of Fragmentation  207 actually be better than extremely dense intergroup communication. And even at moderate or high levels of dogmatism, moderately dense communication networks can still perform nearly as well as extremely dense ones. The concern at hand therefore does not seem conclusive. Integrating some but not all neighbourhoods, as I have recommended, may result in a smaller density of intergroup communication than integrating all neighbourhoods. But this need not significantly impair the epistemic value of public discourse, and it may— depending on the degree of group-­based dogmatism in the system—sometimes even be epistemically beneficial. Let us summarize the argument so far. In highlighting the possibility for networked integration, the systemic perspective on public discourse reveals a first way of lessening the costs of residential integration. Integrated neighbourhoods can serve as mediating nodes in the political system. Because this mediating function allows information to circulate in an epistemically fruitful way to and from segregated neighbourhoods, it obviates the need for all neighbourhoods to  be integrated. So, it creates some space for the voluntary self-­segregation of vulnerable minorities.

4.2  Differentiated Integration The first ‘systemic’ strategy for mitigating the costs of residential integration focuses exclusively on residential forms of integration, and explicitly depends on the existence of some measure of residential integration. But the systemic approach to public discourse can help in a second way. This perspective emphasizes that the discursive system contains highly differentiated types of arenas or nodes. Though some of these (like neighbourhoods) are residential in nature, others are not. What this suggests is that there may be ways of integrating public discourse that do not depend on residential integration—and so, that avoid the distinctive associative costs identified in Section  3.2. Pushed to an extreme, this line of thought might even seem to make residential integration altogether unnecessary: we could, one might think, integrate non-­residential arenas of public discourse instead. I will argue for a more qualified version of this second strategy: while non-­ residential forms of integration cannot altogether supplant residential integration, they can help lessen the emphasis placed on residential integration. Indeed, some non-­residential forms of integration can facilitate inclusive public discourse without requiring estranged groups to live together. As we will see, however, the more extreme suggestion that we ought exclusively to engage in non-­residential integration ultimately underestimates the extent to which spatial proximity matters for intergroup engagement.

208  Democratic Speech in Divided Times What forms of integration might help re-­establish intergroup exchanges without requiring that different groups live together? A first suggestion is workplace integration. The workplace, more than the residential arena, tends to be governed by authoritatively enforced norms of cooperation and civility. As a result, Mutz sees the workplace as a better place to foster intergroup political discussion. And indeed, she finds that people engage more readily with opposite political viewpoints in the workplace than they do in their neighbourhoods, places of worship, or voluntary associations.58 In a similar vein, Shelby cites workplace integration as a potential alternative to residential integration.59 The problem is not that workplace integration is not desirable. After all, we have already seen in Section 2.2 that proponents of the integrative solution typ­ic­ al­ly advocate policies (such as affirmative action policies) aimed at promoting integration in the workplace. Rather, the problem is that workplace integration itself depends heavily on residential integration. Access to jobs is strongly influenced by where one lives. For example, where one lives affects what openings one hears about and profoundly determines the ease or difficulty with which one can travel to particular jobs. Hence, residential segregation is widely held to be a major cause of workplace segregation.60 By implication, it will be very difficult to achieve workplace integration if we leave untouched patterns of residential segregation. This is why, in her comprehensive integrative solution, Anderson insists that spatial integration is an important prelude to workplace integration.61 Workplace integration is therefore unhelpful in our current context: because of its dependence on residential integration, it hardly helps to avoid the costs involved in residential integration. What we need, then, is a type of non-­residential arena whose composition does not depend significantly on prior residential patterns. To this end, one strategy might be to create artificial arenas for political discussion that are specifically designed to transcend existing patterns of residential fragmentation. The most influential example of this strategy is of course James Fishkin’s deliberative polling initiative. Generally, deliberative polls bring several hundred randomly selected participants together to deliberate, in collaboration with balanced panels of experts, on topics of public concern. Because participants are randomly selected, they typically come from diverse groups, including from groups that would normally be spatially and socially fragmented. Partly as a result of this diversity, the deliberation fostered by these experiments has met with significant success. It tends to yield epistemically superior judgments, generally does not give rise to belief polarization, and can even help to

58  Mutz (2006, 28–9). 59  Shelby (2016, 71). 60  See, e.g., Young (2000, 198; 2011, 43–52), Anderson (2010, 27–8), and Fiss (2003, 12). 61  Anderson (2010, 116–17).

The Problem of Fragmentation  209 reduce affective polarization between divided groups.62 The creation of such artificial arenas of inclusive discourse should therefore vigorously be encouraged. Nevertheless, this approach can at best only play a quite limited part in the project of integrating public discourse. First, deliberative polls remain small-­scale initiatives.63 Consequently, even if thousands of deliberative polls were conducted, they would still only involve a small portion of the population in intergroup discussion. Second, deliberative polls are episodic. They normally last no more than two days. So, even for the lucky few who do end up being included in deliberative polls, these discursive arenas only represent a very small part of their political talk.64 The shortcomings of the two approaches considered so far suggest a potential dilemma for non-­residential integrative strategies: either their success depends crucially on prior residential integration; or they are compatible with existing ­patterns of residential fragmentation, but at the cost of being quite limited interventions. Yet one category of arenas, which I have neglected up until now, may escape this dilemma to an important degree: online arenas. It may seem surprising, on the face of it, to look to online speech for more inclusive discourse. After all, some of the worst forms of partisan fragmentation take place online. In particular, social media platforms such as Facebook and Twitter tend to be highly segregated by party. This is partly because people disproportionately select like-­minded users as their contacts on social media platforms. But, as we will see below, it also results from the architecture of social media spaces, which is often designed to make like-­minded perspectives especially visible to users.65 Even so, online spaces—and social media platforms in particular—constitute a promising non-­residential arena in which to implement integrative policies for several reasons. The first reason is that a very significant proportion of people’s political exchanges occur on social media. The case of Facebook, which brings together nearly 70 per cent of the American adult population, is perhaps the most notable. As Jaime Settle has shown, 62  For overviews of deliberative polls, see Fishkin and Luskin (2005), Luskin et al. (2014), and Fishkin (2018). 63  On this point, see also Chapter 5, Section 4. 64  Admittedly, Fishkin (2018, 175–80) has explored attempts at ‘scaling up’ deliberative polls. In particular, he and Ackerman envision a ‘Deliberation Day’, whose design closely parallels that of deliberative polls, and where the entire adult population of the United States is meant to deliberate about upcoming presidential elections. But this proposal faces two difficulties. First, even if it were feasible, it would solve the problem of scale only by exacerbating the problem of frequency. Given its financial cost and logistical difficulty, Deliberation Day is only intended to happen every few years, and would only last a few days. Second, as Fishkin (2018, 176) himself acknowledges, scaling up deliberative polls reintroduces problems related to residential fragmentation: the residential fragmentation of social groups makes it extremely difficult to involve every adult citizen in a socially diverse deliberative group over two days. 65  For analyses of ideological segregation on social media, see, e.g., Bakshy et al. (2015), Sunstein (2017, 117–24), Heatherly et al. (2017, 1284), and Settle (2018, 162).

210  Democratic Speech in Divided Times a far larger group of Americans—with far fewer systematic biases shaping its composition—report being regularly engaged with their [Facebook] News Feed than report watching cable news or reading a newspaper regularly. And when they do, they cannot avoid encountering political content. [. . .] [P]eople report seeing political information half the time they scroll through their News Feed, even when they are not looking explicitly to find it.66

Thus, unlike the political communication involved in deliberative polls,67 pol­it­ ical communication on social media affects an overwhelming majority of the population, and affects them regularly over time. The second reason is that, since social media interactions take place in virtual space, they have unparalleled freedom from physical constraints.68 This freedom is crucially important for our purposes. To begin, because social media spaces are relatively free of physical constraints, they yield unprecedented opportunities for connecting to members of other groups who, in spatially fragmented contexts, may be geographically distant. This is not just a hypothetical possibility. Social media sites have in fact enabled people to maintain far larger networks of ‘weak ties’ than ever before. For ex­ample, Facebook not only allows users to maintain a standing connection to distant acquaintances and to discover their political views, it also connects users to the friends and acquaintances of their friends and acquaintances.69 The fact that social media facilitates greater connections to weak ties is significant. As we saw in Section 3.1, people’s weak ties are more likely than their close ties to come from different groups and to possess different perspectives than them.70 Unsurprisingly, then, people’s social media networks—and, in particular, their Facebook networks—tend to be more socially and politically heterogeneous than their offline networks.71

66  Settle (2018, 13–14; see also 32, 43–4). 67 Note that there are also initiatives aimed at extending deliberative polling online. See, e.g., Fishkin and Peck (2020). However, like their offline counterparts, these online initiatives remain limit­ed in scale and frequency. Other online deliberative experiments involve creating more permanent deliberative spaces, where people can spontaneously come to discuss political issues that they are interested in. See, e.g., Sunstein (2017, 216–17). While these initiatives might be greater in scale and less episodic than deliberative polls, they generate a different problem. Specifically, they suffer from important selection effects: some groups of people (namely, those who are politically engaged) are likely to participate far more than others. By contrast, the reason social media platforms such as Facebook succeed in exposing such a large and diverse range of people to political content is that—as Settle explains—many users encounter political content inadvertently, while they consult their News Feed for other reasons. 68  Settle (2018, 25). 69  Settle (2018, 29, 67, 73). 70 Heatherly et al. (2017, 1273) and Settle (2018, 28) make similar observations in the online context. 71  See, e.g., Settle (2018, 73, 95, 184), Barnidge (2017, 312), Heatherly et al. (2017, 1272).

The Problem of Fragmentation  211 At first sight, the comparative heterogeneity of people’s social media networks seems to contradict the common observation that social media fosters homogeneous encounters within insulated echo chambers. But there is no ­contradiction. The key to reconciling these two observations is that the architecture of social media sites tends to conceal the actual heterogeneity of people’s networks. In the case of Facebook, at least three mechanisms contribute to this effect. First, the algorithms that determine the content of users’ News Feed prioritize content that the user is likely to approve of over content that they are likely to disapprove of.72 Second, the visible quantification of reactions and feedback, combined with the fact that users are more likely to comment on things that they agree with, leads users to overestimate how many people in their network agree with their own posts or with the (overwhelmingly like-­minded) posts that appear on their News Feed.73 And third, when algorithms do allow alternative perspectives onto a user’s News Feed, these tend to be relatively extreme perspectives. This is, in part, because extreme posts elicit stronger interest and more numerous reactions.74 A  ‘false consensus’ effect therefore emerges: the visible architecture of Facebook leads people to believe that their perspectives are nearly ubiquitous and, relatedly, that only fringe ‘extremists’ could disagree with them. This brings us to the final reason why, given its freedom from physical constraints, social media sites constitute a promising alternative arena in which to implement integrative solutions. The architecture of social media sites conceals their actual heterogeneity. But this architecture, being virtual rather than phys­ic­al, can be changed on a large scale without upsetting existing residential patterns. Given the diagnosis offered above, possible architectural changes aimed at promoting integrated political communication on Facebook might include the following: altering the News Feed algorithm so that it promotes more content that users are likely to disagree with;75 adding ‘reaction’ options that give users the opportunity readily to express non-­confrontational or constructive forms of dis­ agree­ment (such as ‘respectfully disagree’ or ‘please provide a validating source’);76 eliminating or reducing the visibility of quantified social reactions;77 and adjusting the News Feed algorithm so that it promotes more moderate content than is currently the case.78 By means of such reforms, we might hope to make the diversity of people’s Facebook networks more visible to them, and, relatedly, to promote greater engagement with that diversity. 72  Sunstein (2017, 14–16, 122–4); Settle (2018, 35–6). 73  Settle (2018, 97, 162, 185–6). 74  Settle (2018, 134). 75  Relatedly, Sunstein (2017, 232–3) suggests introducing ‘opposing viewpoint buttons’ (which, if clicked, expose users to politically uncongenial content), as well as ‘serendipity buttons’ (which expose users to random or unanticipated materials). 76  Settle (2018, 252); Lynch (2019, 48–9). 77  Settle (2018, 253). 78  Settle (2018, 251).

212  Democratic Speech in Divided Times Thus, integrating online speech, particularly on social media, seems a promising way of circumventing residential integration and its attending costs. Unlike deliberative experiments, online speech accounts for a very large portion of people’s political exchanges. Moreover, because online speech occurs in virtual space, patterns of online discourse are, to a significant degree, independent of residential patterns—more independent, at least, than opportunities for workplace integration. This independence matters greatly. It gives people unprecedented op­por­tun­ ities to maintain connections with residentially distant out-­groups. And it means that the architecture of online spaces can be altered to promote regular exposure to these connections without upsetting existing residential patterns. Like networked integration, online integration can therefore lessen the burden placed on residential integration. Residentially self-­segregated communities can be exposed to spatially distant out-­groups not just via the mediation of integrated neighbourhoods, but also via well-­designed social media platforms. By integrating online political discourse, we can therefore give vulnerable groups more leeway to engage in residential self-­segregation. The promise of online integration does not, however, mean that we can fully jettison residential integration when promoting inclusive public discourse. There are several reasons for this. One might worry that some benefits of inclusive political discourse, such as the creation of mutual goodwill discussed in Chapter 5, can only arise through face-­to-­face political discourse.79 Alternatively, one might think that, although social media interactions give users unprecedented op­por­ tun­ities to connect to spatially distant out-­groups, whom users end up encountering still depends meaningfully on their residential location. But even setting these concerns aside, the main reason for not relying solely on online integration is the so-­called digital divide. A large segment of the population (as many as 25 per cent in the United States) remain deprived of Internet access in their homes. This deprivation disproportionately affects poor, non-­ white, elderly, and rural segments of the population.80 An approach to integrating political discourse that focuses solely on online arenas, to the exclusion of spatial integration, is therefore bound to result in insufficiently inclusive and diverse political exchanges. We are left with a qualified conclusion. The systemic view of public discourse highlights the fact that integration can be differentiated between spheres, some residential and others not. This can lessen the emphasis placed on residential integration—and thereby alleviate the costs residential integration imposes on vulnerable groups. But it cannot justify abandoning residential integration al­together. Spatial proximity continues to matter for public discourse.

79  Settle (2018, 212). For a more optimistic view, see Amichai-­Hamburger and McKenna (2006). 80  White House (2013).

The Problem of Fragmentation  213

5. Conclusion Spatial and social fragmentation poses a deep threat to inclusive public discourse. By keeping different social groups apart, by fuelling their mutual dislike, and by amplifying intergroup inequalities, it deprives people of meaningful op­por­tun­ ities to engage with members of other groups. This, as we have seen, undercuts the value of public discourse in a way that risks giving rise to further fragmentation. To tackle this problem, we must adopt wide-­ranging integrative policies aimed at re-­establishing contact and interaction across difference. Yet these policies, particularly those aimed at reducing residential fragmentation, come at a great cost. They impose important restrictions on the associative freedom of those who must integrate. Most worryingly, these restrictions limit historically oppressed groups’ legitimate interest in living among their own. This associative cost does not, I believe, warrant giving up on an integrated public sphere. This is especially clear once we highlight, as I have done, that ­integration creates meaningful associative opportunities at the same time that it curtails others. But it is a great cost nevertheless, and any serious attempt at de-­ fragmenting the public sphere should seek to mitigate it as far as possible. Given the systemic nature of public discourse, there are at least two ways of doing so. The first involves using integrated neighbourhoods as mediating nodes between segregated neighbourhoods. The second emphasizes the opportunities, particularly the online opportunities, for transcending the physical barriers raised by residential segregation. Both strategies reveal that it is possible to reintegrate political discourse while leaving some room for disadvantaged groups voluntarily to self-­segregate. Neither, however, allows us to jettison residential integration altogether. The upshot is clear. We can and should make integration more sensitive to the legitimate associative interests of different groups—far more sensitive, in particular, than previous approaches to integration have been. But there is only so far this will go. However networked and however differentiated, the integrative project will not come cheaply. It will require time, effort, and sacrifice, including unfairly significant sacrifices from those who suffer most from existing social divisions. This is the cost—the necessary cost—of saving democratic public discourse from its present spiral of segregation, mutual dislike, and inequality.

8

Resisting Political Minimalism 1.  Introduction: Political Minimalism 1.1  The Temptation of Political Minimalism When governed by appropriate norms, the public speech of divided democracies offers promising resources for pooling knowledge and thereby promoting accountable policymaking. Yet, as we have seen in previous chapters, infelicitous background conditions may reduce its effectiveness at achieving these aims. While I have suggested that some of these threats are not as great as it is some­ times thought, they do pose important challenges. In particular, when people are preyed upon by campaigns of political misinformation,1 when they are so dis­ trustful that they refuse to listen to one another,2 and when they are dogmatically committed to defending their social groups’ perspectives,3 democratic public discourse may struggle effectively to educate them. In such contexts, people risk ending up ignorant about politics, in a way that limits their ability to hold policy­ makers accountable, and further impairs their ability to engage in fruitful pol­it­ ical discussion.4 Indeed, politically ignorant people may struggle to distinguish good arguments from spurious ones, and to distinguish trustworthy sources of testimony from untrustworthy ones. We have encountered some ways of mitigating these challenges. ‘Positive’ forms of counterspeech can, to some degree, shield or inoculate people from political misinformation.5 As for the problems posed by distrust and dogmatic group cognition, their seriousness may diminish in light of the ‘systemic’ charac­ ter of public discourse. In a discursive system containing a network of intercon­ nected arenas, third parties can mediate between distrustful agents who discuss issues in distinct arenas.6 What is more, dogmatic group cognition can actually play a positive epistemic role when it takes place in a well-­integrated discursive system.7 The problem, however, is that the last two solutions depend on the existence of a discursive system that is well integrated—that is, where information circulates readily (if indirectly) between different arenas of discussion. But in deeply divided 1 Chapter 4. 2 Chapter 5. 3  Chapter 6, Section 5. 4  Chapter 6, Section 1. 5  Chapter 4, Section 4. 6  Chapter 5, Section 3. 7  Chapter 6, Section 5.

Democratic Speech in Divided Times. Maxime Lepoutre, Oxford University Press (2021). © Maxime Lepoutre. DOI: 10.1093/oso/9780198869757.003.0009

Resisting Political Minimalism  215 democracies, the public sphere is often not integrated but fragmented. Different groups live, work, and talk in insulated arenas that are hardly connected, directly or indirectly, to each other. Here too, this is not an insurmountable problem. In the last chapter, I explored morally permissible ways of de-­fragmenting the public sphere. Nevertheless, this task will be laborious. It will require, as we have seen, significant effort, cost, and time.8 As a result, some might lose patience with the ideal of inclusive public dis­ course. A critic might concede that, once robust and wide-­ranging integrative policies have been enacted, inclusive public discourse will indeed be epistemically valuable and accountability-­promoting. But they might nonetheless insist that, in the meantime, it is not. In the meantime, public discourse will struggle to educate people; and their resulting political ignorance will not only limit their ability to hold political power accountable, but it will also further impair the quality of public discourse. So long as this condition persists, the objection continues, we should prefer a more minimalistic political ideal to the broadly deliberative, or discourse-­oriented,9 ideal that I have been defending so far. In other words, we should prefer an ideal that eschews inclusive and large-­scale public discourse. This normative approach to politics is minimalistic in the sense that it does not require the large-­scale participation of ordinary people in political discussion; and indeed, as we will later see, it sometimes actively opposes this large-­scale participation. In sum, one might think of the ‘political minimalism’ objection as making a point about urgency. It holds that, even though the deliberative ideal of inclusive public discourse may become action-­guiding and relevant in the future, it is not yet relevant. Here and now, more minimalistic political arrangements would do better. The purpose of this chapter is to stave off this objection.

1.2  From Robust Political Ignorance to Political Minimalism For the purposes of this chapter, then, let us make two pessimistic assumptions. Let us assume that we find ourselves in highly non-­ideal conditions, where pol­it­ ical misinformation, mutual distrust, group-­based dogmatism, as well as spatial and social fragmentation have combined to make people highly ignorant about 8 Chapter 7. 9  I say ‘broadly’ deliberative because, as discussed in the Introduction, public discourse is strictly speaking broader than public deliberation (understood as the give and take of reasons relating to prac­ tical matters). Having said that, some deliberative democrats use ‘deliberation’ loosely, such that it includes most of what I include under ‘discourse’. See, e.g., Krause (2008). In line with previous uses, I will use ‘deliberative democracy’ synonymously with ‘discourse-­oriented democracy’ in this chapter. The core tenet, and what matters for this chapter, is the commitment to promoting large-­scale partici­ pation in public discourse by ordinary people who come from diverse groups.

216  Democratic Speech in Divided Times politics. And let us further assume that this political ignorance in turn undercuts the value of inclusive public discourse. This is because people end up knowing too little to distinguish good arguments from bad arguments, sound testimony from spurious testimony, and (relatedly) reliable information shortcuts from unreliable shortcuts. Thus, people are incapable of overcoming or circumventing their pol­it­ ical ignorance via inclusive public debate. For convenience, we can call these pessimistic conditions ‘robust political ignorance’. The political ignorance at hand is robust, first, in that it is upheld by other background conditions, such as distrust, social and spatial fragmentation, group-­ based dogmatism, and misinformation. And it is robust, second, in that it contrib­ utes, in turn, to making political learning via inclusive public debate impossible. My purpose in starting from the premise of robust political ignorance here is not to assert that it is always or generally true of divided democracies. Instead, I am making this assumption for the sake of argument. The objection I am consid­ ering in this chapter, which has been put forward by theorists such as Ilya Somin or Jason Brennan, holds (1) that we find ourselves in conditions of robust political ignorance, which undercut the epistemic and accountability-­promoting functions of inclusive public discourse. And from this premise, it infers (2) that we should abandon the deliberative ideal of democracy in favour of a more minimalistic political ideal—an ideal, to reiterate, that does away with the large-­scale partici­ pation of ordinary citizens in public discourse. Underpinning this inference is the belief that, because minimalistic political ideals demand less of ordinary people, they are more sensitive to conditions of robust political ignorance.10 I will argue that this belief—and, with it, the argument for political minimalism— is a mistake. Even if political ignorance (together with the distrust, fragmentation, misinformation, and dogmatism that sustain it) were so great as to substantially undercut the value of inclusive public discourse, this would not yet give us a reason to lose patience with the deliberative or discourse-­oriented ideal of democracy. This is because robust political ignorance is in fact not a special problem for this ideal. On the contrary, exploring more minimalistic political ideals reveals either that their recommendations are equally plagued by robust political ignorance, or that, although their recommendations constitute promis­ ing remedies for robust political ignorance, they can be reconciled with the deliberative ideal. The argument will proceed as follows. In the first place, I will demonstrate that more minimalistic democratic arrangements—which either eliminate inclusive discourse altogether or restrict it to small-­scale settings—neither avoid nor allevi­ ate the problems posed by robust political ignorance. And insofar as these models do appear less vulnerable to robust ignorance, this is because of features that are 10  For versions of this objection, see Anderson (1998, 490–1), Somin (1998, 438–42; 2016, ch. 2), Posner (2003, 163), López-­Guerra (2014), Bell (2016), Brennan (2016).

Resisting Political Minimalism  217 ultimately consistent with deliberative democracy (Section 2). Now, if all models of democracy are inadequate in such non-­ideal conditions, perhaps the solution is to abandon democracy. However, I will show that the main non-­democratic arrangements touted as remedies in fact merely postpone the problems posed by robust political ignorance (Section 3). The point of this companions-­in-­guilt argument is not solely negative. Its posi­ tive purpose is to refocus normative theorizing about democracy. It reveals that criticisms that move from the premise of robust political ignorance to the rejec­ tion of deliberative democracy are distracting. If alternative political ideals are equally plagued by these conditions, then these conditions do not give us any reason to depart from the deliberative ideal, with its support for inclusive public discourse. Instead, our focus should remain on tackling robust political ignorance at its source. In particular, we should not lose patience with the arduous task, outlined in the previous chapter, of tackling the social and spatial fragmentation that impedes inclusive public discourse (Section 4).

2.  Why Robust Ignorance is a Problem for Minimalist Democracy The first version of the political minimalism objection holds that, in conditions of robust political ignorance, we should adopt a more minimalistic ideal of democracy—an ideal of democracy that, unlike deliberative or discourse-­oriented democracy, does not involve large-­scale public discussion between ordinary ­people from diverse groups. If robust political ignorance undermines the value of inclusive discourse, more minimalistic forms of democracy seem as though they might be more adapted to such conditions. Against this influential argument, the present section will demonstrate that the most promising minimalistic democratic models that have been advanced to ­circumvent the difficulties raised by robust political ignorance fail to do so.

2.1  Elite Democracy and Robust Ignorance To avoid the problems robust political ignorance poses for inclusive public discourse, many commentators advocate a highly minimalistic model of democracy—elite democracy—in which ordinary citizens play a smaller part. Elite democracy is a resolutely representative form of democracy. Citizens do not vote on every policy decision. They merely elect more competent political representatives—elites— and throw them out in subsequent elections when they perform poorly. Elite democracy therefore staunchly opposes direct democracy. But this is not what distinguishes it from the deliberative ideal of democracy. Deliberative democrats usually take inclusive public discussion to be intimately

218  Democratic Speech in Divided Times bound up with political representation. That is, a significant part of what citizens are expected to exchange reasons about is their actual or potential representa­ tives. In a representative deliberative democracy, citizens debate which repre­ sentatives they should elect; they argue about whether their elected representatives are enacting good policies; they write to those representatives to demand justifi­ cations; and they publicly criticize representatives who break electoral promises. Accordingly, as we saw in Chapter 6, some of the most sophisticated theories of political representation have emerged out of deliberative theories of democracy.11 Instead, then, elite democracy’s central difference with deliberative democracy is that ordinary citizens are not required to engage in public discussion with each other and with their representatives. The only essential democratic act is voting for representatives.12 This elite theory of democracy has most famously been championed by Joseph Schumpeter and neo-­Schumpeterians such as Richard Anderson, Richard Posner, Jeffrey Green, and Ian Shapiro.13 On the Schumpeterian account, self-­interested elites compete for office. Like firms targeting consumers, they develop packages of policies and attempt to ‘sell’ them to voters. Citizens then vote for the elites who seem most likely to address their concerns, shunning elites whose policies seem undesirable and replacing those who have proven incompetent in power. The competition for office is intended to guarantee that elites will offer adequate policy packages. Likewise, the threat of being voted out of office motivates elected elites to keep their promises. Besides this minimal role, however, citizens should be largely uninvolved in politics. According to Schumpeter, citizens ‘must understand that, once they have elected an individual, political action is his business and not theirs’.14 This passivity extends to public discourse. Richard Anderson, for instance, claims that  or­din­ary citizens do not need to deliberate.15 But elite democrats often go further. Schumpeter positively demands that citizens refrain from writing to their representatives.16 Even more strikingly, Green affirms that democracy should be ‘ocular’: the role of ordinary people is ‘characterized by silence rather than decision, spectatorship rather than activism’.17

11  Chapter 6, Section 3.1. 12  Elite democracy’s rejection of direct democracy does distinguish it from another non-­deliberative theory of democracy: aggregative democracy. Like elite democracy, aggregative democracy empha­ sizes the aggregation of votes over inclusive discussion. But aggregative democracy is consistent with direct democracy, where citizens vote on all issues. 13 Schumpeter (2010 [1942]); Anderson (1998); Posner (2003); Green (2010); Shapiro (2016, 78–83). 14  Schumpeter (2010 [1942], 262). 15 Anderson (1998, 494). Relatedly, Bagg (2018b, 902) suggests that we should de-­prioritize reforms aimed at promoting greater citizen participation in deliberation. 16  Schumpeter (2010 [1942], 262). 17  Green (2010, 189), emphasis added.

Resisting Political Minimalism  219 In sum, elite democracy refrains from promoting large-­scale public discussion by ordinary people. This means that, typically, it does not recommend institu­ tional reforms aimed at getting ordinary people to deliberate and discuss political issues with others (such as the integrative policies I defended in Chapter 6). And, in its purest form, it actively discourages this kind of discursive participation. Schumpeter, Posner, and Anderson all take robust political ignorance as a key motivation for adopting elite democracy. And although they do not explicitly dis­ cuss ignorance, Shapiro and Green prefer elite democracy because they find it more sensitive to non-­ideal conditions than deliberative accounts.18 There is something intuitively appealing about this. Prima facie, if ordinary people are so politically uninformed that their debates and discussions are unlikely to be help­ ful, it makes sense to adopt a model of democracy that is sensitive to these limita­ tions. By jettisoning inclusive public discussion, and restricting citizens’ role to voting out bad leaders, elite democracy does this more emphatically than any other conception of democracy. An immediate concern with elite democracy is that, for all of its ambitions of realism, at least some of its strands seem out of touch with political reality. In particular, there is evidence that ordinary people already engage in significant amounts of political discourse.19 So, insofar as elite democrats require that or­din­ ary citizens refrain from engaging in public deliberation (as Schumpeter does) or conceive of ordinary citizens as silent (as Green does), it seems oddly detached from actual political practice. But there is a more fundamental problem. The problem is that, upon closer inspection, robust political ignorance also constitutes a grave problem for elite democracy. Although elites are responsible for the overwhelming majority of day-­to-­day politics, there remains a bottleneck of democratic decision-­making— elections—where ordinary citizens have the final say. Making good decisions dur­ ing elections requires various kinds of political knowledge. To compare competing politicians’ policy platforms, one must know which politicians support which policies and what the likely outcomes of different policies are. Moreover, to hold politicians accountable, one must know who plays what role in the policymaking process. These, as discussed at the beginning of Chapter 6, are precisely the kinds of knowledge that ordinary citizens are said to lack. Consequently, when individ­ uals are robustly ignorant about politics, elite democracy is problematic: there is no guarantee that politicians will offer proposals that genuinely cater to citizens’

18  Schumpeter (2010 [1942], 234); Posner (2003, 151–2); Anderson (1998, 493–9); Green (2010, 24); Shapiro (2016, 6–9). 19  Jacobs et al. (2009, 4), for example, find that ‘citizens engage in more extensive and meaningful talking than previously suspected. [Their] findings challenge the enduring tendency to attribute the ills of democracy to a lazy and withdrawn citizenry.’ See also Neblo et al. (2010). This is consistent with the observation, discussed in Chapter 7, that people seldom talk politics to the other side.

220  Democratic Speech in Divided Times interests, no guarantee that citizens will choose the best policy platforms among those offered, and no guarantee that politicians will keep their promises. One might nevertheless maintain that robust political ignorance is a greater problem for deliberative democracy than for elite democracy. For Somin, this is because theories of ‘deliberative democracy [. . .] require greater knowledge than less onerous ones such as Schumpeterian retrospective voting’.20 First, while both elite and deliberative democracy demand knowledge of the policymaking pro­ cess, of who supports what policies, and of what the likely outcome of given pol­ icies is, ‘deliberative democracy probably requires greater factual knowledge of this kind, because voters are expected to engage in serious deliberation on policy alternatives’.21 Second, because of the norms that regulate public discourse, delib­ erative democracy also requires further kinds of knowledge, ‘moral and philo­ sophical knowledge’, which elite democracy does not require. Citing Gutmann and Thompson, Rawls, and Habermas, Somin claims that deliberation involves appealing to mutually acceptable or shared considerations.22 Hence, participants must know not only which vote would promote desirable outcomes, but also which considerations are shared. Neither of these two grounds for thinking that the deliberative ideal of democ­ racy demands more knowledge is compelling. First, it is unclear why deliberators must know more about the policymaking process, policy support, and policy out­ comes than voters in elite democracy. The normative appeal of elite democracy depends crucially on voters being able to identify bad representatives and replace them with better elites. Thus, insofar as elite democracy is normatively appealing, it too requires that voters reflect seriously on competing politicians and their policy platforms. Put differently, even if voters in elite democracy do not need to engage in serious interpersonal deliberation, they still need to engage in serious intrapersonal deliberation if they are to vote judiciously. And there is no reason to think that serious intrapersonal deliberation about politicians and their platforms demands less factual knowledge than serious interpersonal deliberation about these issues. The second suggestion—that the norms of public deliberation require further kinds of knowledge—is more promising, but ultimately misses its mark against many deliberative theories of democracy, including my own. Chapter 1 observed that the norm according to which participants must appeal to shared reasons should only apply in formal discursive sites, such as the legislature. Accordingly, the vast majority of citizens need not comply with it, and need not possess the corresponding knowledge. Now, some other kind of moral or philosophical knowledge may be required of ordinary speakers. Debating political issues often involves arguing for the moral desirability of policies, which in turn requires

20  Somin (2014, 160).

21  Somin (2014, 43, emphasis added).

22  Somin (2014, 42–3).

Resisting Political Minimalism  221 knowing about moral values. But this does not seem specific to democratic delib­ eration. To vote for elites in a way that promotes morally good outcomes, one must know not only what the causal effects of one’s vote might be, but also how good those effects are. That simply is moral knowledge. Furthermore, even if engaging in fruitful public discussion did require more information than competently replacing elites, it still would not necessarily follow that robust political ignorance is a greater problem for deliberative democracy. Suppose we are considering two strategies, S1 and S2, for achieving a goal G. Both strategies require us to have information that happens to be unavailable. But S1 requires less unavailable information than S2. Even then, if the unavailable infor­ mation that S1 requires is necessary for S1 to achieve G, and if that information is genuinely inaccessible, then S1 may be just as unsuccessful at achieving G as S2. To see this, imagine that we are trying to open a door, and two passcodes are needed to do so. A only knows one of the passcodes, and B knows neither. In this case, asking A to open the door will be just as unsuccessful as asking B, even though the strategy of asking A requires less unavailable information than that of asking B. Thus, from the fact that a strategy S1 requires less unavailable information than S2, it need not follow that ignorance is a greater problem for S2 than for S1. Consequently, even if elite democrats established that deliberative democracy requires that ordinary citizens be more knowledgeable than elite democracy does, this would not necessarily settle the matter. In conditions of robust political ig­nor­ance, both might conceivably be equally ineffective. At this stage, elite democrats might take a different line of argument, and object that my characterization of elite democracy is uncharitable. I have suggested that in elite democracy, elites simply present their policy platforms, before letting citi­ zens choose. But this is arguably not all there is to elite democracy. Instead, Green claims, elites should also offer arguments for their policies. Notice that this is elite discourse, not the inclusive discourse associated with deliberative democracy: elites exchange arguments, while ordinary citizens are primarily tasked with ­listening.23 The idea is that elites’ arguments for their proposals will educate or­din­ary citizens about politics. Hence, citizens will be more knowledgeable when they elect elites. Deliberative democrats often object that such elite-­driven discourse is defi­ cient. First, Hélène Landemore worries that elites may have substantial epistemic gaps. This is because politically relevant knowledge is distributed between differ­ ent social groups, and some social groups are underrepresented among elites.24 Second, even if elites possessed all the relevant political knowledge, Simone Chambers fears that they may be ill-­motivated. They might collude to keep some 23  Green (2010, 182–6). See above, including notes 15–17, for the lack of deliberation by ordinary citizens. 24  Landemore (2013, ch. 4).

222  Democratic Speech in Divided Times ideas off of the agenda, while offering deceptive arguments aimed at making ­people accept policies that only serve elite interests.25 But let us assume that some elites know what the most morally desirable pol­ icies are, are motivated to advance them, and offer arguments for them. The main problem is this: why think that politically ignorant voters will be swayed by the arguments of wise and benevolent elites, rather than by the sophistical appeals of unwise or self-­interested demagogues? The point is that even if there are real elites, who are knowledgeable and well motivated, not all candidates for office are like this. Therefore, for elite discourse to educate listeners, ordinary citizens would need to discriminate between the sound appeals of real elites and the deceptive ones of mere sophists. However, this is precisely what conditions of robust political ignorance are supposed to preclude. Indeed, as we saw in Section 1.2, the argument for political minimalism rests on the assumption that high levels of political ignorance pre­ vent citizens from making such distinctions, which in turn keeps them from bene­fit­ing from inclusive public discourse. If it turns out that politically ignorant citizens can in fact distinguish between good and bad arguments, testimonies, and other information shortcuts, then they might be able fruitfully to engage in public debate after all. So, in appealing to elite discourse to defend elite democ­ racy, elite democrats are helping themselves to an assumption that they are deny­ ing deliberative democrats. As a final response, elite democrats might suggest that candidates for office are actually almost invariably more knowledgeable than the average citizen, and are often decently motivated. And even if some are not, they might add, anyone who is elected is then made to learn about politics and to behave accountably by their aides and colleagues. If this is the case, ordinary citizens’ robust ignorance about politics might seem relatively unproblematic for elite democracy. Because all ­possible representatives would end up being politically competent, the stakes of elections are low. No matter how robustly ignorant voters are, they cannot go too far wrong. However, this response would not give elite democracy any comparative advantage: if we grant its premises, deliberative democrats too can help them­ selves to this defence. The response under consideration is really an argument for representative democracy: it says that, in conditions of robust political ignorance, having citizens elect representatives is better than having them make policy directly. But, as we noted earlier, deliberative democracy is entirely consistent with representative democracy. Thus, if representative democracy mitigates the negative impact of robust political ignorance, this helps deliberative democracy as much as elite democracy.

25  Chambers (2009, 338–9).

Resisting Political Minimalism  223 Elite democracy responds to conditions of robust political ignorance by getting ordinary citizens to do less. However, it retains a bottleneck of democratic decision-­making, elections, where, if we truly are in conditions of robust political ignorance, then there is no reason to think that voters will choose the correct elites. Consequently, I have argued, robust and widespread ignorance about pol­it­ ics is just as problematic for elite democracy as for deliberative democracy. If, to resist this objection, elite democrats underscore the educational benefits of elite discourse, they replicate the informational burdens of deliberative democracy. And if, instead, elite democrats appeal to the dependability of representatives, this response is equally available to deliberative democrats. Perhaps, then, the proper response to robust political ignorance is not to elim­ in­ate public debate involving ordinary citizens—as elite democracy does—but rather to restrict it to more structured settings. The next section examines this popular suggestion.

2.2  Reintroducing Small-­scale Deliberation To evade the difficulties elite democracy encounters in the face of robust political ignorance, a prominent solution recommends reinstituting deliberation and debate involving ordinary people, but doing so in small-­scale settings, or ‘min­ ipublics’. According to James Fishkin and Robert Luskin, who have most prom­in­ ent­ly advanced this response, the fact that most citizens are politically ignorant and do not engage in high-­quality deliberation is ‘unlikely to change very much. Not for the public as a whole, at least.’26 But it might change for a smaller subset of the public. This proposal aims to achieve the best of both worlds. Since, as Section 2.1 discussed, the good functioning of elite democracy does suffer when ordinary people are highly ignorant about political matters, it aims to educate samples of ordinary people via small-­scale deliberation. At the same time, it promises to avoid the problems diagnosed in previous chapters with using large-­scale discus­ sion to educate ordinary people. Small-­scale discussion, it is said, can be struc­ tured so as to enable ordinary people to learn effectively. In particular, small-­scale deliberative initiatives may be effective in bringing people from different social groups together, insulating them from political misinformation, and inducing them to engage in an open-­minded and trustful manner. If this is correct, then small-­scale deliberative settings have the potential to avoid the chaos of the mass public sphere.

26  Fishkin and Luskin (2005, 286).

224  Democratic Speech in Divided Times These claims are not simply speculative. Political scientists have extensively experimented, with promising results, on the use of deliberative minipublics. Consider a few illustrations. In 2004, a Citizens’ Assembly composed of 160 ran­ domly selected citizens was tasked with deliberating regarding which electoral system British Columbia should adopt.27 Between 2010 and 2011, Iceland charged groups of randomly selected citizens with designing, via deliberative processes, a new national constitution.28 From 2016 to 2018, a randomly selected Citizens’ Assembly in Ireland examined policy issues such as abortion, climate change, and Ireland’s ageing population.29 Over roughly the same period, Citizens’ Initiative Reviews throughout the United States have asked randomly selected citizens to assess referendum proposals (for instance, proposals concerning Genetically Modified Organisms, corporate taxes, or minimum sentencing).30 But the most widely tested form of minipublic, which we have encountered in previous chapters, remains Fishkin’s ‘deliberative polling’ initiative. In one of the first deliberative polls, several hundred randomly selected citizens in Texas delib­ erated on the merits of investing in energy efficiency and renewable resources. Due to their initial success, deliberative polls have since spread across the world. In 2012, for instance, the Japanese government commissioned a deliberative poll to advise it on energy options following the Fukushima disaster.31 In 2014, a deliberative poll in Uganda explored policy options relating to resettlement, land management, and population pressure.32 And, most impressively, Mongolia adopted a Law on Deliberative Polling in 2017, which stipulates, among other things, that deliberative polls must be conducted prior to passing constitutional amendments.33 Consider how minipublics such as these realize the aims outlined above. First, participants tend to be selected at random (or at near random). This helps ensure that the resulting group constitutes a representative sample of the population, which mirrors the broader public’s demographic composition and political attitudes.34 Thus, unlike elite democracy, minipublics actively include ordinary citizens in political discussion. At the same time, unlike much of the political discussion that occurs in the larger public sphere, the composition of minipublics is socially integrated, rather than socially fragmented. Second, the groups are sufficiently small for their discussion to be carefully structured. This is particularly well illustrated by deliberative polls. Participants in deliberative polls receive balanced briefing materials prior to deliberating. Their subsequent deliberation is facilitated by trained moderators, who ensure that all participants are respected by others and can make themselves heard.

27  Goodin and Dryzek (2006, 225). 28  Landemore (2015, 166–91). 29  Farrell et al. (2019). 30  Gastil and Knobloch (2020). 31  Fishkin (2018, 173–4). 32  Fishkin (2018, 101–11). 33  Fishkin (2018, 189–95). 34  Fishkin (2018, 16, 73–4).

Resisting Political Minimalism  225 Finally, small-­group deliberation is interwoven with plenary sessions during which participants can question panels of experts.35 This careful structuring encourages participants to engage with one another in an informed, open-­minded, and trustful manner, which in turn has impressive epistemic results. In the case of deliberative polls, Fishkin observes that partici­ pants habitually become significantly more knowledgeable regarding the issue at hand, and that their policy preferences often change accordingly.36 In sum, small-­scale structured deliberation can include ordinary people in public discourse in a way that is epistemically effective, even when they are ini­ tially uninformed about the issues at hand. It does so, in part, by taking people out of non-­ideal contexts that are rife with social fragmentation and misinforma­ tion, and putting them into contexts regulated by strong norms of open-­minded and trustful engagement. The use of minipublics therefore seems a prima facie promising tool for overcoming the difficulties raised by robust political ignorance. Nevertheless, as it stands, the idea of introducing small-­scale structured deliberation by representative samples of the population leaves open a crucial question: the question of what position minipublics should have in the policy­ making process—and, relatedly, how empowered they should be. At one extreme, minipublics could conceivably have exclusive decision-­making power, and replace existing decision-­making arenas (such as legislatures populated by elected repre­ sentatives). While this proposal has some influential supporters,37 most pro­pon­ ents of minipublics take an alternative view: they suggest that minipublics should complement, rather than replace, existing processes of mass-­decision-­making.38 This complementary role can take various forms. Many minipublics have an advisory role, whereby they provide information and recommendations to the demos and their representatives during the decision-­ making process. For ex­ample, Citizens’ Initiative Reviews aim to produce a succinct one-­page analysis of the referendum proposal they examine, which is then passed on to voters in the official Voters’ Pamphlet.39 Similarly, the Irish Citizens’ Assembly on abortion culminated in a report which it relayed to the government, and which informed broader public debates during the referendum that followed.40 But minipublics can also occupy more empowered roles that fall short of full decisional power. They can, for instance, have agenda-­setting power—the power to initiate reforms by putting proposals on the agenda of parliamentary or popu­ lar votes. Notably, in the case of the British Columbia Citizens’ Assembly, the 35  Fishkin and Luskin (2005, 287–9). 36  Fishkin (2018, 97–8, 143–6, 157–8, 196–7). See also Fishkin and Luskin (2005, 290–1). Gastil and Knobloch (2020, chs. 6, 8) report similarly encouraging epistemic effects in their study of Citizens’ Initiative Reviews. 37  Guerrero (2014); López-­Guerra (2014); Hennig (2019); Bouricius (2019). 38  E.g., Goodin and Dryzek (2006), MacKenzie and Warren (2012), Curato and Böker (2016), Fishkin (2018), Farrell et al. (2019), Gastil and Knobloch (2020). 39  Gastil and Knobloch (2020, 16). 40  Farrell et al. (2019, 119–20).

226  Democratic Speech in Divided Times electoral system it recommended needed to be put to a referendum vote. Even more ambitiously, some theorists have suggested that minipublics could operate as a second legislative chamber that initiates and decides on legislative proposals alongside a chamber of elected representatives.41 As these examples indicate, there are many ways in which minipublics might complement without replacing existing mass decision-­making arenas. The crucial point for our purposes is that, so long as minipublics are under­ stood in this broad complementary sense, the proposal of using minipublics to address the problem of robust political ignorance is entirely consistent with the picture of inclusive public discourse I have been defending throughout this book. Recall that on the systemic understanding of public discourse, public discourse is composed of many interconnected arenas of political talk, which perform com­ plementary roles. Considered in this light, deliberative minipublics are just one arena in the broader discursive system, which interact with, and feed into, other discursive spheres.42 What this means is that, insofar as small-­scale deliberative initiatives help overcome difficulties stemming from robust political ignorance, this is not an argument for departing from deliberative democracy. Rather, it helps identify how deliberative democracy, and the inclusive political talk it calls for, might be improved in these conditions. Now, one might doubt whether using minipublics as a complementary arena in the broader system of public discourse really does help overcome the difficulties raised by robust political ignorance. After all, if minipublics are just one arena among others, why think that these other arenas will heed their advice or pro­ posals? Just as robustly ignorant people may fail to identify reliable information shortcuts, or fail to recognize good arguments, so too they may fail to recognize the epistemic value of minipublics’ recommendations and proposals. The British Columbia Citizens’ Assembly exemplifies this possibility. Chambers complains that shallow rhetoric dominated public discussion of the Assembly’s recom­ mended electoral system. Consequently, the proposal was ultimately defeated.43 A similar observation applies to the ‘Australia Deliberates’ deliberative poll con­ ducted in 1999. As Fishkin notes, not only did the minipublics’ deliberations struggle to reach many voters, but, in addition, they were partly drowned out by a slogan-­driven television response from the opposition.44 The worry, then, is that even if ignorant citizens can learn effectively inside minipublics, the robust political ignorance that obtains outside of minipublics risks nullifying their influence on other arenas that play a necessary role in policymaking. On the basis of such considerations, Cristina Lafont concludes that complementary 41  Gastil and Wright (2019); Abizadeh (2020); Landemore (forthcoming); see Owen and Smith (2019) for a critical assessment. 42 For a similar view of complementary minipublics, see Chambers (2009), Curato and Böker (2016, esp. 187). 43  Chambers (2009, 331). 44  Fishkin (2018, 170–1).

Resisting Political Minimalism  227 minipublics will probably make ‘no difference whatsoever’ to the broader public sphere.45 Lafont’s pessimistic conclusion is almost certainly too strong. For one thing, there is evidence that minipublics do at least sometimes have a positive influence on the broader public. In their examination of Citizens’ Initiative Reviews, for example, Gastil and Knobloch find that voters who read the Review’s report tend to become more knowledgeable about the issue at hand; that a substantial number of them change their views as a result; and that, typically, the report is con­sidered to be trustworthy.46 This evidence is not altogether surprising. According to Michael MacKenzie and Mark Warren, minipublics have distinctive properties in virtue of which members of the public can more easily recognize their trustworthiness. First, ran­ dom selection means that minipublics, unlike typical groups of experts, are likely to be descriptively representative of the population. So insofar as people are more likely to trust individuals from similar social backgrounds, they are more likely to  trust minipublics.47 Second, Alex Guerrero argues that because minipublics are selected randomly, they do not rely on electoral campaign financing, and are therefore less vulnerable to co-­optation by powerful lobbies. If so, minipublics may be more visibly trustworthy than elected officials or the experts they appoint.48 Let us take stock. I have argued (1) that, used in a complementary capacity, deliberative minipublics are entirely consistent with the deliberative ideal of large-­scale and inclusive public discourse, and (2) that they arguably help to attenuate the problems that afflict large-­scale public discourse in conditions of robust political ignorance. So, the proposal that we implement complementary minipublics cannot yield an argument for departing from deliberative democ­ racy. On the contrary, it reveals a promising avenue for improving inclusive ­public discourse in non-­ideal conditions. Having said that, exactly how far minipublics succeed in positively impacting other parts of the discursive system remains an empirically open question. Moreover, although minipublics may well improve the broader discursive system, it is also likely that conditions of robust political ignorance set limits on the extent to which they do so. As the British Columbia and Australia Deliberates cases illustrate, minipublics sometimes have a limited impact on uninformed non-­participants.49 In light of these limitations, one might be tempted to employ minipublics in the more radical way I mentioned above. On this understanding, recall, relatively small randomly selected deliberative groups are fully empowered to actually 45  Lafont (2015, 88), emphasis added. 46  Gastil and Knobloch (2020, 129–40). 47  MacKenzie and Warren (2012, 105–7). 48  Guerrero (2014, 164). 49  See Curato and Böker (2016, 179–82) for a similar point.

228  Democratic Speech in Divided Times make political decisions. Thus, they replace normal sites of mass democratic decision-­making, such as legislatures composed of elected officials. In recent years, Alex Guerrero, Claudio López-­Guerra, Brett Hennig, and Terrill Bouricius have all defended this more extreme proposal.50 Following Guerrero, let us call this proposal ‘lottocracy’. In our context, lottocracy’s appeal is straightforward. Since the randomly selected deliberative group no longer needs to have its recommendations ratified by the broader public (or the representatives they have elected), these are no longer hostage to the widespread political ignorance of ordinary citizens. Lottocracy apparently circumvents the difficulties raised by robust political ignorance. An immediate concern is that lottocracy may seem undemocratic. Decision makers are not selected through elections where citizens have an equal vote. Consequently, citizens are subjected to rule by individuals who are not formally accountable to them.51 Lottocrats could respond in two ways. First, like Guerrero, they could argue that rule by randomly selected deliberative groups is democratic in important respects.52 Alternatively, they could claim that it is relatively unprob­ lematic that lottocratic arrangements are undemocratic. This might be because they are unconvinced that there is anything intrinsically valuable about demo­ cratic decision-­making procedures. Or, like López-­Guerra, lottocrats might think that democracy’s intrinsic value is outweighed by the instrumental value of having a political system that remains epistemically capable of producing good decisions in conditions of robust political ignorance.53 To sidestep these responses, the rest of this chapter will develop a different line of criticism, which does not appeal to democracy’s intrinsic value. So far, I have only examined responses to the difficulties raised by robust political ignorance that are traditionally democratic—where ‘traditionally democratic’ means that, unlike in lottocracy, ultimate decisional authority lies with the entire demos, each of whom at least has an equal right to vote. However, because all the proposals examined so far are traditionally democratic, they involve a bottleneck of inclu­ sive decision-­making where political ignorance generates difficulties. Ignorant citizens are liable to be swayed by the wrong arguments, testimonies, or cues (Section 1.2), to vote for the wrong elites (Section 2.1), and to disregard the re­com­menda­tions or proposals of randomly selected deliberative minipublics (Section 2.2). In the face of these problems, one might therefore be tempted to remove the democratic bottleneck, and opt for a non-­democratic response to conditions of robust political ignorance. The next section will show that, even if 50  Guerrero (2014); López-­Guerra (2014); Hennig (2019); Bouricius (2019). 51  See, e.g., Lafont (2015, 42–54), López-­Guerra (2014, 27), and Tuck (2019). 52  Guerrero (2014, 167). 53  López-­Guerra (2014, 26–8, 52–6).

Resisting Political Minimalism  229 we bracket the intrinsic value of democracy, this temptation is misleading. Ultimately, the most influential non-­democratic strategies for avoiding the diffi­ culties raised by robust political ignorance—including lottocracy—merely postpone these difficulties.

3.  Why Robust Ignorance is a Problem for Non-­democratic Views As advertised, I will be arguing that the second and more radical version of the political minimalism objection, which recommends shifting to non-­democratic political ideals, in fact also fails to circumvent the challenges posed by robust political ignorance. Now, one might worry that whether or not this is true depends on the non-­ democratic system in question. I will focus on the two most prominent non-­ democratic arrangements that have been advanced in response to the observation that citizens tend to be widely and robustly uninformed about politics. I begin by examining epistocracy—rule by knowers—before returning to lottocracy. Starting with epistocracy, which is more obviously non-­democratic, will be useful for two reasons. On the one hand, its shortcomings serve to highlight the prima facie appeal of lottocracy. On the other, the argumentative strategy developed against epistocracy will help identify how lottocracy also falls short. Although I restrict my focus to specific attempts at realizing epistocracy and lottocracy, I will supply reasons for thinking that the problems they encounter do not depend on their institutional specificities.

3.1  Epistocracy and Robust Ignorance Epistocracy is a political system that distributes decision-­making power on the basis of prior knowledge. Though its most famous incarnation is Plato’s recom­ mendation of rule by philosopher kings, Jason Brennan has influentially rearticu­ lated this ideal. Against the background of widespread and robust political ignorance, epistocracy’s appeal is clear: it proposes to empower those and only those who are knowledgeable. So, if epistemically privileged knowers exist, and if they can be identified and empowered, then epistocracy straightforwardly avoids the problem raised by robust political ignorance. One might doubt that there are privileged knowers in politics, generally con­ strued. Here, I wish to set aside such concerns. Instead, the key question for epis­ tocracy is how knowers are to be identified and empowered. How, in other words, do we transition to rule by knowers? Brennan’s principal suggestion is that we establish a voter qualification examination. The examination asks questions

230  Democratic Speech in Divided Times regarding basic facts that are crucial to making competent political decisions. Only those who perform sufficiently well are then enfranchised.54 However, this recommendation only pushes back the transitional question. Even if there could be an adequate qualification examination, the question becomes how we are to design and implement it. Here, the difficulties raised by conditions of robust political ignorance resurface. A first option, which Brennan leans towards, is that democratic decision-­making could decide what counts as competence.55 But why think that the demos—robustly ignorant as it supposedly is—will be competent at designing the examination? The examination tests citi­ zens for knowledge that is crucial to good political decision-­making. Hence, insofar as democratic decision-­making can identify the knowledge requirements that the examination tests for, it is capable of good political decision-­making. Putting the same point differently, insofar as democratic decision-­making is incompetent at identifying good options—as the ignorance-­ driven argument for political minimalism alleges—it is incompetent at identifying the examination’s know­ ledge requirements. (Notice that this observation equally applies if the demos is supposed to identify competent individuals who then design the test. Insofar as the demos can do so, it can elect competent representatives.) The key point is not that it is impossible democratically to design a voter qualification examination. Rather, it is that Brennan’s proposal for implementing epistocracy works only insofar as widespread and robust political ignorance in fact fails to undermine democratic decision-­making.56 Therefore, on this way of proceeding, epistocracy cannot be better than democracy at weathering robust political ignorance. If the democratic route to designing and implementing the voter qualification examination is unpromising, maybe, instead, those who are politically know­ ledge­able should somehow forcibly impose the examination. However, this rec­ ommendation is question-­begging. In political contexts, very often, what is in dispute just is who is knowledgeable regarding what we should do. Parties in political disagreements typically take themselves to be right, and their adversaries wrong. Moreover, a characteristic feature of non-­ideal conditions characterized by widespread and robust political ignorance is that (as Brennan himself suggests57) the powerful and the wise seldom coincide. 54  Brennan (2016, 211–14). 55  Brennan (2016, 224). 56  Though Brennan considers this objection head-­on (2016, 224–6), his answer is unsatisfactory. He challenges the assumption that if you know what the competence requirements are then you are yourself competent, by suggesting that you might know the competence requirements de dicto but not de re. Simply put, you might know that rulers should know the basic facts of economics, without know­ ing what the basic facts of economics are. But if this is what Brennan means when saying that demo­ cratic decision-­making bodies should decide what counts as competence, it again merely postpones the transitional question. If democratic decision-­making only determines the competence require­ ments de dicto, then elaborating the test still requires having someone determine what they are de re. But how are they chosen? If the answer is ‘democratically’, the problem resurfaces. 57 See, e.g., Brennan’s (2016, 32) assertion that ignorant citizens are far more numerous than knowledgeable citizens.

Resisting Political Minimalism  231 Putting these two observations together yields an undesirable result. If most competing political actors take themselves to be politically knowledgeable, and if the powerful and the knowledgeable rarely coincide, we have reason to expect that the question-­begging recommendation at hand will in practice not achieve rule by the knowledgeable. Rather, it will achieve rule by whomever the powerful take—however erroneously—to be knowledgeable.58 Thus, even on this alterna­ tive proposal, circumstances of widespread and robust political ignorance plague the transition to epistocracy: when the powerful are also ignorant, a policy of having the knowledgeable impose epistocracy seems ill-­suited to achieving rule by genuine knowers. We can in fact go further than this. While I am emphasizing the difficulties that arise for epistocracy at the transitional stage, there are reasons to think that robust political ignorance does not merely pose a one-­off transitional problem for epistocracy. Political circumstances are constantly changing: regional wars insti­ gate flows of migration, financial crises disrupt the existing economic system, new technologies give rise to novel kinds of social division, and so on. When political circumstances change in this way, the problems that states must solve also change. And, since different bodies of knowledge are needed to solve differ­ ent problems, this means that the knowledge requirements that define political competence change over time. Consequently, maintaining epistocracy in the face of continuously changing political circumstances requires continuously designing and implementing new tests for identifying and selecting political knowers. If—as I have argued—robust political ignorance plagues the process of designing and implementing such epistocratic tests, then this entails that robust political ig­nor­ ance in fact poses an ongoing problem for epistocracy.59 Epistocrats might respond that these worries about identifying and empower­ ing knowers are simply theoretical, and rest upon speculative claims—for ex­ample, about whether or not the powerful and the knowledgeable coincide. But the theoretical worries I have raised are far from speculative. The real-­world track record of voter qualification examinations is terrible. In the American South, for instance, Jim Crow laws implementing literacy tests were widely used to disen­ franchise black Americans after the Civil War. Brennan considers this observation, and retorts that it only shows that voter qualification examinations have been poorly implemented in the past. ‘These tests were administered in bad faith [. . .] the fact that governments used to hide their

58  For a similar argument, which emphasizes the self-­conception as knowers of tribal citizens, see Aikin and Talisse (2020, 51–2). 59  Bagg (2018b, 896–8) also points out that, even setting aside the epistemic difficulty of designing these tests, it is likely that powerful forces would deliberately design the tests to benefit themselves and to entrench their own power. I agree. Here, however, I am bracketing this concern to make the more specific point that robust political ignorance poses a problem for transitioning to and maintaining epis­ tocracy. So epistocracy does not even avoid the problem it is designed to avoid.

232  Democratic Speech in Divided Times racism beneath an epistocratic guise does not show that epistocratic exams are inherently objectionable’.60 But this response jars seriously with the broader epis­ tocratic challenge. The relevant question, Brennan continuously stresses, is not ‘is democracy inherently objectionable?’ or ‘is epistocracy inherently objectionable?’. Brennan persistently criticizes democratic theorists—particularly deliberative democratic theorists—for defending theories which, though not inherently bad, are overly idealized and therefore impracticable in real-­world conditions.61 Nor is this specific to Brennan. As we saw in Section 1.1, the whole point of the argument for political minimalism is that, though deliberative accounts of democracy are appealing in the ideal, they may operate under unrealistic assump­ tions. So, on pain of inconsistently applying their standards, epistocrats should be asking: given realistic assumptions about what people and social circumstances are like, which political system should we strive for? In this context, the empirical evidence regarding epistocratic arrangements, including voter examinations, is frightful.62 And this is no accident. It accords with the theoretical argument I offered for thinking that, in non-­ideal conditions where power and knowledge come apart, we should expect epistocratic experiments to be misapplied. Perhaps, however, in suggesting that epistocracy is theoretically and em­pir­ic­ al­ly inadequate, I am overlooking an alternative version of epistocracy, that— unlike Brennan’s—has been empirically tested and has proven successful in non-­ideal conditions. In particular, Daniel Bell has prominently defended the ‘China Model’ of epistocracy, which is inspired by China’s political reforms in the 1990s.63 In this sophisticated model, which involves different forms of rule across central, middle, and local government, central government positions are distrib­ uted according to competence and virtue. Competence and virtue, in turn, are identified via examinations that test candidates’ intellectual ability (which includes politically relevant knowledge), emotional intelligence, and virtue. These examinations are numerous and ongoing: after passing public service exam­in­ ations, candidates ‘must perform well at lower levels of government, with more rigorous tests and evaluations at each step of the way, to move further up the chain of political command’.64 Prima facie, the China Model might seem to constitute a counterexample to my theoretical argument that conditions of robust political ignorance jeopardize the transition to epistocracy. After all, it is a form of epistocracy that has actually been established, via non-­democratic means. Accordingly, insofar as the China Model has proven successful—and Bell emphasizes its ‘stunning economic success’65—this may seem to contradict my above objection: that, in non-­ideal conditions where the powerful and the wise seldom coincide, non-­democratic 60  Brennan (2016, 223–4). 61  Brennan (2016, 69–73, 186–7, 206–7). 62  See also Popper’s (1995[1945], 145n25) critique of Plato’s Academy, which ‘was notorious for breeding tyrants’. 63  Bell (2016). 64  Bell (2016, 169). 65  Bell (2016, 4).

Resisting Political Minimalism  233 attempts at establishing epistocracy are likely to fail to empower those who are genuinely wise and competent. Nevertheless, Bell’s recommendation ultimately does not escape my objection. Despite its impressive record on poverty alleviation, China continues to face immense problems, including grave inequalities, widespread corruption, a disas­ trous environmental record, extensive human rights abuses, and declining social mobility.66 Bell himself therefore acknowledges ‘that democracies have the best record overall compared to other forms of government in the past’.67 Hence, closer inspection of the empirical record actually corroborates my concern: in real-­world conditions, non-­democratic attempts at establishing epistocracy seem comparatively unsuccessful. In reply, Bell insists that this evidence is not necessarily problematic for the epistocratic ideal, because real-­world China only imperfectly realizes this ideal. Because of the significant social divisions and inequalities that exist in Chinese society, China’s current selection procedure, he notes, falls short of identifying the best candidates. Thus, he insists that ‘there remains a large gap between the ideal and the reality’.68 Bell’s point is that, if this gap were filled, the China Model could potentially outperform democracies. However, like Brennan’s, this reply jars with the spirit of the broader argument for political minimalism. The challenge, to reiterate, holds that deliberative democracy is bound to perform badly given how uninformed people actually are in non-­ideal conditions, which are marked by political misinformation, dogma­ tism, social fragmentation, and distrust. So, defenders of epistocracy must estab­ lish that, in non-­ideal conditions, with people as they are, epistocracy would do better. The above evidence suggests that it does not: those charged with develop­ ing Chinese selection examinations have in fact struggled to design them so as to identify and empower the best candidates. In this context, to reply, as Bell does, that the China Model would succeed if inequalities and social divisions were lower is to apply a double standard. Doing so compares democracy in non-­ideal conditions to epistocracy in idealized conditions: while epistocrats criticize democracy for being unworkable given how politically ignorant people are in non-­ideal conditions, the argument for the China Model’s comparative desirabil­ ity depends on people and circumstances being significantly better-­than-­actual. Despite its greater sophistication, then, Bell’s account of epistocracy ultimately reproduces the problems of Brennan’s account: either its normative appeal is threatened by the same non-­ideal conditions that allegedly undermine (delib­ erative) democracy, or it helps itself to idealizations that it denies (deliberative) democratic theorists.69 66  Bell (2016, 43, 52, 112–13, 125–31, 173). 67  Bell (2016, 19). 68  Bell (2016, 9). 69  In addition, some of Bell’s recommendations for improving Chinese epistocracy also run foul of the first horn of my above objection to epistocracy: that democratically transitioning to epistocracy replicates the ignorance-­related difficulties of democracy. Indeed, Bell (2016, 176) recommends

234  Democratic Speech in Divided Times Note, finally, that while I have focused on Brennan and Bell’s theories of e­ pistocracy, the worries their accounts face should generalize. What unites all epistocratic theories is the claim that political decision-­making power should be distributed on the basis of prior knowledge. As such, they all face the transitional question of how we are to identify and empower knowers. And this transitional issue is precisely where conditions of robust political ignorance generate difficul­ ties. So, contrary to what the epistocratic variant of the argument for political minimalism claims, robust political ignorance does not recommend abandoning the deliberative ideal, with its emphasis on inclusive public discourse, in favour of epistocracy.

3.2  Lottocracy and Robust Ignorance This problem might motivate a turn to lottocracy, or rule by randomly selected people. Remember that, as I am using the term, ‘lottocracy’ refers to a political system that replaces existing assemblies of democratic decision-­making (such as arenas composed of elected representatives) with assemblies of people who have been selected by lot. Some define ‘lottocracy’ in a different way, such that it involves lottery-­selected groups that complement existing democratic arenas of decision-­making. I am setting aside this latter proposal because, as we have already seen in Section 2.2, it is in fact consistent with the deliberative ideal of democracy. So it does not constitute an alternative to this ideal. Lottocracy proposes to overcome the difficulties that stem from robust pol­it­ ical ignorance as follows. First, it distributes political power randomly. Then, once decision makers have been randomly selected, they undergo a competence-­ building process where small-­scale deliberation features centrally. And as Section 2.2 discussed, there is substantial evidence that small-­ scale deliberative competence-­building processes significantly improve the knowledge of randomly selected citizens. Thus, López-­Guerra claims, by breeding rather than screening for knowers, lot­ tocracy improves on epistocracy’s attempt at avoiding the difficulties raised by robust political ignorance.70 It achieves rule by knowers, thereby keeping power away from ignorant hands. But since there is no need to screen for pre-­existing knowers, it apparently circumvents the transitional troubles of epistocracy. Predictably, however, the turn to lottocracy in the end merely succeeds in post­ poning the difficulties raised by robust political ignorance. The crucial premise

implementing a referendum to approve the epistocratic arrangement. But if this is the case, then the ignorance-­related objection to democracy presumably also plagues the transition to, or maintenance of, Bell’s epistocracy. 70  López-­Guerra (2014, 25–6).

Resisting Political Minimalism  235 underpinning lottocracy’s appeal is that a competence-­building process can make randomly selected citizens sufficiently knowledgeable to rule well. Correspondingly, the crucial question is how this competence-­building process operates. Either the competence-­building process involves substantial interventions by experts, or it does not. If it does not, then, in the present context, it seems unclear why delib­ eration would make the randomly selected group more know­ledge­able. After all, as we saw in Section 1.2, the argument for political minimalism starts from the premise that ordinary citizens are so robustly politically uninformed that dis­ cussion between them is epistemically fruitless. If, however, experts do play an essential role—and this is what theorists of minipublics generally suggest71—then we have another iteration of the problem epistocracy faced: transitioning into ­lottocracy requires identifying and selecting expert knowers. And just as robust ignorance plagued epistocracy with respect to this issue, so it should plague lottocracy.72 Guerrero’s influential articulation of lottocracy brings this out. Like all lotto­ crats, he asserts that expert presentations constitute an essential feature of the competence-­building process.73 How do we identify experts in such cases? If the answer is through democratic decision-­making, then Guerrero cannot consist­ ently hold both that this will work well and that democratic decision-­making will be incompetent at electing representatives. As in the critique of epistocracy, the point is that if—as the ignorance-­driven argument for political minimalism alleges—democratic decision-­making is ineffective at identifying competent and well-­motivated representatives, it is unclear why it would be effective at identify­ ing competent and well-­motivated experts. So this strategy for identifying know­ ers functions only insofar as conditions of robust political ignorance fail to undermine democratic decision-­making. Alternatively, if the suggestion is that some competent agency should impose experts through non-­democratic means, we have seen when critiquing epistoc­ racy that this solution is objectionably question-­begging: in context, who is com­ petent just is what is in dispute. And since the powerful and the knowledgeable often come apart in non-­ideal contexts marked by widespread and robust pol­it­ ical ignorance, there are reasons to expect that the most powerful, rather than those who are actually good at identifying experts, will get their way. Ultimately, 71  Among lottocrats, see Guerrero (2014, 156), López-­Guerra (2014, 35–7). Among defenders of complementary minipublics, see, e.g., Fishkin and Luskin (2005, 288–9), Goodin and Dryzek (2006, 223–4), Smith et al. (2015, 247), Farrell et al. (2019, 116). 72  In Section 3.1, I suggested that the problem of public ignorance applies not simply when first transitioning to epistocracy, but also when maintaining epistocracy in changing circumstances. A similar consideration applies to lottocracy. Over time, political circumstances change and require dif­ ferent experts. Hence, the problem of identifying experts may well be an ongoing problem for lottoc­ racy. For the sake of simplicity, however, this section will focus on how this problem arises at the transitional stage. 73  Guerrero (2014, 156).

236  Democratic Speech in Divided Times Guerrero concedes that ‘a full defense of the lottocratic alternative will have to do more to specify the details about the [expert] qualification assessment process’.74 In other words, he acknowledges that the selection of experts in contexts of pol­it­ ical ignorance is a problem for lottocracy, but does not offer a solution. Because of these difficulties, lottocrats might be tempted to criticize the other horn of the dilemma, and suggest that small-­scale deliberation remains epi­stem­ ic­al­ly fruitful even without experts. Guerrero, for instance, briefly mentions the possibility of ‘eliminat[ing] the expert stage’.75 The problem with this suggestion is that the evidence supporting the epistemic value of deliberative minipublics over­ whelmingly concerns cases involving expert contributions. In a recent overview, Graham Smith goes so far as to assert that ‘in all mini-­publics, participants are exposed to competing expert viewpoints’.76 The upshot is that, insofar as lotto­ crats recommend eliminating the expert stage of lottocracy, they can no longer appeal to the existing empirical evidence on the success of minipublics. Therefore, they lose a key incentive for embracing lottocracy in conditions of robust political ignorance: the concrete evidence that small-­scale deliberation within randomly selected groups is epistemically fruitful.77 Perhaps, however, I am overstating this problem. Some cases of successful small-­scale deliberation arguably did not rely—or relied far less substantially—on close interaction, guidance and framing by experts. This is plausibly the case with Chicago’s ‘community policing’ initiatives, where a neighbourhood’s citizens and police officers come together to deliberate about that neighbourhood’s public safety issues.78 And it is also arguably true of the hugely successful ‘participatory budgeting’ initiatives pioneered in Porto Alegre in the 1980s, and which have since proliferated across the world, where assemblies of ordinary citizens discuss which public spending projects should be prioritized.79 Nevertheless, it is unclear whether this further evidence can genuinely support the ignorance-­driven argument for lottocracy. First, unlike typical minipublics, neither Chicago’s community policing nor Porto Alegre’s participatory budgeting involve randomly selected participants. Instead, participants are self-­selected. This difference matters. By contrast with lottocratic deliberative groups, self-­ selected deliberative groups are unlikely to constitute a representative sample of the public.80 Furthermore, Archon Fung suggests that the overrepresentation of some groups—in particular, of disadvantaged groups—may have contributed 74  Guerrero (2014, 162). 75  Guerrero (2014, 175). 76  Smith (2012, 94). 77  The problem may in fact be even worse. There is some evidence that more empowered minipub­ lics tend to be less deliberative than other minipublics. See, e.g., Smith et al. (2015, 253–4), Bächtiger and Parkinson (2019, 57–8). This is concerning for lottocrats, since they embrace the most em­powered form of minipublic. Assuming that deliberation is epistemically fruitful, this evidence raises the worry that, even with experts, lottocracy will be less epistemically fruitful than the less empowered minipub­ lics with which political scientists have experimented. 78  Fung (2003, 359–60). 79  Fung (2003, 360–2). 80  Goodin and Dryzek (2006, 222).

Resisting Political Minimalism  237 to the success of participatory budgeting and community policing.81 Hence, the  success of these initiatives does not necessarily lend support to expert-­less lottocracy. Even setting this concern aside, there is a second problem. Insofar as there is evidence that deliberation between ordinary citizens is successful even without carefully structured guidance by experts, this weakens the core starting premise of the argument for political minimalism. As explained in Section 1.2, the ­argument for political minimalism starts from the assumption that, because of conditions such as fragmentation, misinformation, distrust, and group-­based dogmatism, ordinary people tend to be deeply ignorant about political affairs, in  a way that prevents them from learning from inclusive discussion. Contrary to  this assumption, the evidence mentioned above suggests that, in real-­world conditions, deliberation between ordinary people may after all be epistemically fruitful. So, even if cases of participatory budgeting and community policing did support lottocracy, they also support deliberative democracy. As a result, they do not lend credibility to the argument for abandoning deliberative democracy in favour of this more minimalistic political ideal. The broader conclusion is this: if lottocracy involves experts, then it replicates the difficulties stemming from robust political ignorance that epistocracy faces; and if, by contrast, lottocracy eliminates experts, this should reduce our confi­ dence in the claim that lottocracy is epistemically superior to alternatives (such as deliberative democracy) in conditions of robust political ignorance. Either way, robust political ignorance does not give us a reason to abandon the deliberative ideal of democracy in favour of lottocracy.

4. Conclusion In previous chapters, we have seen that, while conditions such as political misin­ formation, group-­based dogmatism, social and spatial fragmentation, and dis­ trust do not strictly undermine the ideal of democratic public discourse I have put forward, they do threaten to limit or reduce its good functioning. They do so, in part, by making individuals politically ignorant in a way that reduces their abil­ ity to benefit from democratic public debate. Moreover, tackling these problem­ atic background conditions—particularly fragmentation—will be an arduous task. In response to these remaining difficulties, some are tempted to abandon the deliberative ideal of democracy, with its emphasis on inclusive public discourse, in favour of a more minimalistic political ideal—an ideal that abandons the aim of involving, on a large scale, ordinary people in public discussion.

81  Fung (2003, 359–62).

238  Democratic Speech in Divided Times The present chapter has argued that this move to political minimalism is a ­ istake. Even if we assume for the sake of argument that we find ourselves in m conditions of robust political ignorance—conditions, that is, where citizens are widely ignorant about politics, in a way that makes them unable to engage fruit­ fully in public debate—it still does not follow that we should do away with the deliberative ideal. This is because such conditions of robust political ignorance are equally problematic for the more minimalistic ideals that have been put ­forward as alternatives. Elite democracy does not avoid the problem, since it relies on ordinary c­ itizens, ignorant as they supposedly are, to select and learn from elites. Nor is advocating small-­scale deliberation within minipublics viable as an alternative to deliberative democracy. Not only does robust political ignorance pose difficulties for using minipublics in a complementary capacity within a broader discursive system, but (more importantly) even insofar as there are reasons to think that such minipub­ lics help attenuate these non-­ideal conditions, their use is entirely consistent with deliberative democracy. A more extreme alternative recommends departing from democracy al­together. However, neither of the two most promising non-­democratic proposals—rule by empowered minipublics and rule by knowers—truly circumvents the difficulties posed by robust political ignorance. Instead, they simply relocate these difficulties to an earlier stage, where we ask how we are to transition to and maintain these non-­democratic models. So, if widespread and robust political ig­nor­ance seriously threatens deliberative democracy, then it also seriously threatens the political ideals advanced as replacements. In this light, the present chapter should steer deliberative theorists away from the siren songs calling for us to abandon inclusive public discourse as a political ideal. Because robust political ignorance is equally problematic for all the com­ peting political models, this phenomenon gives us no reason to depart from the deliberative ideal. Put differently, since inclusive and large-­scale public discourse is not the source of the problem, our focus should not be on de-­emphasizing it. Instead, our focus should be, first and foremost, on tackling robust political ig­nor­ance and the problematic background conditions that sustain it. In par­ ticular, we should not lose patience with the task, outlined in Chapter  7, of ­de-­fragmenting the democratic public sphere. At the same time, deliberative theorists should continue to rethink public speech, conceptually and institutionally, in a way that makes it less vulnerable to conditions of robust political ignorance. For instance, we have seen in this chap­ ter that deliberative institutional innovations—such as deliberative polls—may be helpful in educating uninformed citizens, and in transmitting their pol­it­ical knowledge to the broader discursive system. If abandoning inclusive discourse were a promising way of avoiding the difficulties raised by robust pol­it­ical ig­nor­ ance, we might give up on developing such innovations. But since it is not, we should instead redouble our efforts.

Conclusion 1.  Summary of the Argument Public speech that brings together, on a large scale, people from diverse walks of life is one of the cornerstones of a healthy democratic order. It is a crucial resource for pooling the politically relevant knowledge that is distributed across different social groups. And, relatedly, it constitutes an invaluable tool for holding political power accountable, and thereby protecting people from domination. But not all kinds of public speech or public discourse achieve these democratic functions to the same degree. Moreover, what kinds of speech best serve these functions may differ based on the social and political context in which we find ourselves. Accordingly, the central question I have been concerned with is the following: what rules or norms should govern inclusive public discourse, if we want it to perform these epistemic and anti-­domination functions in non-­ideal circumstances characterized by gross and often unjust divisions? The first broad conclusion is that some of the discursive practices that are symptomatic of deep divisions actually play a key positive role in non-­ideal circumstances. Specifically, I have argued that we should welcome the presence of emotionally charged speech, most notably expressions of profound anger, in the public discourse of divided democracies. To see why, remember the core dilemma that we started with. On the one hand, having speakers in formal arenas of public discourse frame their discussions in terms of shared reasons is important, because this practice helps to facilitate the subsequent contestation of political power. But, at the same time, this practice risks being too exclusive. As a result of the epistemic divisions that obtain between groups living in segregated and unequal conditions, even the weakest interpretation of what it means for a consideration to be ‘shared’ will exclude many important considerations.1 Emotionally charged narrative helps to attenuate this dilemma. Because emotions are sources of salience, and because they allow us to register evaluative properties that outstrip our existing conceptual categories, conveying emotions to others enables us to bridge our epistemic divisions—and, in doing so, to expand the pool of shared resources from which public debate might then ­proceed. Anger is especially important here, because the distinctive property it 1  Chapter 1, Section 2. Democratic Speech in Divided Times. Maxime Lepoutre, Oxford University Press (2021). © Maxime Lepoutre. DOI: 10.1093/oso/9780198869757.003.0010

240  Democratic Speech in Divided Times makes salient, and allows us to register, is injustice. Since injustice is both rampant and often misunderstood in divided democracies, we therefore need anger in public speech.2 By contrast, some characteristic features of the public discourse of divided democracies are not welcome. Hate speech, which erodes the basic standing of vulnerable groups, and misinformation, which deceives its targets about pressing political matters, both constitute dire threats to the value of democratic public discourse. Yet public speech can give us powerful tools for countering these harmful forms of speech. In particular, I have argued that counterspeech is often a preferable way of countering harmful speech than the use of coercive legal bans— provided, that is, counterspeech is governed by the right norms. Among other things, counterspeech should be driven by the state, rather than simply by private individuals. It should involve the positive reaffirmation of egalitarian ideals and political truths, rather than the mere negative rebuttal of hateful or deceptive propositions. And, importantly, it should be a continuous process, which can just as well pre-­empt as follow harmful speech.3 This normative picture of public discourse is explicitly designed to be sensitive to, and action-­guiding in, non-­ideal conditions. Still, we have seen that infelicitous background conditions can pose problems for it. When people dislike or distrust each other so much that they refuse to speak or listen to one another;4 when they are so spatially and socially fragmented that they lack the opportunity to do so;5 when they are too dogmatically attached to their group-­based perspective;6 and when (as a result), they know too little to distinguish good public interventions from bad public interventions7—when some or all of these conditions hold, this may limit the extent to which public discourse can perform its epistemic and anti-­domination functions. Though problematic, these conditions are not fatal to the normative ideal of democratic speech I have offered. The first thing to note is that, as we have seen, the kind of public discourse I recommend can go some way towards remediating these conditions. For instance, I have argued that, over and above their epistemic function, expressions of anger can help to rebuild trust and goodwill between alienated social groups.8 Second, even when it does not directly alter these conditions, the picture of public discourse I have put forward has resources in virtue of which it can mitigate the problems they raise. Here, a recurrent theme has been the importance of viewing public discourse as a system of interconnected arenas.9 In a discursive

2  Chapter 1, Section 5; Chapter 2. 3  Chapters 3–4. 4 Chapter 5. 5 Chapter 7. 6  Chapter 6, Section 5. 7  Chapter 6, Section 1. 8  Chapter 5, Section 4. 9  Chapter 1, Section 4; Chapter 2, Section 5; Chapter 5, Section 3; Chapter 6, Section 5; Chapter 7, Section 4.

Conclusion  241 system, third party speakers can mediate between distrustful participants, ­integrated neighbourhoods can allow information to circulate to and from s­ egregated neighbourhoods, and group-­based dogmatism can even be turned into an epistemically useful tool. Finally, we can and should supplement these two first responses, which operate fundamentally via public speech, with a response that is not speech-­based. Most importantly, to improve public discourse in non-­ideal circumstances, we ought to adopt policies aimed at creating a more integrated, or tightly connected, discursive system.10 In other words, we should implement policies aimed at creating more sustained interaction between diverse groups online, in the workplace, in neighbourhoods, and so on. These integrative policies are costly. They will take time and an enormous amount of effort, and will sometimes impose unfair burdens on vulnerable groups. And in the meantime, the conditions cited above (dislike, fragmentation, dogmatism, and ignorance) will limit the success with which we can promote more and better talk between diverse social groups. But, crucially, these costs do not constitute a reason to desist from this project. The conditions that pose problems for inclusive public speech—and, by extension for the deliberative ideal of democracy—also pose problems for more minimalistic political ideals (democratic or non-­ democratic).11 So, we should resist the temptation to jettison inclusive public discourse. Instead, as I stressed in the last chapter, our energy is better spent on more productive endeavours: we should continue investigating, first, how we can fairly eliminate these problematic social conditions; and, second, how we can further refine our discursive practices to make them more fruitful in the meantime.

2.  The Challenge Remaining My defence of this account of democratic public discourse does not purport to be exhaustive. As I have noted throughout, I take philosophical and empirical research to be continuous with one another. Philosophical insights regarding language, epistemology, emotion, and other concepts allow us to formulate working hypotheses regarding what norms of democratic speech may be desirable in divided circumstances. These hypotheses in turn call for further, more targeted, lines of empirical research. Armed with this empirical research, we will be able to validate, eliminate, or refine our theories of democratic speech. In closing, let me note three domains where the argument I have put forward directs us towards more specific empirical explorations. The first pertains to

10 Chapter 7.

11 Chapter 8.

242  Democratic Speech in Divided Times anger. While my thesis clarifies the democratic role of emotional speech by exploring angry narratives, the discussion of anger itself invites further research. My epistemic defence of anger depends on the successful transmission of anger to one’s audience.12 To substantiate this prescription, we must therefore examine under what specific social conditions expressions of anger are most capable of arousing anger in listeners. Relatedly, we must consider whether the kinds of expressions of anger that perform this epistemic function are of the same type as those that, I have suggested, facilitate trust-­building. The second domain has to do with the verbal correction of misinformation. Despite the growing literature on fake news and the difficulty of responding to it, I  noted earlier that most of the existing empirical evidence concerns ‘negative’ forms of counterspeech—the paradigmatic example of this being fact-­checking.13 But we have seen that there are reasons, most clearly articulated in applied philosophy of language, for thinking that ‘positive’ forms of counterspeech are more likely to be successful. Going forward, our experimental investigations of counterspeech should distinguish more clearly between positive and negative counterspeech, and disentangle their respective success at correcting misinformation. Finally, the third domain concerns the epistemology of systems. Philosophers have used formal modelling techniques to demonstrate (a) that dogmatism needn’t be epistemically bad provided information circulates well enough between dogmatic agents, and (b) that more circulation of information is not always better than less. I have hypothesized, on this basis, (a) that the dogmatism of social groups can be epistemically fruitful,14 and (b) that we need not integrate all discursive arenas to produce an epistemically valuable democratic system.15 But what we know here still remains too limited to be directly actionable. To arrive at actionable prescriptions, we need to find out, first, precisely how much circulation of information is needed given the levels of dogmatism that we observe in the political domain; and, second, under what kinds of social conditions information tends to circulate well. The former task will require, among other things, finer-­grained modelling from social epistemologists. And the latter will require more sociological investigations of the predictors of intergroup informational ties. We have encountered some preliminary research on this topic, both for online and offline networks. But we are still far from a firm understanding of the social conditions that incentivize intergroup engagement. Deepening this understanding will be crucial to determining the specific kinds of integrative policies we ought to adopt. Investigating these complementary lines of research is a challenging task. But this, once more, is no reason not to undertake it. We have seen that attempted ‘shortcuts’ that jettison inclusive discourse actually reproduce some of its greatest 12  Chapter 2, Section 4. 15  Chapter 7, Section 4.1.

13  Chapter 4, Section 5.

14  Chapter 6, Section 5.

Conclusion  243 problems. Furthermore, arduous though this task may be, it is also extremely rewarding. By integrating the wide-­ranging insights of philosophers and social scientists within democratic theory, it yields a broader understanding of the democratic ideal, of its relation to responsive decision-­making, and of the place of speech within that ideal. The reward is also political. Where democratic public speech is intimately bound up with exercises of power, and where that speech circulates in increasingly complex ways, we urgently need a theoretical framework that enables us to evaluate and improve upon existing discursive practices. The present book has outlined such a framework and demonstrated its promise. The task ahead is to fulfil that promise.

Bibliography Abizadeh, Arash. ‘On the Philosophy/Rhetoric Binaries’. Philosophy & Social Criticism 33, no. 4 (2007): 445–72. Abizadeh, Arash. ‘Representation, Bicameralism, Political Equality, and Sortition’. Perspectives on Politics (2020): 1–16. Achen, Christopher, and Larry Bartels. Democracy for Realists. Princeton, NJ: Princeton University Press, 2016. Aiken, Alex. ‘Alex Aiken Introduces the Rapid Response Unit’. Government Communication Service, 2018. https://gcs.civilservice.gov.uk/news/alex-aikenintroduces-the-rapid-response-unit/. Aikin, Scott, and Robert Talisse. ‘On “Fake News” ’. 3 Quarks Daily, 2018. https:// www.3quarksdaily.com/3quarksdaily/2018/05/on-fake-news.html. Aikin, Scott, and Robert Talisse. Political Argument in a Polarized Age. Cambridge: Polity Press, 2020. Allcott, Hunt, and Matthew Gentzkow. ‘Social Media and Fake News in the 2016 Election’. Journal of Economic Perspectives 31, no. 2 (2017): 211–36. Alter, Adam, and Daniel Oppenheimer. ‘Uniting the Tribes of Fluency to Form a Metacognitive Nation’. Personality and Social Psychology Review 13 (2009): 219–35. Amichai-Hamburger, Yair, and Katelyn McKenna. ‘The Contact Hypothesis Reconsidered: Interacting via the Internet’. Journal of Computer-Mediated Communication 11, no. 3 (2006): 825–43. Anderson, Elizabeth. ‘The Epistemology of Democracy’. Episteme 3, no. 1–2 (2006): 8–22. Anderson, Elizabeth. ‘Democracy: Instrumental vs. Non-Instrumental Value’. In Contemporary Debates in Political Philosophy, edited by Thomas Christiano and John Christman, 213–28. Oxford: Blackwell, 2009. Anderson, Elizabeth. The Imperative of Integration. Princeton, NJ: Princeton University Press, 2010. Anderson, Elizabeth. ‘Democracy, Public Policy, and Lay Assessments of Scientific Testimony’. Episteme 8 (2011): 144–64. Anderson, Elizabeth. ‘Reply to My Critics’. Symposia on Gender, Race, and Philosophy 9, no. 2 (2013): 1–8. Anderson, Richard. ‘The Place of the Media in Popular Democracy’. Critical Review 12, no. 4 (1998): 481–500. Asser, Martin. ‘What the Muhammad Cartoons Portray’. BBC, 2010. http://news.bbc. co.uk/1/hi/world/middle_east/4693292.stm. Bächtiger, Andre, and John Parkinson. Mapping and Measuring Deliberation. Oxford: Oxford University Press, 2019. Bächtiger, Andre, Simon Niemeyer, Michael Neblo, Marco Steenbergen, and Jurg Steiner. ‘Disentangling Diversity in Deliberative Democracy’. Journal of Political Philosophy 18, no. 1 (2010): 32–63. Badano, Gabriele, and Matteo Bonotti. ‘Rescuing (and Enriching) Public Reason Liberalism’s Accessibility Requirement’. Law and Philosophy 39 (2020): 35–69.

246 Bibliography Bagg, Samuel. ‘Can Deliberation Neutralize Power’. European Journal of Political Theory 17 (2018a): 257–79. Bagg, Samuel. ‘The Dispersion of Power’. American Political Science Review 112, no. 4 (2018b): 891–904. Baier, Annette. ‘Trust and Antitrust’. Ethics 96, no. 2 (1986): 231–60. Baker, Edwin. ‘Harm, Liberty, and Free Speech’. Southern California Law Review 70 (1996): 979–1020. Bakshy, Eytan, Solomon Messing, and Lada Adamic. ‘Exposure to Ideologically Diverse News and Opinion on Facebook’. Science 348 (2015): 1130–2. Baldwin, James, Emile Capouya, Lorraine Hansberry, Nat Hentoff, Langston Hughes, and Alfred Kazin. ‘The Negro in American Culture’. CrossCurrents 11, no. 3 (1961): 205–24. Baldwin, Lewis. ‘Malcolm X and Martin Luther King, Jr.: What They Thought About Each Other’. Islamic Studies 25, no. 4 (1986): 395–416. Barnidge, Matthew. ‘Exposure to Political Disagreement in Social Media Versus Face-toFace and Anonymous Online Settings’. Political Communication 34, no. 2 (2017): 302–21. Barsade, Sigal. ‘The Ripple Effect: Emotional Contagion and Its Influence on Group Behaviour’. Administrative Science Quarterly 47 (2002): 644–75. BBC. ‘9/11 Conspiracy Theories’. BBC, 29 August 2011. https://www.bbc.co.uk/news/ magazine-14665953. BBC. ‘The Saga of “Pizzagate” ’. BBC, 2016. https://www.bbc.co.uk/news/blogs-trending38156985. Bell, Daniel. The China Model. Princeton, NJ: Princeton University Press, 2016. Bell, Macalester. ‘Anger, Virtue, and Oppression’. In Feminist Ethics and Social and Political Philosophy, edited by Lisa Tessman, 165–83. London: Springer, 2009. Benhabib, Seyla. ‘Toward a Deliberative Model of Democratic Legitimacy’. In Democracy and Difference, edited by Seyla Benhabib, 67–94. Princeton, NJ: Princeton University Press, 1996. Ben-Ze’Ev, Aaron. ‘Emotions Are Not Mere Judgments’. Philosophy and Phenomenological Research 68, no. 2 (2004): 450–7. Berinsky, Adam. ‘Rumors and Health Care Reform’. British Journal of Political Science 47 (2017): 241–62. Billingham, Paul and Anthony Taylor. ‘A Framework for Analyzing Public Reason Theories’. European Journal of Political Theory (2020): 1–21. Blairy, Sylvie, Pedro Herrera, and Ursula Hess. ‘Mimicry and the Judgment of Emotional Facial Expressions’. Journal of Nonverbal Behavior 23, no. 1 (1999): 5–41. Blake, Michael. ‘Liberal Foundationalism and Agonistic Democracy’. Nomos 46 (2005): 230–43. Blitz, James. ‘Brexit and the “NHS Dividend” ’. Financial Times, 2018. https://www.ft.com/ content/1fcaf8b0-277e-11e8-b27e-cc62a39d57a0. Bohman, James, and Henry Richardson. ‘Liberalism, Democracy, and “Reasons that All Can Accept” ’. Journal of Political Philosophy 17, no. 3 (2009): 253–74. Bolsen, Tony, James Druckman, and Fay Lomax Cook. ‘Citizens’, Scientists’, and Policy Advisors’ Beliefs about Global Warming’. Annals of the American Academy 658 (2015): 271–95. Borgwald, Kristin. ‘Women’s Anger, Epistemic Personhood, and Self-respect’. Philosophical Studies 162, no. 1 (2012): 69–76. Bourgeois, Patrick, and Ursula Hess. ‘The Impact of Social Context on Mimicry’. Biological Psychology 77 (2008): 343–52. Bouricius, Terrill. ‘Why Hybrid Bicameralism is Not Right for Sortition’. In Legislature by Lot, edited by John Gastil and Erik Olin Wright, 313–35. London: Verso: 2019.

Bibliography  247 Bradshaw, Peter. ‘Blue is the Warmest Colour—Review’. The Guardian, 2013. http://www. theguardian.com/film/2013/nov/21/blue-is-the-warmest-colour-review. Brady, Michael. Emotional Insight. Oxford: Oxford University Press, 2013. Brandeis, Louis. ‘Opinion in Whitney v California, 274 US 357’, 1927. Brennan, Jason. Against Democracy. Princeton, NJ: Princeton University Press, 2016. Brettschneider, Corey. When the State Speaks, What Should It Say? Princeton, NJ: Princeton University Press, 2012. Brison, Susan. ‘The Autonomy Defense of Free Speech’. Ethics 108, no. 2 (1998): 312–39. Brown, Alexander. Hate Speech Law. Abingdon: Routledge, 2015. Brown, Etienne. ‘Propaganda, Misinformation, and the Epistemic Value of Democracy’. Critical Review 30 (2018): 194–218. Brownlee, Kimberley. ‘Freedom of Association: It’s Not What You Think’. Oxford Journal of Legal Studies 35, no. 2 (2015): 267–82. Brownlee, Kimberley. ‘Freedom of Association’. Stanford Encyclopedia of Philosophy, 2019. https://plato.stanford.edu/entries/freedom-association/. Brun, Georg, Ulvi Doguoglu, and Dominique Kuenzle. ‘Introduction: A New Role for Emotions in Epistemology?’. In Epistemology and Emotions, edited by Georg Brun, Ulvi Doguoglu, and Dominique Kuenzle, 1–32. Farnham: Ashgate, 2008. Campbell, Bradley, and Jason Manning. ‘The New Millennial “Morality”: Highly Sensitive and Easily Offended’. Time, 2015. http://time.com/4115439/student-protests-micro aggressions/. Cassam, Quassim. Conspiracy Theories. Cambridge: Polity Press, 2019. Catala, Amandine. ‘Democracy, Trust, and Epistemic Justice’. The Monist 98 (2015): 424–40. Center for Disease Control and Prevention. ‘Vaccines Do Not Cause Autism’. CDC, 2015. https://www.cdc.gov/vaccinesafety/concerns/autism.html. Chakravarti, Sonali. Sing the Rage. Chicago, IL: University of Chicago Press, 2014. Chambers, Simone. ‘Behind Closed Doors: Publicity, Secrecy, and the Quality of Deliberation’. Journal of Political Philosophy 12, no. 4 (2004): 389–410. Chambers, Simone. ‘Rhetoric and the Public Sphere’. Political Theory 37, no. 3 (2009): 323–50. Chambers, Simone. ‘Deliberation and Mass Democracy’. In Deliberative Systems, edited by Jane Mansbridge and John Parkinson, 52–71. Cambridge: Cambridge University Press, 2012. Chambers, Simone. ‘Balancing Epistemic Quality and Equal Participation in a System Approach to Deliberative Democracy’. Social Epistemology 31, no. 3 (2017): 266–76. Christiano, Thomas. ‘The Significance of Public Deliberation’. In Deliberative Democracy, edited by James Bohman and William Rehg, 243–78. Cambridge, MA: MIT Press, 1997. Christiano, Thomas. ‘Rational Deliberation among Experts and Citizens’. In Deliberative Systems, edited by John Parkinson and Jane Mansbridge, 27–51. Cambridge: Cambridge University Press, 2012. Clift, Tom. ‘Blue Is the Warmest Colour (Review)’. Moviedex, 2014. http://moviedex.com/ reviews/blue-is-the-warmest-colour-review/. Cogley, Zac. ‘A Study of Virtuous and Vicious Anger’. In Virtues and their Vices, edited by Kevin Timpe and Craig Boyd, 199–224. Oxford: Oxford University Press, 2014. Cohen, Joshua. ‘Deliberation and Democratic Legitimacy’. In The Good Polity: Normative Analysis of the State, edited by Alan Hamlin and Phillip Pettit, 17–34. New York: Blackwell, 1989. Cohen, Joshua. ‘Procedure and Substance in Deliberative Democracy’. In Democracy and Difference, edited by Seyla Benhabib, 95–119. Princeton, NJ: Princeton University Press, 1996.

248 Bibliography Cohen, Joshua. ‘Reflections on Deliberative Democracy’. In Contemporary Debates in Political Philosophy, edited by Thomas Christiano and John Christman, 247–63. Malden, MA: Wiley-Blackwell, 2009. Conover, Pamela, and Donald Searing. ‘Studying “Everyday Political Talk” in the Deliberative System’. Acta Politica 40 (2005): 269–83. Cook, John, Stephan Lewandowsky, and Ullrich Ecker. ‘Neutralizing Misinformation through Inoculation’. PLOS One 12 (2017): 1–21. Cooper, Joel, and Russell Fazio. ‘A New Look at Dissonance Theory’. Advances in Experimental Social Psychology 17 (1984): 229–66. Coplan, Amy. ‘Understanding Empathy’. In Empathy, edited by Amy Coplan and Peter Goldie, 3–18. Oxford: Oxford University Press, 2011. Cramer, Katherine. The Politics of Resentment. Chicago, IL: Chicago University Press, 2016. Cudd, Ann. Analyzing Oppression. Oxford: Oxford University Press, 2007. Curato, Nicole, and Marit Böker. ‘Linking Mini-publics to the Deliberative System: A Research Agenda’. Policy Sci 49 (2016): 173–90. Curato, Nicole, Marit Hammond, and John Min. Power in Deliberative Democracy. Cham: Palgrave Macmillan, 2019. Cyran, Olivier. ‘ “Charlie Hebdo”, Pas Raciste? Si Vous Le Dites. . .’ . Article 11, 2013. http:// www.article11.info/?Charlie-Hebdo-pas-raciste-Si-vous. Delgado, Richard. ‘Words That Wound: A Tort Action for Racial Insults, Epithets and Name Calling’. In Words That Wound, edited by Mari Matsuda, 89–110. Boulder, CO: Westview Press, 1993. Delgado, Richard, and Jean Stefancic. Understanding Words That Wound. Boulder, CO: Westview Press, 2004. Deonna, Julien, and Fabrice Teroni. The Emotions. New York: Routledge, 2012. Dilnot, Andrew. ‘UK Statistics Authority Statement on the Use of Official Statistics on Contributions to the European Union’. UK Statistics Authority, 2016. https://www. statisticsauthority.gov.uk/news/uk-statistics-authority-statement-on-the-use-of-officialstatistics-on-contributions-to-the-european-union/. Döring, Sabine. ‘Affective Perception and Rational Motivation’. Dialectica 61, no. 3 (2007): 363–94. Dotson, Kristie. ‘Tracking Epistemic Violence, Tracking Practices of Silencing’. Hypatia 26, no. 2 (2011): 236–57. Dottle, Rachael. ‘Where Democrats and Republicans Live in Your City’. FiveThirtyEight, 20 May 2019. https://projects.fivethirtyeight.com/republicans-democrats-cities/. Douglas, Heather. Science, Policy, and the Value-free Ideal. Pittsburgh, PA: University of Pittsburgh Press, 2009. Douglass, Frederick. My Bondage and My Freedom. Buffalo, NY: Miller, Orton, and Mulligan, 1855. Dowding, Keith. ‘Emotional Appeals in Politics and Deliberation’. Critical Review of International Social and Political Philosophy 21, no. 2 (2018): 242–60. Dryzek, John. Deliberative Democracy and Beyond: Liberals, Critics, Contestations. Oxford: Oxford University Press, 2000. Dryzek, John. Foundations and Frontiers of Deliberative Governance. Oxford: Oxford University Press, 2010. Dryzek, John, and Christian List. ‘Social Theory and Deliberative Democracy: A Reconciliation’. British Journal of Political Science 33 (2003): 1–28. Dworkin, Ronald. ‘Is There a Right to Pornography?’. Oxford Journal of Legal Studies 1 (1981): 177–212.

Bibliography  249 Dworkin, Ronald. ‘Foreword’. In Extreme Speech and Democracy, edited by Ivan Hare and James Weinstein, v–ix. Oxford: Oxford University Press, 2009. Ecker, Ullrich, Stephan Lewandowsky, Briony Swire, and Darren Chang. ‘Correcting False Information in Memory’. Psychonomic Bulletin & Review 18 (2011): 570–8. Edwards, Paul. ‘Global Climate Science, Uncertainty and Politics’. Science as Culture 8 (1999): 437–72. Elgin, Catherine. ‘Emotion and Understanding’. In Epistemology and Emotions, edited by Georg Brun, Ulvi Doguoglu, and Dominique Kuenzle, 33–50. Farnham: Ashgate, 2008. Ellison, Ralph. Invisible Man. New York: Vintage, 1952. Elster, Jon. ‘Strategic Uses of Argument’. In Barriers to Conflict Resolution, edited by Kenneth Arrow et al, 236–57. Stanford, CA: Stanford Center on Conflict and Negotiation, 1995. Elster, Jon. ‘The Market and the Forum’. In Deliberative Democracy, edited by James Bohman and William Rehg, 3–34. Cambridge, MA: MIT Press, 1997. Elster, Jon. ‘Deliberation and Constitution Making’. In Deliberative Democracy, edited by Jon Elster, 97–122. Cambridge: Cambridge University Press, 1998. Enoch, David. ‘Against Public Reason’. In Oxford Studies in Political Philosophy, Vol. 1, edited by David Sobel, Peter Vallentyne, and Steven Wall, 112–40. Oxford: Oxford University Press, 2015. Estlund, David. Democratic Authority. Oxford: Oxford University Press, 2008. European Commission. ‘A Multi-dimensional Approach to Disinformation’. Communications Networks, Content and Technology. Brussels: European Union, 2018. Fabre, Cécile. Cosmopolitan Peace. Oxford: Oxford University Press, 2016. Farrell, David, Jane Suiter, and Clodagh Harris. ‘  “Systematizing” Constitutional Deliberation: The 2016–2018 Citizens’ Assembly in Ireland’. Irish Political Studies 34, no. 1 (2019): 113–23. Ferguson, Amber. ‘Elizabeth Warren’s Message at Netroots Nation’. Huffington Post, 2015. http://www.huffingtonpost.com/entry/elizabeth-to-speak-at-netroots-nation-rally_ us_55a938f8e4b03f76c5ee2be2. Festinger, Leon. A Theory of Cognitive Dissonance. Stanford, CA: Stanford University Press, 1957. Financial Times Editorial Board. ‘Legislation against Fake News Is Open to Abuses’. Financial Times, 2019. https://www.ft.com/content/b1d78fc2-57b4-11e9-a3db1fe89bedc16e. Finnis, John. ‘Marriage: A Basic and Exigent Good’. The Monist 91, no. 3/4 (2008): 388–406. Fishkin, James. Democracy When the People Are Thinking. Oxford: Oxford University Press, 2018. Fishkin, James, and Janet Peck. ‘Online Deliberative Polling Gives Picture of Informed Public Opinion in Election’. Center for Deliberative Democracy (2020). cdd. stanford.edu. Fishkin, James, and Robert Luskin. ‘Experimenting with a Democratic Ideal’. Acta Politica 40 (2005): 284–98. Fiss, Owen. ‘What Should Be Done for Those Who Have Been Left Behind?’. In A Way Out: America’s Ghettos and the Legacy of Racism, 3–46. Princeton, NJ: Princeton University Press, 2003. Forst, Rainer. The Right to Justification. New York, NY: Columbia University Press, 2012. Fricker, Miranda. Epistemic Injustice. Oxford: Oxford University Press, 2007. Fricker, Miranda. ‘Epistemic Justice as a Condition of Political Freedom?’. Synthese 190, no. 7 (2013): 1317–32.

250 Bibliography Friedman, Howard, and Ronald Riggio. ‘Effect of Individual Differences in Nonverbal Expressiveness on Transmission of Emotion’. Journal of Nonverbal Behavior 6, no. 2 (1981): 96–104. Friedman, Jeffrey. ‘Public Ignorance and Democratic Theory’. Critical Review 12, no. 4 (1998): 397–411. Friedman, Marilyn. ‘John Rawls and the Political Coercion of Unreasonable People’. In The Idea of a Political Liberalism, edited by Victoria Davion and Clark Wolf, 16–33. Lanham, MD: Rowman & Littlefield, 2000. Frye, Marilyn. The Politics of Reality: Essays in Feminist Theory. Trumansburg, NY: The Crossing Press, 1983. Fung, Archon. ‘Recipes for Public Spheres’. Journal of Political Philosophy 11 (2003): 338–67. Fung, Archon. ‘Deliberation before the Revolution’. Political Theory 33, no. 2 (2005): 397–419. Funke, Daniel, and Daniela Flamini. ‘A Guide to Anti-misinformation Actions around the World’. Poynter, 2019. https://www.poynter.org/ifcn/anti-misinformation-actions/. Garrison, William Lloyd. ‘Preface’. In Narrative of the Life of Frederick Douglass, by Frederick Douglass, iii–xii. Boston, MA: Anti-Slavery Office, 1845. Garsten, Bryan. ‘The Rhetorical Revival in Political Theory’. Annual Review of Political Science 14 (2011): 159–80. Gastil, John, and Katherine Knobloch. Hope for Democracy. Oxford: Oxford University Press, 2020. Gastil, John, and Erik Olin Wright. Legislature by Lot. London: Verso, 2019. Gaus, Gerald. The Order of Public Reason. Cambridge: Cambridge University Press, 2010. Gaus, Gerald, and Kevin Vallier. ‘The Roles of Religious Conviction in a Publicly Justified Polity’. Philosophy & Social Criticism 35, no. 1–2 (2009): 51–76. Gelber, Katharine. ‘Reconceptualizing Counterspeech in Hate Speech Policy (with a Focus on Australia)’. In The Content and Context of Hate Speech, edited by Michael Herz and Peter Molnar, 198–216. Cambridge: Cambridge University Press, 2012. Gelber, Katharine, and Luke McNamara. ‘Evidencing the Harms of Hate Speech’. Social Identities 22, no. 3 (2015a): 324–41. Gelber, Katharine, and Luke McNamara. ‘The Effects of Civil Hate Speech Laws: Lessons from Australia’. Law & Society Review 49, no. 3 (2015b): 631–64. Gelfert, Axel. ‘Fake News: A Definition’. Informal Logic 38, no. 1 (2018): 84–117. Gentzkow, Matthew, and Jesse Shapiro. ‘Ideological Segregation Online and Offline’. Quarterly Journal of Economics 126, no. 4 (2011): 1799–839. Gibson, James. ‘Does Truth Lead to Reconciliation?’. American Journal of Political Science 48, no. 2 (2004): 201–17. Godlee, Fiona, and Harvey Marcovitch. ‘Wakefield’s Article Linking MMR Vaccine and  Autism Was Fraudulent’. thebmj, 2011. https://www.bmj.com/content/342/bmj. c7452.full. Goldie, Peter. The Emotions. Oxford: Oxford University Press, 2002. Goldman, Alvin. ‘Two Routes to Empathy’. In Empathy, edited by Amy Coplan and Peter Goldie, 31–44. Oxford: Oxford University Press, 2011. Goodin, Robert. Motivating Political Morality. Oxford: Wiley-Blackwell, 1992. Goodin, Robert. ‘Sequencing Deliberative Moments’. Acta Politica 40 (2005): 182–96. Goodin, Robert, and John Dryzek. ‘Deliberative Impacts: The Macro-political Uptake of Mini-publics’. Politics & Society 34, no. 2 (2006): 219–44. Goodin, Robert, and Kai Spiekermann. An Epistemic Theory of Democracy. Oxford: Oxford University Press, 2018.

Bibliography  251 Government Communication Service. ‘5 Trends in Leading-edge Communications’. UK Government Communication Service, 2018. https://gcs.civilservice.gov.uk/wp-content/ uploads/2018/10/5-Trends.pdf. Green, Jeffrey. The Eyes of the People. Oxford: Oxford University Press, 2010. Guerrero, Alexander. ‘Against Elections: The Lottocratic Alternative’. Philosophy & Public Affairs 42 (2014): 135–78. Guerrero, Laura. ‘ “I’m so Mad I Could Scream” ’. Southern Communication Journal 59 (1994): 125–41. Guess, Andrew, and Alexander Coppock. ‘Does Counter-attitudinal Information Cause Backlash?’. British Journal of Political Science 50, no. 4 (2020): 1–19. Gutmann, Amy, and Dennis Thompson. Democracy and Disagreement. Cambridge, MA: Belknap Press, 1996. Gutmann, Amy, and Dennis Thompson. Why Deliberative Democracy? Princeton, NJ: Princeton University Press, 2004. Habermas, Jürgen. The Theory of Communicative Action. Frankfurt am Main: Suhrkamp, 1981. Habermas, Jürgen. Between Facts and Norms. Cambridge, MA: MIT Press, 1996. Habermas, Jürgen. ‘Religion in the Public Sphere: Cognitive Presuppositions of the “Public Use of Reason” by Religious and Secular Citizens’. In Between Naturalism and Religion: Philosophical Essays, 114–48. Cambridge: Polity Press, 2008. Halley, Keaton. ‘How Gay Marriage Harms People’. Creation.com, 2017. https://creation. com/gay-marriage-harm. Halpern, Jodi, and Harvey Weinstein. ‘Rehumanizing the Other: Empathy and Reconciliation’. Human Rights Quarterly 26 (2004): 561–83. Hannon, Michael. ‘Empathetic Understanding and Deliberative Democracy’. Philosophy and Phenomenological Research 101, no. 3 (2020): 1–21. Hardin, Russell. Trust & Trustworthiness. New York: Russell Sage Foundation, 2002. Havekes, Esther, Michael Bader, and Maria Krysan. ‘Realizing Racial and Ethnic Neighborhood Preferences? Exploring the Mismatch between What People Want, Where They Search, and Where They Live’. Population Research and Policy Review 35, no. 1 (2016): 101–26. Hawley, Katherine. ‘Trust, Distrust, and Commitment’. Noûs 48, no. 1 (2014): 1–20. Heatherly, Kyle, Yanquin Lu, and Jae Kook Lee. ‘Filtering out the Other Side? Cross-cutting and Like-minded Discussions on Social Networking Sites’. New Media & Society 19 (2017): 1271–89. Heinze, Eric. ‘Hate Speech and the Normative Foundations of Regulation’. International Journal of Law in Context 9, no. 4 (2013): 590–617. Heinze, Eric. Hate Speech in Democratic Citizenship. Oxford: Oxford University Press, 2016. Hennig, Brett. ‘Who Needs Elections?’. In Legislature by Lot, edited by John Gastil and Erik Olin Wright, 301–12. London: Verso: 2019. Hertzberg, Benjamin. Chains of Persuasion. Oxford: Oxford University Press, 2018. Hess, Ursula, and Sylvie Blairy. ‘Facial Mimicry and Emotional Contagion to Dynamic Facial Expressions and Their Influence on Decoding Accuracy’. International Journal of Psychophysiology 40 (2001): 129–41. Hochschild, Arlie Russell. Strangers in Their Own Land: Anger and Mourning on the American Right. New York: New Press, 2016. Hochschild, Jennifer. ‘Creating Options’. In A Way Out: America’s Ghettos and the Legacy of Racism, 68–73. Princeton, NJ: Princeton University Press, 2003. Hoffman, Andrew. How Culture Shapes the Climate Change Debate. Palo Alto, CA: Stanford University Press, 2015.

252 Bibliography Holan, Angie. ‘PolitiFact’s Lie of the Year: “Death Panels” ’. Politifact, 2009. https://www. politifact.com/truth-o-meter/article/2009/dec/18/politifact-lie-year-death-panels. Holmes, Oliver Wendell. Dissent: Abrams v United (1919). Howard, Jeffrey. ‘Free Speech and Hate Speech’. Annual Review of Political Science 22 (2019a): 93–109. Howard, Jeffrey. ‘Terror, Hate, and the Demands of Counterspeech’. British Journal of Political Science 30, nos. 1–2 (2019b): 1–16. Huddy, Leonie. ‘The Group Foundations of Democratic Political Behavior’. Critical Review 30 (2018): 71–86. Huddy, Leonie, Lilliana Mason, and Lene Aaroe. ‘Expressive Partisanship’. American Political Science Review 109 (2015): 1–17. Huijer, Marli. ‘Storytelling to Enrich Democratic Debate: The Dutch Discussion on Embryo Selection for Hereditary Breast Cancer’. BioSocieties 4 (2009): 223–38. Hume, David. A Treatise of Human Nature. Oxford: Oxford University Press, 2009[1738]. Ingraham, Christopher. ‘The “Smoking Gun” Proving North Carolina Republicans Tried to Disenfranchise Black Voters’. The Washington Post, 29 July 2016. https://www. washingtonpost.com/news/wonk/wp/2016/07/29/the-smoking-gun-proving-northcarolina-republicans-tried-to-disenfranchise-black-voters/?utm_term=.89848a1fdff6. Isenberg, Nancy. White Trash. New York: Viking, 2016. Iyengar, Shanto, Gaurav Sood, and Yphtach Lelkes. ‘Affect, not Ideology: A Social Identity Perspective on Polarization’. Public Opinion Quarterly 76, no. 3 (2012): 405–31. Jacobs, Lawrence, Fay Lomax Cook, and Michael Carpini. Talking Together. Chicago, IL: University of Chicago Press, 2009. Jacobson, Louis. ‘ “What We Said Was, You Can Keep (Your Plan) If It Hasn’t Changed since the Law Passed” ’. PolitiFact, 2013. https://www.politifact.com/factchecks/2013/ nov/06/barack-obama/barack-obama-says-what-hed-said-was-you-could-keep/. Jaggar, Alison. ‘Love and Knowledge: Emotion in Feminist Epistemology’. Inquiry 32, no. 2 (1989): 151–76. Jerit, Jennifer, and Yangzi Zhao. ‘Political Misinformation’. Annual Review of Political Science 23 (2020). John, Stephen. ‘From Social Values to P-Values: The Social Epistemology of the Intergovernmental Panel on Climate Change’. Journal of Applied Philosophy 34 (2017): 157–71. Johnson, Hollyn, and Colleen Seifert. ‘Sources of the Continued Influence Effect’. Journal of Experimental Psychology 20, no. 6 (1994): 1420–6. Johnson, James. ‘Arguing for Deliberation’. In Deliberative Democracy, edited by Jon Elster, 161–84. Cambridge: Cambridge University Press, 1998. Jones, Karen. ‘Trust as an Affective Attitude’. Ethics 107, no. 1 (1996): 4–25. Jones, Karen. ‘Second-hand Moral Knowledge’. Journal of Philosophy 94, no. 2 (1999): 55–78. Kahan, Dan. ‘The Politically Motivated Reasoning Paradigm, Part 1’. In Emerging Trends in the Social and Behavioral Sciences, 1–16. Hoboken, NJ: Wiley & Sons, 2015. Kermode, Mark. ‘Blue Is the Warmest Colour—Review’, The Guardian, 27 February 2016. http://www.theguardian.com/film/2013/nov/24/blue-is-the-warmest-colour-review. King, Martin Luther. ‘I  Have a Dream’. Lincoln Memorial, Washington DC, 28 August 1963. https://americanrhetoric.com/speeches/mlkihaveadream.htm. Krause, Sharon. Civil Passions. Princeton, NJ: Princeton University Press, 2008. Krysan, Maria, and Sarah Moberg. ‘Trends in Racial Attitudes’. University of Illinois System, 2020. https://igpa.uillinois.edu/programs/racial-attitudes#section-8.

Bibliography  253 Kuhn, Thomas. The Structure of Scientific Revolutions. Chicago, IL: Chicago University Press, 1962. Lafont, Cristina. ‘Deliberation, Participation, and Democratic Legitimacy’. Journal of Political Philosophy 23, no. 1 (2015): 40–63. Lakatos, Imre. ‘Falsification and the Methodology of Scientific Research Programmes’. In Criticism and the Growth of Knowledge, edited by Imre Lakatos and Alan Musgrave, 91–196. Cambridge: Cambridge University Press, 1970. Landemore, Hélène. Democratic Reason. Princeton, NJ: Princeton University Press, 2013. Landemore, Hélène. ‘Inclusion Constitution-making: The Icelandic Experiment’. Journal of Political Philosophy 23, no. 2 (2015): 166–91. Landemore, Hélène. Open Democracy. Oxford: Oxford University Press, forthcoming. Langton, Rae. ‘Speech Acts and Unspeakable Acts’. Philosophy & Public Affairs 22 (1993): 293–330. Langton, Rae. ‘Beyond Belief: Pragmatics in Hate Speech and Pornography’. In Speech and Harm, edited by Ishani Maitra and Mary Kate McGowan, 72–93. Oxford: Oxford University Press, 2012. Langton, Rae. ‘Hate Speech and the Epistemology of Justice’. Law and Philosophy 10 (2014): 1–9. Langton, Rae. ‘Blocking as Counterspeech’. In New Work on Speech Acts, edited by Daniel Fogal, Daniel Harris, and Matt Moss, 144–62. Oxford: Oxford University Press, 2018a. Langton, Rae. ‘The Authority of Hate Speech’. Oxford Studies in Philosophy of Law 3 (2018b): 123–48. Langton, Rae, Sally Haslanger, and Luvell Anderson. ‘Language and Race’. In Routledge Companion to Philosophy of Language, edited by Gillian Russell and Fara Delia Graff, 753–67. Abingdon: Routledge, 2012. Lee, Carol. ‘Obama Reacts to “Death Panel” Claims’. Politico, 2009. https://www.politico. com/story/2009/08/obama-reacts-to-death-panel-claims-026149. Lenard, Patti. ‘The Decline of Trust, the Decline of Democracy?’. Critical Review of International Social and Political Philosophy 8, no. 3 (2005): 363–78. Lenard, Patti. Trust, Democracy, and Multicultural Challenges. University Park, PA: Pennsylvania State University Press, 2012. Lepoutre, Maxime. ‘Hate Speech Laws: Expressive Power Is Not the Answer’. Legal Theory, 2020. Lewandowsky, Stephan, Ulrich Ecker, Colleen Seifert, Norbert Schwarz, and John Cook. ‘Misinformation and Its Correction’. Psychological Science in the Public Interest 13 (2012): 106–31. Lewis, David. ‘Scorekeeping in a Language Game’. Journal of Philosophical Logic 8 (1979): 339–59. Lewis, Kristi. ‘When Leaders Display Emotion’. Journal of Organizational Behavior 21, no. 2 (2000): 221–34. Library of Congress, ‘Initiatives to Counter Fake News in Selected Countries’. Library of Congress (2019): 1–108. Lincoln, Abraham. ‘Letter to John A. Gilmer’. In The Collected Works of Abraham Lincoln, edited by Roy Basler, Vol. 4. New Brunswick, NJ: Rutgers University Press, 1860. List, Christian, and Mathias Koenig-Archibugi. ‘Can There Be a Global Demos?’. Philosophy & Public Affairs 38, no. 1 (2010): 76–110. Lister, Andrew. ‘Public Reason and Democracy’. Critical Review of International Social and Political Philosophy 11, no. 3 (2008): 273–89.

254 Bibliography López-Guerra, Claudio. Democracy and Disenfranchisement. Oxford: Oxford University Press, 2014. Lorde, Audre. ‘The Uses of Anger’. Women’s Studies Quarterly 25, no. 1/2 (1997): 278–85. Lovett, Frank, and Philip Pettit. ‘Neorepublicanism: A Normative and Institutional Research Program’. Annual Review of Political Science 12, no. 1 (2009): 11–29. Lukianoff, Greg, and Jonathan Haidt. ‘The Coddling of the American Mind’. The Atlantic, 2015. https://www.theatlantic.com/magazine/archive/2015/09/the-coddling-of-theamerican-mind/399356/. Lupia, Arthur. Uninformed. Oxford: Oxford University Press, 2016. Lupia, Arthur, and Mathew McCubbins. The Democratic Dilemma. Cambridge: Cambridge University Press, 1998. Luskin, Robert, Ian O’Flynn, James Fishkin, and David Russell. ‘Deliberating across Deep Divides’. Political Studies 62 (2014): 116–35. Lynch, Michael. Know-It-All Society. New York: W.W. Norton, 2019. MacKenzie, Michael, and Mark Warren. ‘Two Trust-based Uses of Minipublics in Democratic Systems’. In Deliberative Systems, edited by Jane Mansbridge and John Parkinson, 95–124. Cambridge: Cambridge University Press, 2012. MacKinnon, Catharine. Only Words. Cambridge, MA: Harvard University Press, 1993. Maddison, Sarah. ‘Relational Transformation and Agonistic Dialogue in Divided Societies’. Political Studies 63 (2015): 1014–30. Maitra, Ishani. ‘Subordinating Speech’. In Speech and Harm, edited by Ishani Maitra and Mary Kate McGowan, 94–118. Oxford: Oxford University Press, 2012. Maitra, Ishani, and Mary Kate McGowan. ‘Introduction and Overview’. In Speech and Harm, edited by Ishani Maitra and Mary Kate McGowan, 1–23. Oxford: Oxford University Press, 2012. Malcolm  X. The Ballot or the Bullet. Cleveland, OH, 1964. https://www.youtube.com/ watch?v=D9BVEnEsn6Y. Manin, Bernard. ‘On Legitimacy and Political Deliberation’. Political Theory 15, no. 3 (1987): 338–68. Mansbridge, Jane. ‘Everyday Talk in the Deliberative System’. In Deliberative Politics, edited by Stephen Macedo, 211–39. Oxford: Oxford University Press, 1999a. Mansbridge, Jane. ‘Should Blacks Represent Blacks and Women Represent Women? A Contingent “Yes” ’. Journal of Politics 61 (1999b): 628–57. Mansbridge, Jane, James Bohman, Simone Chambers, David Estlund, Andreas Føllesdal, Archon Fung, Cristina Lafont, Bernard Manin, and José Luis Martí. ‘The Place of Selfinterest and the Role of Power in Deliberative Democracy’. Journal of Political Philosophy 18, no. 1 (2010): 64–100. Mansbridge, Jane, James Bohman, Simone Chambers, Archon Fung, John Parkinson, Dennis Thompson, and Mark Warren. ‘A Systemic Approach to Deliberative Democracy’. In Deliberative Systems, edited by John Parkinson and Jane Mansbridge, 1–26. Cambridge: Cambridge University Press, 2012. Marin, Ludovic. ‘“FAKE NEWS!” (1/5)—Loi Contre La “Manipulation de l’information”: Un Combat Perdu d’avance?’. LCI, 2019. https://www.lci.fr/population/fake-news-1-5-loicontre-la-manipulation-de-l-information-un-combat-perdu-d-avance-2117268.html. Markovits, Elizabeth. ‘The Trouble with Being Earnest’. Journal of Political Philosophy 14, no. 3 (2006): 249–69. Maroh, Julie. ‘Le Bleu d’Adèle’. Coeurs-Forêts, 2013. http://www.juliemaroh.com/2013/05 /27/le-bleu-dadele/. Massey, Douglas, and Nancy Denton. American Apartheid. Cambridge, MA: MIT Press, 1993.

Bibliography  255 Matravers, Derek. Empathy. Cambridge: Polity Press, 2017. Matsuda, Mari. Words That Wound. Boulder, CO: Westview Press, 1993. McAdams, Richard. The Expressive Powers of Law. Cambridge, MA: Harvard University Press, 2015. McAuley, James. ‘France Weighs a Law to Rein in “Fake News,” Raising Fears for Freedom of Speech’. The Washington Post, 2018. https://www.washingtonpost.com/world/europe/ france-weighs-a-law-to-rein-in-fake-news-raising-fears-for-freedom-of-speech/ 2018/01/10/78256962-f558-11e7-9af7-a50bc3300042_story.html. McGary, Howard. ‘Forgiveness’. American Philosophical Association 26, no. 4 (1989): 343–51. McGowan, Mary Kate. ‘Oppressive Speech’. Australasian Journal of Philosophy 87 (2009): 389–407. McGowan, Mary Kate. ‘Responding to Harmful Speech’. In Voicing Dissent, edited by Casey Rebecca Johnson, 182–200. Abingdon: Routledge, 2018. McGowan, Mary Kate. Just Words. Oxford: Oxford University Press, 2019. McTernan, Emily. ‘Microaggressions, Equality, and Social Practices’. Journal of Political Philosophy 26, no. 3 (2018): 261–81. Mihai, Mihaela. Negative Emotions and Transitional Justice. New York: Columbia University Press, 2016. Mikkelson, David. ‘Will Garlic Cure Coronavirus?’. Snopes, 2020. https://www.snopes. com/fact-check/garlic-cure-coronavirus/. Mill, John Stuart. On Liberty. London: Penguin Classics, 2006[1859]. Mills, Charles. ‘White Ignorance’. In Race and Epistemologies of Ignorance, edited by Shannon Sullivan and Nancy Tuana, 11–38. Albany, NY: State University of New York Press, 2007. Milton, John. ‘The Areopagitica’. St Lawrence Institute, 1644. http://www.stlawrenceinstitute. org/vol14mit.html. Mouffe, Chantal. ‘Deliberative Democracy or Agonistic Pluralism?’. Social Research 66, no. 3 (1999): 745–58. Murray, Jessica. ‘Schools to Teach Pupils about Perils of Fake News and Catfishing’. The Guardian, 2019. https://www.theguardian.com/technology/2019/jun/26/schoolchildrenother to-get-online-safety-advice-on-catfishing-and-fake-news. Mutz, Diana. Hearing the Other Side. Cambridge: Cambridge University Press, 2006. National Health Service. ‘How to Treat Coronavirus Symptoms at Home’. National Health Service, 2020. https://www.nhs.uk/conditions/coronavirus-covid-19/what-to-do-ifyou-or-someone-you-live-with-has-coronavirus-symptoms/how-to-treat-coronavirussymptoms-at-home/. Naurin, Daniel. Deliberation Behind Closed Doors. Colchester: ECPR Press, 2007. Neblo, Michael, Keavin Esterling, Ryan Kennedy, David Lazer, and Anand Sokhey. ‘Who Wants to Deliberate—And Why?’. American Political Science Review 104, no. 3 (2010): 566–83. Nelson, Sara. ‘Katie Hopkins Reflects on Branding Migrants “Cockroaches” and ‘Feral Humans’. Huffpost, 2015. https://www.huffingtonpost.co.uk/2015/07/27/katie-hopkinsreflects-migrants-cockroaches-feral-humans_n_7877124.html. Nielsen, Laura Beth. ‘Power in Public: Reactions, Responses and Resistance to Offensive Public Speech’. In Speech and Harm, edited by Ishani Maitra and Mary Kate McGowan, 148–73. Oxford: Oxford University Press, 2012. Nussbaum, Martha. Upheavals of Thought. Cambridge: Cambridge University Press, 2001. Nussbaum, Martha. ‘Perfectionist Liberalism and Political Liberalism’. Philosophy & Public Affairs 39, no. 1 (2011): 3–45.

256 Bibliography Nussbaum, Martha. ‘Transitional Anger’. Journal of the American Philosophical Association 1, no. 1 (2015): 41–56. Nussbaum, Martha. Anger and Forgiveness. Oxford: Oxford University Press, 2016. Nyhan, Brendan, and Jason Reifler. ‘Displacing Misinformation about Events’. Journal of Experimental Political Science 2, no. 1 (2015): 81–93. Oakes, James. The Radical and the Republican. New York: W.W. Norton, 2007. O’Neill, John. ‘The Rhetoric of Deliberation: Some Problems in Kantian Theories of Deliberative Democracy’. Res Publica 8 (2002): 249–68. Owen, David, and Graham Smith, ‘Sortition, Rotation, and Mandate’. In Legislature by Lot, edited by John Gastil and Erik Olin Wright, 279–300. London: Verso: 2019. Page, Benjamin, and Robert Shapiro. The Rational Public. Chicago, IL: Chicago University Press, 1992. Palma, Stefania, Neil Munshi, and John Reed. ‘Singapore “Falsehoods” Law Shows Perils of Fake News Fight’. Financial Times, 2020. https://www.ft.com/content/e50eb042-3db311ea-a01a-bae547046735. Parkinson, John. Deliberation in the Real World. Oxford: Oxford University Press, 2006. Parvin, Phil. ‘Is Deliberative Democracy Feasible?’. The Monist 98 (2015): 407–23. Peacocke, Christopher. ‘Does Perception have a Nonconceptual Content?’. Journal of Philosophy 98 (2001): 239–64. Pennycook, Gordon, Tyrone Cannon, and David Rand. ‘Prior Exposure Increases Perceived Accuracy of Fake News’. Journal of Experimental Psychology 147, no. 12 (2018): 1865–80. Peoples, Steve, and Emily Swanson. ‘AP-GfK Poll: Republican Voters on Trump: No Compassion, No Problem’. AP-GfK, 2015. http://ap-gfkpoll.com/uncategorized/ our-latest-poll-findings-45. Pepp, Jessica, Eliot Michaelson, and Rachel Sterken. ‘What’s New About Fake News?’. Journal of Ethics and Social Philosophy 16, no. 2 (2019): 67–94. Pettigrove, Glen. ‘Meekness and “Moral” Anger’. Ethics 122, no. 2 (2012): 341–70. Pettit, Philip. ‘The Cunning of Trust’. Philosophy & Public Affairs 24, no. 3 (1995): 202–25. Pettit, Philip. On the People’s Terms. Cambridge: Cambridge University Press, 2012. Plato. The Republic. Translated by Desmond Lee. London: Penguin Classics, 2007[c. 380 BC]. Poama, Andrei. ‘Staying In or Moving Out? Justice and the Abolition of the Dark Ghetto’. European Journal of Political Theory 19, no. 1 (2020): 128–38. Polletta, Francesca, and John Lee. ‘Is Telling Stories Good for Democracy? Rhetoric in Public Deliberation after 9/11’. American Sociological Review 71 (2006): 699–723. Popper, Karl. The Open Society and Its Enemies, Vol. I. Oxford: Routledge, 1995[1945]. Posner, Richard. Law, Pragmatism, and Democracy. Cambridge, MA: Harvard University Press, 2003. Post, Robert. ‘Racist Speech, Democracy, and the First Amendment’. William and Mary Law Review 32 (1990): 267–328. Post, Robert. Constitutional Domains: Democracy, Community, Management. Cambridge, MA: Harvard University Press, 1995. Putnam, Robert. ‘Bowling Alone: America’s Declining Social Capital’. Journal of Democracy 6, no. 1 (1995): 65–78. Quayle, Dan. ‘The Quayle Quarterly’. Wall Street Journal. 8 June 1992. Quirke, Antonia. ‘Review—Blue Is the Warmest Colour’. Financial Times, 2013. http:// www.ft.com/cms/s/2/86a0b1cc-52af-11e3-8586-00144feabdc0.html#axzz3y4zPKKYZ. Quong, Jonathan. Liberalism without Perfection. Oxford: Oxford University Press, 2011.

Bibliography  257 Quong, Jonathan. ‘On the Idea of Public Reason’. In A Companion to Rawls, edited by Jon Mandle and David Reidy, 265–80. Oxford: Blackwell, 2013. Quong, Jonathan. ‘Public Reason’. Stanford Encyclopedia of Philosophy, 2017. https://plato. stanford.edu/entries/public-reason/. Rasmussen Reports. ‘58% Still Trust A Jury’s Verdict More Than A Judge’s’. Rasmussen Reports, 2014. http://www.rasmussenreports.com/public_content/lifestyle/general_ lifestyle/february_2014/58_still_trust_a_jury_s_verdict_more_than_a_judge_s. Rawls, John. Political Liberalism. New York: Columbia University Press, 1993. Rawls, John. ‘The Idea of Public Reason Revisited’. The University of Chicago Law Review 64, no. 3 (1997): 765–807. Rawls, John. Justice as Fairness: A Restatement. Cambridge, MA: Harvard University Press, 2001. Raz, Joseph. ‘Disagreement in Politics’. American Journal of Jurisprudence 43, no. 1 (1998): 25–52. Reilly, Katie. ‘Here Are All the Times Donald Trump Insulted Mexico’. Time, 2016. http:// time.com/4473972/donald-trump-mexico-meeting-insult/. Resnik, David and Kevin Elliott. ‘Value-entanglement and the Integrity of Scientific Research’. Studies in History and Philosophy of Science 75 (2019): 1–11. Richards, David. Free Speech and the Politics of Identity. Oxford: Oxford University Press, 1999. Richards, Robert, and Clay Calvert. ‘Counterspeech 2000: A New Look at the Old Remedy for “Bad Speech” ’. BYU Law Review 2000 (2000): 553–86. Rini, Regina. ‘Fake News and Partisan Epistemology’. Kennedy Institute of Ethics Journal 27 (2017): 43–64. Rothman, Lily. ‘When Spousal Rape First Became a Crime in the US’. Time, 2015. http:// time.com/3975175/spousal-rape-case-history/. Rowbottom, Jacob. ‘Lies, Manipulation and Elections—Controlling False Campaign Statements’. Oxford Journal of Legal Studies 32, no. 3 (2012): 507–35. Runciman, David. Political Hypocrisy. Princeton, NJ: Princeton University Press, 2008. Sampson, Robert. Great American City: Chicago and the Enduring Neighborhood Effect. Chicago, IL: Chicago University Press, 2012. Sanders, Lynn. ‘Against Deliberation’. Political Theory 25, no. 3 (1997): 347–76. Saul, Jennifer. ‘What Is Happening to Our Norms Against Racist Speech?’. Aristotelian Society Supplementary Volume 93, no. 1 (2019): 1–23. Savransky, Rebecca. ‘Giuliani: “Black Lives Matter” Phrase “Inherently Racist” ’. The Hill, October 2016. http://thehill.com/blogs/blog-briefing-room/news/287157-giuliani-blacklives-matter-inherently-racist. Scanlon, Thomas. ‘A Theory of Freedom of Expression’. Philosophy & Public Affairs 1, no. 2 (1972): 204–26. Scanlon, Thomas. What We Owe to Each Other. Cambridge, MA: Harvard University Press, 1998. Schaap, Andrew. ‘Agonism in Divided Societies’. Philosophy & Social Criticism 32, no. 22 (2006): 255–77. Schimmelfennig, Frank. ‘The Community Trap’. International Organization 55, no. 1 (2001): 47–80. Schulzke, Marcus. ‘The Social Benefits of Protecting Hate Speech and Exposing Sources of Prejudice’. Res Publica 22 (2016): 225–42. Schumpeter, Joseph. Capitalism, Socialism, and Democracy. Abingdon: Routledge, 2010 [1942].

258 Bibliography Schwartzman, Micah. ‘The Sincerity of Public Reason’. Journal of Political Philosophy 19, no. 4 (2011): 375–98. Scruton, Roger. Sexual Desire. London: Phoenix Press, 2001. Searing, Donald, Frederick Solt, Pamela Conover, and Ivor Crewe. ‘Public Discussion in the Deliberative System: Does It Make Better Citizens?’. British Journal of Political Science 37 (2007): 587–618. Sénat Vidéo. ‘Questions Au Gouvernement: David Rachline’, 2015. http://videos.senat.fr/ video/videos/2015/video30342.html. Seneca, L.A. ‘On Anger’. In Moral Essays. Translated by John Basore. London: Heinemann, 1928[c.45]. Sereno, Kenneth, Melinda Welch, and David Braaten. ‘Interpersonal Conflict’. Journal of Applied Communication Research 15 (1987): 128–43. Settle, Jaime. Frenemies. Cambridge: Cambridge University Press, 2018. Shapiro, Ian. The State of Democratic Theory. Princeton, NJ: Princeton University Press, 2003. Shapiro, Ian. Politics against Domination. Cambridge, MA: Belknap Press, 2016. Shaw, Rosalind. ‘Rethinking Truth and Reconciliation Commissions: Lessons from Sierra Leone’. United States Institute of Peace. Washington, DC, 2005. http://www.usip.org/ sites/default/files/resources/sr130.pdf. Shelby, Tommie. ‘Integration, Inequality, and Imperatives of Justice: A Review Essay’. Philosophy & Public Affairs 42, no. 3 (2014): 253–85. Shelby, Tommie. Dark Ghettos: Injustice, Dissent, and Reform. Cambridge, MA: Harvard University Press, 2016. Siegman, Aron Wolfe, Robert Anderson, and Tal Berger. ‘The Angry Voice’. Psychosomatic Medicine 52 (1990): 631–43. Silvermint, Daniel. ‘Rage and Virtuous Resistance’. Journal of Political Philosophy 25, no. 4 (2017): 461–86. Simpson, Robert Mark. ‘Dignity, Harm, and Hate Speech’. Law and Philosophy 32 (2013a): 701–28. Simpson, Robert Mark. ‘Un-ringing the Bell’. Australasian Journal of Philosophy 91 (2013b): 555–75. Sinaceur, Marwan, and Larissa Tiedens. ‘Get Mad and Get More than Even’. Journal of Experimental Social Psychology 42, no. 3 (2006): 314–22. Sleat, Matt. ‘Coercing Non-liberal Persons: Considerations on a More Realistic Liberalism’. European Journal of Political Theory 12, no. 4 (2013): 347–67. Slovic, Paul, Torbjorn Malmfors, Daniel Kewski, C.K.  Merz, Nancy Neil, and Sheryl Bartlett. ‘Intuitive Toxicology’. Risk Analysis 15 (1995): 661–75. Smith, Adam. The Theory of Moral Sentiments. Indianapolis, IN: Liberty Classics, 1982[1759]. Smith, Ben, and Byron Tau. ‘Birtherism: Where It All Began’. Politifact, 2011. https://www. politico.com/story/2011/04/birtherism-where-it-all-began-053563. Smith, Graham. ‘Deliberative Democracy and Mini-publics’. In Evaluating Democratic Innovations, ed. Brigitte Geissel and Kenneth Newton. Abingdon: Routledge, 2012. Smith, Graham, Robert Richards Jr., and John Gastil. ‘The Potential of Participedia as a Crowdsourcing Tool for Comparative Analysis of Democratic Innovations’. Policy and Internet 7, no. 2 (2015): 243–62. Smith, Kimberly. ‘Storytelling, Sympathy and Moral Judgment in American Abolitionism’. Journal of Political Philosophy 6, no. 4 (1998): 356–77. Somin, Ilya. ‘Voter Ignorance and the Democratic Ideal’. Critical Review 12, no. 4 (1998): 413–58.

Bibliography  259 Somin, Ilya. ‘Richard Posner’s Democratic Pragmatism and the Problem of Ignorance’. Critical Review 16, no. 1 (2004): 1–22. Somin, Ilya. ‘Deliberative Democracy and Political Ignorance’. Critical Review 22, no. 2–3 (2010): 253–79. Somin, Ilya. ‘Why Political Ignorance Undermines the Wisdom of the Many’. Critical Review 26, no. 1–2 (2014): 151–69. Somin, Ilya. Democracy and Political Ignorance. Stanford, CA: Stanford University Press, 2016. Spencer, Saranac. ‘Fake Coronavirus Cures, Part 2: Garlic Isn’t a “Cure” ’. FactCheck, 2020. https://www.factcheck.org/2020/02/fake-coronavirus-cures-part-2-garlic-isnt-a-cure/. Spragens, Thomas. Reason and Democracy. Durham, NC: Duke University Press, 1990. Srinivasan, Amia. ‘The Aptness of Anger’. Journal of Political Philosophy 26, no. 2 (2018): 123–44. Stanley, Jason. How Propaganda Works. Princeton, NJ: Princeton University Press, 2015. Steinel, Wolfgang, Gerben Van Kleef, and Fiecke Harinck. ‘Are You Talking to Me?!’. Journal of Experimental Social Psychology 44 (2008): 362–69. Steiner, Jurg. The Foundations of Deliberative Democracy. Cambridge: Cambridge University Press, 2012. Strossen, Nadine. ‘Interview with Nadine Strossen’. In The Content and Context of Hate Speech, edited by Michael Herz and Peter Molnar, 378–98. Cambridge: Cambridge University Press, 2012. Strossen, Nadine. HATE: Why We Should Resist It with Free Speech, Not Censorship. Oxford: Oxford University Press, 2018. Sunstein, Cass. ‘The Law of Group Polarization’. Journal of Political Philosophy 10, no. 2 (2002): 175–95. Sunstein, Cass. #Republic. Princeton, NJ: Princeton University Press, 2017. Swedish Civil Contingencies Agency. 2018. https://www.msb.se/RibData/Filer/pdf/28697.pdf. Talisse, Robert. Democracy and Moral Conflict. Cambridge University Press, 2009. Talisse, Robert. Overdoing Democracy. Oxford: Oxford University Press, 2019. Tappolet, Christine. ‘Values and Emotions’. In Morality and the Emotions, edited by Carla Bagnoli, 117–31. Oxford: Oxford University Press, 2011. Tiedens, Larissa. ‘Anger and Advancement versus Sadness and Subjugation’. Journal of Personality & Social Psychology 80, no. 1 (2001): 86–94. Tirrell, Lynne. ‘Genocidal Language Games’. In Speech and Harm, edited by Ishani Maitra and Mary Kate McGowan, 174–221. Oxford: Oxford University Press, 2012. Tirrell, Lynne. ‘Transitional Justice in Post-genocide Rwanda’. In Theorizing Transitional Justice, edited by Claudio Corraetti, Nir Eisikovits, and Jack Rotondi, 237–52. Farnham: Ashgate, 2015. Tirrell, Lynne. ‘Toxic Speech: Inoculations and Antidotes’. Southern Journal of Philosophy 56 (2018): 116–44. Tuck, Richard. ‘Active and Passive Citizens’. Tanner Lectures on Human Values, 2019. Tully, James. ‘Exclusion and Assimilation: Two Forms of Domination in Relation to Freedom’. Nomos 45 (2005): 191–229. Twohey, Megan. ‘Rudy Giuliani Lashes Out at Black Lives Matter’. New York Times, 2016. https://www.nytimes.com/2016/07/11/us/politics/rudy-giuliani-black-lives-matter. html?_r=0. UN. International Convention on the Elimination of All Forms of Racial Discrimination (1965). Usborne, Simon. ‘Katie Hopkins Has Just Written a Piece So Hateful That It Might Give Hitler Pause—Why Was It Published?’. Independent, 2015. http://www.independent. co.uk/voices/katie-hopkins-when-is-enough-enough-10186490.html.

260 Bibliography Vaidhyanathan, Siva. Anti-social Media. Oxford: Oxford University Press, 2018. Vallier, Kevin. ‘Against Public Reason Liberalism’s Accessibility Requirement’. Journal of Moral Philosophy 8 (2011): 366–89. Vallier, Kevin. ‘Public Justification versus Public Deliberation: The Case for Divorce’. Canadian Journal of Philosophy 45, no. 2 (2015): 139–58. Vallier, Kevin. ‘In Defence of Intelligible Reasons in Public Justification’. The Philosophical Quarterly 66, no. 264 (2016): 596–616. Valls, Andrew. Rethinking Racial Justice. Oxford: Oxford University Press, 2018. Van Beest, Ilja, Gerben Van Kleef, and Eric Van Dijk. ‘Get Angry, Get out’. Journal of Experimental Social Psychology 44 (2008): 993–1002. Van der Linden, Sander, Anthony Leiserowitz, Seth Rosenthal, and Edward Maibach. ‘Inoculating the Public against Misinformation about Climate Change’. Global Challenges 1, no. 2 (2017): 1–7. Waldron, Jeremy. Law and Disagreement. Oxford: Oxford University Press, 1999. Waldron, Jeremy. The Harm in Hate Speech. Cambridge, MA: Harvard University Press, 2012. Walker, Shaun. ‘Vladimir Putin: Gay People at Winter Olympics Must “Leave Children Alone” ’. The Guardian, 2014. http://www.theguardian.com/world/2014/jan/17/vladimirputin-gay-winter-olympics-children. Weigel, David. ‘Three Words that Republicans Wrestle With’. Washington Post, 2016. https://www.washingtonpost.com/politics/three-words-that-republicans-wrestle-withblack-lives-matter/2016/07/12/f5a9dfdc-4878-11e6-90a8-fb84201e0645_story. html?utm_term=.0bd766ba0ceb. Weinstock, Daniel. ‘Compromise, Pluralism, and Deliberation’. Critical Review of International Social and Political Philosophy 20, no. 5 (2017): 636–55. Welch, Susan, Lee Sigelman, Timothy Bledsoe, and Michael Combs. Race & Place: Race Relations in an American City. Cambridge: Cambridge University Press, 2001. Wesolowska, Elzbieta. ‘Social Processes of Antagonism and Synergy in Deliberating Groups’. Swiss Political Science Review 13, no. 4 (2007): 663–80. West, Robin. ‘Liberty, Equality, and State Responsibilities’. Brooklyn Law Review 79, no. 3 (2014): 1031–45. White House. ‘Connecting America’. The White House—President Barack Obama, 2013. https://obamawhitehouse.archives.gov/connect-america. White, Stuart. ‘Freedom of Association and the Right to Exclude’. Journal of Political Philosophy 5 (1997): 373–91. Williams, Melissa. ‘The Uneasy Alliance of Group Representation and Deliberative Democracy’. In Citizenship in Diverse Societies, edited by Will Kymlicka and Wayne Norman, 124–52. Oxford: Oxford University Press, 2000. Willsher, Kim. ‘Abandoned’ French Working Class Ready to Punish Left’s Neglect by Voting for Far Right’. The Guardian, 2015. http://www.theguardian.com/world/2015/ mar/22/alienated-french-working-class-vote-far-right-claims-analyst. Wood, Thomas, and Ethan Porter. ‘The Elusive Backfire Effect’. Political Behavior 41 (2019): 135–63. Wubben, Maarten, David De Cremer, and Eric Van Dijk. ‘How Emotion Communication Guides Reciprocity’. Journal of Experimental Social Psychology 45 (2009): 987–90. Young, Iris Marion. ‘Communication and the Other: Beyond Deliberative Democracy’. In Democracy and Difference, edited by Seyla Benhabib, 120–36. Princeton, NJ: Princeton University Press, 1996. Young, Iris Marion. Inclusion and Democracy. Oxford: Oxford University Press, 2000. Young, Iris Marion. Responsibility for Justice. Oxford: Oxford University Press, 2011. Zollman, Kevin. ‘The Epistemic Benefit of Transient Diversity’. Erkenntnis 72 (2010): 17–35.

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. 9/11  106, 110 see also conspiracy theories abortion  160–1, 166–7 accountability and group cognition  158–9, 163–4, 186 and hypocrisy  140–4 and lottocracy  228 and public contestation  2–3, 15, 19–22, 29, 109, 136–7, 188 see also domination Achen, Christopher  158–64, 166–9, 171, 173, 185 affective polarisation  6–7, 9–10, 132–5, 184–5, 192–3, 208–9 see also goodwill Affordable Care Act death panel rumour  111, 116, 119 Obama’s false statement regarding  106, 109 see also misinformation agonism 5 Anderson, Elizabeth  22–3, 146, 171, 184, 194–7, 201, 208 Anderson, Richard  218–19 anger and contagion  61–2 and revenge  52–3, 56–7, 67–8 and trust  55, 72–4, 76–80, 144–51 counterproductivity objection to  8, 51, 54–9, 81–2 defined 51–4 epistemic cost of  69–72 epistemic value of  59–74 formal and informal sites of  77–8 gendered dimension of  73 intrinsic value of  56–8 networked communication of  73–4 misdirected  52, 69–71, 150–1 systemic role of  74–81 transitional  52–4, 60, 67 see also Douglass, Frederick; King, Martin Luther; Malcolm X argument  2, 38 abuses of  46, 70–1

contrasted with narrative  38–41 defined 38 exclusionary character of  38–40 Baldwin, James  49, 53–4, 60–1, 169 Baldwin, Lewis  79–80 bans on hate speech  8, 84–6, 90–2 absence of in the United States  87 and the ‘cessation of serious controversy’ 112–13 deterrence function of  100–1 risk of misuse  96–7, 112 salience of  99–101 their cost to freedom  90–2 bans on misinformation  107, 110–14 and intentions  113–14 chilling effect of  111–13 compared to bans on hate speech  111–13 examples of  111 risk of misuse  111–12 Bartels, Larry  158–64, 166–9, 171, 173, 185 beliefs see judgements Bell, Daniel  232–4 Benhabib, Seyla  18–19, 45 Berinsky, Adam  116, 119, 176n.62 Birtherism  106, 121, 124 see also conspiracy theories Black Lives Matter  96–7 Bota, Zimasile  153 Brennan, Jason  158–9, 162–3, 216, 229–32 Brettschneider, Corey  93–5 Brexit  106, 108, 111, 114–15, 117–18, 120–1 see also misinformation Brownlee, Kimberley  197–8 Bush, George W.  156 Cassam, Quassim  110, 121 Catala, Amandine  138–9 Catholics  138, 168n.38 Center for Disease Control  115–16, 119 Chakarvarti, Sonali  145–6, 148, 150 Chambers, Simone  134n.11, 221–2, 226 Chaplinsky v. New Hampshire 87n.12 Citizens’ Assembly  224–6

262 Index Citizens’ Initiative Review  224–5, 225n.36, 227 civility duty of  18–19, 29, 33 and trust  134–6 see also public reason climate change and group identity  173–9, 182 misinformation about  106–9 pre-emptive messaging about  124 see also group cognition; IPCC; misinformation; motivated reasoning; testimony coercion  4, 17, 47–8, 84, 90–2, 96, 196 cognitive dissonance see hypocrisy Condorcet Jury Theorem  163–4 Connor, ‘Bull’  147–8 conspiracy theories difficulty of countering  9, 121, 124–5 examples of  106 see also misinformation contagion see emotion, anger context of evaluation defined  58, 74–5 of anger  75–7 continued influence effect see misinformation, stickiness of Cook, John  124 counterspeech and authority  8, 93–6, 99, 114–15, 123–6, 152 and hate speech  8, 85–6, 92–9, 101–4, 151–4 and marketplace of ideas metaphor  92–3 and misinformation  9, 107, 114–25 and persuasion  95, 152 and salience  97–9, 101–4, 115–20 contrasted with bans  84, 90, 96–7 diachronic conceptions of  122–5 positive and negative conceptions of  8–9, 101–4, 117–22, 152, 240, 242 risk of backfire  97–9, 116–17 risk of misuse  96–7 state-driven conception of  93–7 unfairness of  93–4 see also hate speech; misinformation COVID-19  106–7, 114, 118 see also misinformation Cramer, Katherine  146, 160, 166, 171 cross-cutting discourse see discourse, ideal of inclusive Cudd, Ann  165 Danish cartoons  83–4 see also hate speech death panels see Affordable Care Act

deliberation defined  1, 4 structured see deliberative polling; minipublics see also deliberative democracy; discourse deliberative polling  138–9, 208–9, 210n.67, 224–6 as an integrative strategy  208–9, 210n.67 as a trust-building tool  138–9 epistemic success of  208–9, 225 see also minipublics deliberative democracy and argument  38–41 and Deliberation Day  209n.64 and emotion  39–47, 50 see also anger, emotion and representation  164–6, 217–18, 222 as a system see systemic approach to discourse defined  1–2, 132–3, 216–18 pessimism about  2–3, 214–15 see also deliberation; discourse demagogue see Trump, Donald Democratic Party  134–5, 160, 166, 171, 173, 192–3, 195 democracy and public discourse  1–2 and voting  1 see also deliberative democracy; elite democracy; representation Denton, Nancy  200–2 dignity and counterspeech  8, 93–6, 99 and hate speech  8, 88–90 assurance of  88–9, 91 defined 88 disagreement see division discourse anti-domination function of  2, 15, 19–22, 24–5, 37, 47–8, 90, 109, 157, 188, 239 as a system see systemic approach to discourse defined 3–4 elite 221–2 epistemic function of  1, 15, 41–5, 51, 108–9, 157, 163–4, 188 see also anger, epistemic value of; emotion, epistemic value of; narrative, as an epistemic tool homogeneous  188–9, 191–3 ideal of inclusive  1–2, 188, 214–17, 239 non-ideal approach to  2–3, 15, 53–4, 131–2, 239

Index  263 presuppositions of  9–12, 131–2, 214–15 quantity of  134n.11, 219 trust-building function of  9–10, 138–54 see also norms of discourse discursive norm see norm of discourse disinformation 109–10 see also misinformation disrespect  8, 47, 83–4 see also hate speech distrust see trust, lack of division affective see goodwill, lack of; trust, lack of and inequality  22–5, 28–9, 191–2 epistemic  22–5, 61, 165, 170–2, 193 ethical  19–22, 35–7, 165, 175–8, 193 partisan see partisanship racial  22–3, 35–6, 134–5, 146, 171, 192 see also racism social and spatial see fragmentation urban/rural  146, 175–6, 192 see also rural consciousness dogmatism and group cognition  10–11, 171–2, 179–85, 193 epistemic value of  11, 180–5 in science  11, 180–2 domination and hate speech  90–1 and misinformation  109 and public discourse see discourse, anti-domination function of and the shared reasons constraint  19–22, 24–5, 27–30, 33–4, 37, 44, 47–8 defined  2, 20 of the unreasonable  27–8, 90–1 Dotson, Kristi  23–4 Douglass, Frederick interaction with Lincoln  79 use of anger  53–4, 62, 65–7, 72–4

emotion and group identity  162, 169–70 contagious quality of  41–2, 61–2 epistemic costs of  71–2 epistemic value of  41–2, 59–74, 169 exclusion of  39–41 in public discourse  2–4, 37–47, 50–1 misdirected  46–7, 52, 69–71 nonconceptual content of  64–5 salience function of  42, 63, 169 see also anger; deliberative democracy; narrative empathy see perspective-taking environment  175–9, 182 see also climate change epistocracy 229–34 and ideal theory  231–3 and the China Model  232–3 as a response to ignorance  229 defined 229 undermined by ignorance  230–2 equality as a shared reason  35–6 colour-blind conception of  35–6 expressed in speech  102–4 in opportunity for discursive participation 1 in standing  88; see also dignity in voting power  1 see also inequality European enlargement  141 expertise abuses of  46, 70–1 and lottocracy  234–7 the problem of identifying  124, 173, 226–7, 229–31, 234–6 see also testimony; motivated reasoning

echo chamber see fragmentation education  94–5, 117–18, 123 elections and political ignorance  219–20 misinformation during  106, 110–11, 121–2 see also ignorance; misinformation elite democracy and elite discourse  221–2 and non-ideal theory  219 defined 217–19 as a response to ignorance  219 vulnerability to ignorance  219–23 see also democracy; Schumpeter Ellison, Ralph  41, 49, 60–1, 82 Elster, Jon  140–3

Fabre, Cécile  144, 146 Facebook and social fragmentation  209–11 misinformation on  110 scale of political discourse  209–10 strategies for integrating  211–12 fact-checking limited effectiveness of  114–17 as negative counterspeech  117 see also counterspeech; misinformation fake news  109–10 see misinformation feminism  10–11, 23–4, 59–60, 73, 97–8, 164–6 fighting words  87 see also bans on hate speech; hate speech

264 Index Fishkin, James  208–9, 223–6 see also deliberative polling fluency 115–16 see also misinformation; stickiness of fragmentation as a source of inequality  191–2 as a source of distrust and ill-will  184, 192–3 epistemic cost of  22–4, 171–2, 184, 192–3 as a problem for public discourse  184, 188–93 residential  11–12, 22–3, 170–1, 184, 188–9, 202 social  11–12, 184, 188–9 see also integration freedom as non-domination see domination and slavery  35, 65–7 of association  11–12, 195–203 of speech  90–2, 96–7, 103–4, 111–13 see also bans on hate speech; bans on misinformation; counterspeech; hate speech; marketplace of ideas Fricker, Miranda  23–4, 65 Friedman, Jeffrey  156 Friedman, Marilyn  27 Frye, Marilyn  73 Fung, Archon  236–7 Galela, Lendiso Richard Ndumo  145–9 Gandhi, Mohandas  55–6, 77–8 Garrison, William Lloyd  62, 65–6, 72–4 Gastil, John  227 Giuliani, Rudolph  96 Gelber, Katharine  93–4, 101 genetic fallacy  168 goodwill dependence on trust  133–7 lack of  9–10, 72–3, 132–5, 176, 184, 192–3 necessary for fruitful discourse  72–4, 133–8, 192–3 strategies for rebuilding see trust-building relation to social and spatial fragmentation  184, 192–3 and the system of public discourse  10, 73–4, 135–8 see also trust; trust-building; Truth and Reconciliation Commissions groups and emotion  162, 169–70 arbitrariness of membership in  161–3, 168 cognition of see group cognition cultural and social  168n.38 defined  165, 168–73 divisions between see goodwill, lack of; trust, lack of

perspective see group perspective representation 164–8 see also fragmentation; partisanship group cognition applied to technical questions  167, 173–9 as a problem for public discourse  160–4 as a solution for public discourse  167, 185–7 as part of human nature  163, 185 defined 158–61 dogmatic  171–2, 179–85 epistemic criticisms of  161–3, 169–74, 179–80 epistemic value of  164–85 see also group perspective group perspective and anger  169–70 and narrative  41–5 and privilege  22–5, 170–2, 179–80, 182 defined 165 epistemic costs of  170–2, 179–80 epistemic value of  180–5 relevance to scientific and technical matters  167, 174–9 role of emotion in  162–3, 169–70 see also group cognition; perspective-taking Green, Jeffrey  218–19, 221 Goodin, Robert  143 Guerrero, Alex  227–8, 235–6 Gutmann, Amy  18–19, 50, 84 Habermas, Jurgen  39–40 harm of hate speech see hate speech causal and constitutive dimensions of  89 hate speech and domination  90–2 and ideal theory  89 and trust  151–4 defined 86–8 extensiveness of  83–4, 91 harmfulness of  88–90 publicity of  87 salience of  97–104 see also bans on hate speech; counterspeech Havekes, Esther  202 Heinze, Eric  100–1 hermeneutical injustice  23–5, 65, 71–2 heuristic see information shortcuts; Lakatos Hochschild, Arlie Russell  69–70 Hume, David  41–2, 61 hypocrisy 140–4 see also sincerity, accountability

Index  265 ideal and non-ideal theory  2–6, 9, 15–16, 131–2, 138–40, 214–15, 240–1 see also discourse; elite democracy; epistocracy; hate speech; political minimalism, public reason; shared reasons constraint; trust-building identity see groups ignorance about political issues  10, 156–7 about race  22–4, 71–2, 171, 192 and anger  69–72 and group cognition  157–63, 166–7, 185–7 and hypocrisy  143–4 as a problem for elite democracy  219–23 as a problem for epistocracy  230–4 as a problem for lottocracy  234–7 as a problem for inclusive discourse  156–60, 163–4, 167, 185–7, 214–17 rationality of  156 robust 215–17 see also information shortcuts inclusion see discourse, ideal of inclusive inductive risk  174–5 inequality and segregation  190–2 and speaker authority  93 expressed in speech  83–4, 86–8 see also hate speech inference, see argument information shortcuts  157–60, 163–4, 185–7 see also ignorance; group cognition injustice as the object of anger  52 hidden 71–2 hermeneutical  23–5, 65 systemic 68 understood as an assault on dignity  88–90 inoculation 122 see also counterspeech, diachronic conceptions of integration and freedom of association  11–12, 195–203 as a trust-building tool  135, 138, 185, 192–4 as a tool for promoting inclusive discourse 191–5 costs to historically oppressed groups  199–207, 212–13 defined and illustrated  193–5 differentiation across sites  207–13 networked integration  203–7 online 209–12 of the workplace  208 racial  189–90, 194–5, 199–203 residential  185, 194–6, 199–213

viewed from a systemic perspective  203–13 see also fragmentation International Criminal Court  139 International Panel on Climate Change  177–8 is/ought fallacy  173 John, Stephen  177–8 Johnson, James  133–4 judgments political, defined  160 dependent on information shortcuts  157–60, 163–4 driven by group identity see group cognition justification see deliberation, public reason, shared reasons constraint Kahan, Dan  173–4 King, Martin Luther and Malcolm X  53, 55, 79–80 appeals to shared reasons  20, 29 ‘I have a dream’ speech  103 repudiation of anger  52–3, 55–6, 76, 78, 80 Martin Luther King Day  104 Knobloch, Katherine  227 knowledge based in group perspectives see group perspective, epistemic value of social dispersion of  1, 165, 239 see also anger, epistemic value of; emotion, epistemic value of; epistocracy; ignorance; discourse, epistemic function of Koenig-Archibugi, Matthias  139 Krause, Sharon  17–18, 20, 26, 33–5, 40–1, 44–5 Ku Klux Klan  88–9, 103–4 see also hate speech; racism Lafont, Cristina  226–7 Lakatos, Imre  180–1 Landemore, Hélène  221–2 Langton, Rae  98n.47 Le Pen, Marine  46 legal restrictions on hate speech see bans on hate speech on misinformation see bans on misinformation Lenard, Patti  143 Lewandowsky, Stephan  115–16, 119–20, 124 Lincoln, Abraham  29, 57, 60, 79 List, Christian  139 López-Guerra, Claudio  227–8, 234 Lorde, Audre  49, 59–61, 63 lottocracy and expertise  234–7 as a response to robust ignorance  227–8, 234

266 Index lottocracy (Cont.) defined  227–8, 234 undemocratic character of  228 undermined by robust ignorance  234–7 see also Guerrero; López-Guerra love as alternative to anger  51, 78–80 disagreement over meaning  36, 42–3 Luskin, Robert  138–9, 223 MacKenzie, Michael  227 Maitra, Ishani  92–3 Malcolm X  53 and King  53, 55, 79–80 use of anger  53–4, 79–80 ‘The Ballot or the Bullet’  67–9, 77 Mandela, Nelson  55–6 Mansbridge, Jane  78, 164–6, 183 Massey, Douglas  200–2 marital rape  24 marketplace of ideas  92–3 see also counterspeech McGowan, Mary Kate  92–3, 98–9 McNamara, Luke  101 microaggression  23–4, 41, 165 Mill, John Stuart  88n.15, 92n.31, 95n.40 Mills, Charles  22–3, 35–6 minipublics as a response to non-ideal conditions  223–9 complementary role of  225–7 defined and illustrated  223–5 replacement role of  225, 227–9 see also deliberative polling; lottocracy misinformation and intentions  109–10, 113–14 and ‘sharing’  116n.27 as a source of domination  109 as an epistemic threat  108–9 contrasted with hate speech  108–9 defined 108–10 examples of  106–7, 109–10 salience of  115–21 scale of  107 stickiness of  9, 114–20 see also bans on misinformation; conspiracy theories; counterspeech; fake news monitoring 143–4 monuments  93–4, 104 see also counterspeech motivated reasoning  10–11, 120, 158–60, 173–9 see also group cognition; Kahan, Dan; values Mutz, Diana  188, 197–8, 208 narrative and emotion  7–8, 38, 41–2, 144–5

and impartiality  45–6 and shared reasons  7–8, 16, 41–7 as a trust-building tool  144–51 as an epistemic tool  7–8, 38–9, 41–4 defined 38–9 omission from the deliberative ideal  39–41 see also anger; testimony National Health Service see Brexit; misinformation national holidays  93–4, 104 see also counterspeech neighbourhoods see fragmentation, residential; integration, residential; self-segregation networks epistemic analysis of  181, 205–7 of public discourse  39, 73–4, 137–8, 183, 188, 197–8, 203–7 see also systemic approach to public discourse norm of discourse defined 3–4 enforcement of  4, 17, 84, 107 summary of norms defended  7–9, 131–2, 239–40 Nussbaum, Martha  52–7, 59–60, 74–8, 81, 145 Nyhan, Brendan  119–20 Oakes, James  62, 79–80 Obama, Barack  106, 108–9, 114, 121, 156–7 see also Affordable Care Act; Birtherism; TARP online speech digital divide  212 fragmentation of  188–9, 209–11 integration of  194, 211 online counterspeech  118–19 online misinformation  107 see also misinformation see also social media; Facebook participatory budgeting  236–7 partisanship as a source of political judgment  158–61 driven by social group identity  160, 162–3 emotional in nature  158, 162–3 partisan divisions  134–5, 173, 192–3, 199 see also Democratic Party; Republican Party; Tea Party Parvin, Phil  133–6, 138 perspective see group perspective perspective-taking  38, 41–2, 61–5, 72–4 Pettit, Philip  20, 147 Pettigrove, Glen  54, 60, 71–2 political minimalism and non-ideal theory  214–15

Index  267 as a response to robust ignorance  215–16 defined 215 undermined by robust ignorance  216–17, 228–9, 237–8 see also elite democracy, epistocracy, lottocracy, Schumpeter, Shapiro Pizzagate 106 see also conspiracy theories Plessy v. Ferguson 35 polarisation affective see goodwill, lack of belief 35 police violence  1, 96, 153 political ignorance see ignorance Posner, Richard  3, 218–19 power arbitrary  2, 15, 19–22, 67, 90 see also domination coercive see coercion protests  1, 96, 147–8 see also anger Protestants 138 Prout, William  180 proviso see public reason public reason  4–5, 17–19, 25–30, 34–5, 39–40 and deliberation  4–5 and ideal theory  4–5, 28–30 and reasonableness  25–30 burdens of judgment  26 proviso  18n.8, 34–5, 34n.54 see also civility; Rawls; shared reasons constraint public sphere see systemic approach to discourse publicity and hypocrisy  140–1 and the harm in hate speech  87–9, 98–9 and the identifiability of hate speakers  151–4 compatible with secrecy  143n.49 see also salience Quong, Jonathan  18–19, 29–30, 40 rage see anger racism  46, 83–4, 91, 96, 103, 199–200 and abuse or harassment  23–4, 165, 169–70, 199–201 and colour-blindness  35–6 systemic quality of  67–9 see also Black Lives Matter; hate speech; Ku Klux Klan; slavery; stereotypes Rapid Response Unit  118–19 rationalism  39–40, 162, 169 Rawls, John  4–5, 17–19, 26, 29–30, 33–5, 39–40, 134, 136 see also public reason reason 15n.1 see also rationalism

reasonableness see public reason, and reasonableness referendum  114, 224–6 Reifler, Jason  119–20 religion  17–18, 20–1 representation and deliberation  164–6, 217–18, 222 and elite democracy  217–19, 221–2 group-based 164–6 Republican Party  134–5, 160, 173, 192–3, 199 rhetoric see argument; emotion; narrative; trust Rini, Regina  109–10, 162 robust ignorance see ignorance Rowbottom, Jacob  114 rural consciousness  166, 172 salience in anger  63, 169–70 see also anger, epistemic value of in emotional representations, see emotion, saalience function of of bans see bans on hate speech, salience of of counterspeech see counterspeech, and salience of hate speech see hate speech, salience of of misinformation see misinformation, salience of same-sex relationships and hermeneutical injustice  24 and homophobia  22, 24, 36 and marriage  21, 24, 35–6, 43 in parliamentary debates  43 media representations of  42–3 Sampson, Robert  204–5 Schimmelfennig, Frank  141 Schulzke, Marcus  151–4 Schumpeter, Joseph  218–19 science role of dogmatism in  180–2 role of values in  174–9 scientists as a source of expert testimony  114–15, 117–18, 174–9 effects of values on standards of  175 see also Center for Disease Control; climate change; COVID-19; IPCC; vaccines segregation see fragmentation self-segregation possible benefits of  199–200 preference for mixed neighbourhoods  202 tension with integration  195–6, 200 see also fragmentation; integration Seneca 54 Settle, Jaime  209–11 sexual harassment  23, 65, 165–6

268 Index Shapiro, Ian  133–4, 138, 218–19 shared reasons constraint and acceptability  17–19 and accessibility  19 and domination  19–22, 24–5, 27–30, 33–4, 37, 44, 47–8 and ideal theory  16, 28–30, 39 constituency of  17, 25–30, 44 defined 16–19 exclusiveness of  22–5 intelligibility 19 moral, not legal  17 presupposes trust  134–5 sites of  17, 33–7, 136 see also public reason; religion; Rawls Shelby, Tommie  190, 196, 199–201, 208 silencing 123–4 Simpson, Robert  98–9 sincerity and deliberation  136–8, 142–3 and trust  137 as a widespread public commitment  143 see also hypocrisy sites of discourse see systemic approach to discourse slavery  35, 53, 60, 62, 65–7 see also Douglass, Frederick Smith, Graham  236 social media  11, 31, 107, 110–11, 116n.27, 195–6, 209–12 see also Facebook; online speech social structure  165 Somin, Ilya  216, 220 Southern realignment  160, 171 speech see discourse speech-act(s) causal and constitutive dimensions of  89 difficulty of reversing  97–9 silencing of  93, 123–4 Srinivasan, Amia  56–8, 75–6 standing 88–9 see also dignity Stanley, Jason  99 state and counterspeech see counterspeech and bans see bans on hate speech; bans on misinformation stereotypes  73, 98, 103, 192 Supreme Court  18n.6, 33, 35 systemic approach to discourse and anger  74–81 and dogma  180–5, 205–7, 214, 240–1 and integration  32, 203–15

and minipublics  226–7 and networked communication  73–4, 137–8, 203–7 and norms  32–3, 136 and the shared reasons constraint  33–7, 136 and trust or goodwill  73–4, 76–81, 135–8 defined 30–3 division of labour within  31–4, 78–81, 135–6 variety of sites or arenas  31 TARP (Troubled Asset Relief Program)  156–7 Taubira, Christiane  43, 94–5 Talisse, Robert  193 Tea Party  160, 166 testimony as a trust-building tool  144–51 as an information shortcut  157–60, 163, 167 scientific  114–15, 117–18, 173–9 under conditions of uncertainty  174–5 see also narrative; scientists; information shortcuts Thompson, Dennis  18–19, 50, 84 threats 87 ties, weak and close  197–8, 210 Tirrell, Lynne  122 transitional justice see trust building; Truth and Reconciliation Commissions; Tutu Trump, Donald and criticism of Black Lives Matter  96 and misinformation  109–10 and misdirection of anger  46, 69 and use of hate speech  83–4 trust and anger  55, 72–4, 76–81 see also trust-building, role of anger in and counterspeech  120 and risk  146–8, 152–4 and the system of discourse  135–8 and trust-building see trust-building lack of  9–10, 72–3, 133–5, 144–5, 184, 192 warranted and unwarranted  152–3, 214 see also goodwill; trust-building; Truth and Reconciliation Commissions trust-building discursive 138–55 non-discursive  133, 135, 138, 185, 191–5 non-ideal approach to  138–40 role of anger in  144–51 role of hate speech in  151–4 role of hypocrisy in  140–4 role of narrative in  144–51 see also trust; Truth and Reconciliation Commissions

Index  269 Truth and Reconciliation Commissions  55, 77, 144–51, 153 Tutu, Desmond  55, 77–8 Twitter 255 see also Facebook; online speech;  social media vaccines  106–9, 114–16, 119 see also misinformation values epistemic 174–7 ethical see division, ethical relevance to technical and scientific matters 173–9 see also motivated reasoning

Virginia v. Black 87n.12 vote  1, 228–9 see also elections Waldron, Jeremy  84–5, 87–91, 112–13 Warren, Mark  227 Williams, Melissa  164 Wood, Carmita  65 see also sexual harassment well-ordered society  29, 88–9 Young, Iris Marion  5, 39, 84, 136–7, 164–6, 168–73, 190–2, 196, 200 Zollman, Kevin  181, 205–7