Freedom of expression

This article surveys the classic and contemporary literature on the nature and limits of freedom of expression (or free speech). It begins by surveying the main philosophical justifications for free speech, before moving to consider the two most discussed topics in the free speech literature: hate speech and pornography. The article offers some brief reflections on the large number of arguments which have been offered on these topics. Three newer battlegrounds for free speech are examined at the end: no platforming, fake news and online shaming.

itself does and those that concentrate on its further effects. Section 4 examines longstanding debates on pornography and free speech, illustrating the various ways in which feminists have characterized the harm that pornography causes to women. Section 5 offers a short overview of new debates in the free speech literature, namely those on no-platforming, fake news, and online public shaming, while Section 6 briefly concludes.

| THEORIES OF FREE SPEECH
Free speech has been defended as a special liberty-distinct from other liberties and/or more robust than liberty in general (Kendrick, 2017)-on different grounds. It is said to be vital to protect the diversity and pluralism liberal societies value; an adjunct of liberal toleration and means of promoting tolerant attitudes; a safety valve that enables public grievances to be raised peacefully, or a route to civic order and stability more generally; and a bulwark against the slippery slope of governments' tendency to arrogate ever greater legal and policy powers to themselves. Relatedly, it can be defended via a reasonable distrust towards governments making correct decisions on when to regulate or not regulate speech: given that distrust, it is safest to have a constitutional right to free speech (Schauer, 1989). While there have been attempts to develop feminist (Williams, 2009) and Marxist (Heinze, 2018a(Heinze, , 2018b accounts of free speech, we focus here on the most prominent liberal views, given its status as a key liberal principle. J. S. Mill's (2006) famous argument that widespread freedom of speech is necessary to realise the truth in human affairs-since we can only gain justified confidence in our views in the open clash of debate-remains influential in the literature but has been subject to sustained criticism. For a start, like all consequentialist views, Mill's argument does not account for our intuition that free speech is valuable in itself, not simply as an instrument (e.g., see Greenawalt, 1989), as in the autonomy and democracy views we consider in a moment.
Furthermore, Mill's position has been regarded as over-intellectualised (Barendt, 2005, pp. 9-10;Haworth, 1998, pp. 24-32), ignoring the fact that much speech is not really concerned with truth-for example football chants, picketing strikers, or graffiti that stake out a gang's territory (McKinnon, 2006, pp. 123-4). If consequences matter, it also needs to be explained why truth should be prioritised over other potential effects of free speech, such as the hurt and distress caused by hate speech. Mill does not neglect the fact that speech can harm, as in his famous example of someone inciting a mob outside the house of a corn-dealer thought to be responsible for starving them (Mill, 2006, p. 64), or less obviously harms such as false advertising or malicious invasions of privacy (Riley, 2005).
However, for Mill free speech and the discovery of truth serve our happiness of a higher kind as progressive beings committed to deliberation, and their loss is generally more serious than any harms potentially resulting from speech. Moreover, such harms may be short-lived since with time people are generally able to identify true opinions and reject false ones (Schauer, 2012). 2 Mill's view is often associated with the 'marketplace of ideas' metaphor (Abrams v United States, 1919) but this has been vigorously questioned (Gordon, 1997;Howard, 2019a, p. 96;Simpson, 2013, p. 722;Sparrow & Goodin, 2001;Waldron, 2012, pp. 155-6). A free market for speech today gives us fake news on social media (discussed below) and dangerous political populism (Barendt, 2005, p. 9). Where truth is valued, speech is usually highly regulated to foster its emergence, in ways antithetical to the economic free market. For example, academic journals will not publish articles containing claims that are clearly wrong or misguided, contrary to Mill's view that false claims should be allowed to be aired for the truth to emerge (Alexander, 2005, p. 128;Post, 2000;cf. Simpson, 2020). A more charitable interpretation of Mill's argument holds that it is the process of truth discovery through speech that is valuable as it exercises and realises our critical deliberative capacities for reasoning (Brink, 2008;cf. Miles, 2012).
In addressing the view that autonomy grounds free speech, it is helpful to differentiate between speaker-and listener-centred theories, 3 and also between a procedural right to autonomy and the substantive ideal of a life which includes critical deliberation. A procedural, speaker-based view will claim that viewpoint-based restrictions on 2 of 13 -BONOTTI AND SEGLOW free speech usurp an individual's own judgement in choosing what to express to others; Baker's (2011) account contains elements of this view. Scanlon (1972), Strauss (1991), and Dworkin (1996, p. 200) in different ways advance a procedural, listener-based view claiming that individuals are sovereign over what they come to believe based on what they read, see, and hear. For the state to pre-empt their judgement, for example on the basis that hateful, seditious or other dangerous speech will incite individuals to commit harms, fails to respect their right and responsibility to determine for themselves how to act on the basis of the speech they receive (but cf. Howard, 2019b). One common criticism of this argument is that much speech does not consist of propositional content apt for individuals rationally to evaluate: many instances of hate speech, pornography, fake news, and deceptive advertising fall into this category (cf. Scanlon, 1979), like some of the examples we just directed against Mill. Another criticism is that individuals who value their autonomy would be reasonably concerned to protect it from threats by others who are incited to endanger it, and thus the argument would licence lesser scope for free speech than its more absolutist proponents imagine (Brison, 1998a;pp. 326-9;Howard, 2019a, p. 97).
Defending free speech based on substantive autonomy involves the claim that the self-determination at its core is necessary for individuals' self-realisation in some morally valuable way (see Redish, 1982). Baker (1997Baker ( , 2011 defends a speaker-based substantive position that emphasises how autonomy involves disclosing one's authentic beliefs to a social world shared with others. Individuals 'must be permitted to act in and sometimes affect the world […] by trying to persuade or criticize others' (Baker, 1997, p. 992). A listener-based substantive approach would claim instead that access to diverse, competing views in a plural society is necessary for autonomous individuals to review and endorse only those views they value.
It has been argued that the autonomy argument promotes a sectarian view that is reasonably rejectable in political liberal terms (Cohen, 1993, p. 222). It can also be objected that this approach is insufficiently discriminating: if individuals themselves determine what is valuable, then any more stringent protection granted to 'highvalue'-for example political or artistic-forms of speech over 'low-value' kinds will need to invoke some further claim, for example that politics and culture especially foster our deliberative capacities. A third criticism is that substantive autonomy is over-inclusive: since autonomy involves myriad life choices and there is plausibly some right to it, it is unclear why freedom of speech, as just one incident of autonomy, requires its own special right (Ladenson, 1997;Schauer, 2015;cf. Kendrick, 2017). Seana Shiffrin's (2011) recent thinker-based argument for free speech is a version of the autonomy argument that involves sparer assumptions than the substantive view and has attracted critical interest (e.g., Barendt, 2019;Scanlon, 2011). The core of Shiffrin's view is that we cannot understand the unique content of our own minds without free speech; those contents only come to fruition through speech (cf. Gilmore, 2011;Macklem, 2008). Shiffrin also argues that our interest in being recognised by others as persons with our own views is impossible without unimpaired communication between persons (Shiffrin, 2011, pp. 289-97), besides other specific human interests she enumerates.
The democracy-based defence of free speech appeals to the conceptual and institutional connections between those two values. Though an exclusively listener-based version of the democratic view is rarely advanced, it can be argued that citizens require access to reliable information and the means to critically assess each other's views to exercise their democratic rights responsibly. The speaker-based democratic view comes in more than one variety. Dworkin (2009) and Weinstein (2011) argue that it is illegitimate for governments to enact laws that affect citizens' vital interests unless they have had the opportunity to speak out in favour of or against them. In a similar vein, Post (2011) maintains that citizens cannot reasonably regard themselves as authors of the law unless they have had the substantive opportunity to influence the course of public discussion. Another democracy-based account is that defended by Sunstein (1993), who emphasises the role of high-value political and artistic speech in democratic deliberation. When it comes to this kind of speech, Sunstein argues, only viewpoint-and content-neutral time, place and manner restrictions are justified, while there could be greater regulation of low-value speech.
The democratic defence of free speech fits naturally with deliberative interpretations of the democratic ideal.
Redish (2013), however, argues for a more adversarial interpretation, which emphasises how each democratically BONOTTI AND SEGLOW -3 of 13 minded citizen has her own interests that free speech helps her to realise. By contrast, Estlund's (2009) epistemic interpretation of democracy, which sees it as the political arrangement most conducive to arriving at cogent or well-founded views, combines the democratic ideal with the insight of Mill's consequentialist view.
The democratic view can explain how certain apparent restrictions on free speech are in fact means to promote it. For example, this approach might support regulating the speech of powerful actors (e.g., via limits on campaign advertising) and fostering the speech of relatively marginalised groups (e.g., via guaranteed air time) (cf. Redish, 2001). Yet one objection to the democratic view may be that it cannot defend the free speech rights of resident non-citizens who lack formal voting rights. Moreover, while the argument's political focus offers stringent protection for political speech and broader social commentary that is most valuable or vulnerable to external interference, it may find it hard to accommodate modes of intuitively 'high-value' speech such as scientific speech or speech between members of a closed religious order.

| HATE SPEECH
Hate speech is perhaps the most debated topic in the contemporary free speech literature, with the contrast between US First Amendment protection and laws against hate speech in other states providing practical import to that debate. The rise of online hate speech (e.g. on social media) -instantaneous, rapidly dispersed, and often anonymous (Brown, 2018) -has made the discussion more urgent, not least because so much free speech censorship is now carried out by unaccountable corporations.
Hate speech maligns or vilifies people based on their group identity: race, ethnicity, religion, or sexual orientation, and other categories besides. As a form of conduct, the conceptual core of hate speech is arguably its assault on the civic standing of those it targets, where it exploits their already marginal or vulnerable status (Bonotti & Seglow, 2021, pp. 32-4). This definition helps explain why hate directed towards more privileged members of society is not (generally) hate speech and also helps make the analytical distinction between hate speech and nonhateful offensive speech, such as that directed at religious majorities. While hate speech attacks people's identities, offensive speech is directed at their beliefs (Bonotti & Seglow, 2019).
In specifying the harms of hate speech, it is helpful to distinguish between approaches that centre on what hate speech itself does and those that look at its further effects (this is very roughly the difference between its illocutionary and perlocutionary force).
In the former category belong those writers who see hate speech as a form of subordinating speech, for example one that helps instil the idea of racial inferiority in its victims' minds (Matsuda, 1989, p. 233), undermines their self-respect (Lawrence, 1987, p. 351), or is an affront to their dignity (Delgado, 1982). Related to subordination, hate speech has been seen as a form of discrimination, as when a 'Whites Only' sign was hung outside a firm's restroom door (EOOC v Tyson Foods, Inc., 2006), or as a form of stigmatising speech, or one that spreads harmful stereotypes about some group. Thus in Beauharnais v Illinois (1952), which concerned a racist pamphlet, the US Supreme Court ruled against speech which exposed African Americans to 'contempt, derision, or obloquy'. Hate speech has also been seen as a form of silencing because of the way in which, if they have sufficient social authority, hate speakers can derank their victims in an illocutionary sense (Maitra, 2012), or help maintain a social climate in which certain minority groups are not listened to or not trusted, what Fricker (2007) calls 'testimonial injustice'.
Silencing is clearly relevant to the autonomy-based and democratic defences of free speech.
Hate speech can also be considered a case of group defamation, one that attacks the social reputation of those it targets. This strategy has proved attractive for a number of hate speech scholars (e.g., Matsuda 1989), and is at the core of the influential account of hate speech provided by Jeremy Waldron (2012). According to Waldron, hate speech is harmful and should be regulated because it attacks the equal civic status of members of vulnerable minorities and undermines the public good of 'assurance', 'the 'pervasive, diffuse, general, sustained, and reliable underpinning of people's basic dignity and social standing, provided by all to and for all' (Waldron, 2012, p. 93). Hate speech targets vulnerable minorities and assaults their civic dignity by leveraging (and compounding) their already disadvantaged status; it could not gain purchase on members of privileged groups (Waldron, 2012, p. 237, note 13; see also Bonotti & Seglow 2021, pp. 33-5). Waldron's argument has been criticised on the grounds that the equal civic status of minorities could be protected via laws against discrimination, violence, harassment and other hate crimes, as well as through policies aimed at improving such groups' social and economic circumstances, without the need for hate speech laws (Brown, 2015;Seglow, 2016;Simpson, 2013).
As far as its effects (or perlocutionary force) are concerned, a standard way of understanding the harm of hate speech is that it can incite violence, public disorder, and a climate of hatred, and that in the long term it may result in injustice and discrimination against vulnerable minorities (e.g., Brown, 2008). While the short-term effects of hate speech may be straightforward to ascertain, its longer-term consequences can be difficult to compute. The presence of various intervening factors and possible counterfactual claims makes it hard to identify hate speech as a significant cause of long-term injustices, and a speaker's responsibility may reduce as a result (cf. Ten, 1980). Similar problems attend the view that hate speech harms the health of its victims (Brown, 2015, pp. 56-7), since their ailments may result from other social and economic injustices they suffer (Heinze, 2016;Strossen, 2018).
This raises a more general methodological question related to the study of hate speech. When normative arguments regarding the regulation of hate speech appeal to its harmful effects, it is important that those who advance them engage with relevant empirical research in the area (Simpson, 2019). Some scholars maintain that this engagement has been limited. According to Eric Heinze, for example, '[t]he problem with countless prohibitionist writings is that they start with those empirically demonstrable harms of immediate, interpersonal situa-tions…but then extrapolate straightforwardly from them to a purely rhetorical empiricism, lacking any empirical references, about equivalent harms putatively caused by hatred expressed within public discourse' (Heinze, 2016, p. 126). However, this claim seems exaggerated. Some of the theoretical literature on hate speech has in fact engaged with empirical research (e.g., Brown, 2015;Gelber & McNamara, 2013. Perhaps a more plausible claim, then, is that the social scientific literature struggles to provide conclusive evidence of hate speech's distinctive contribution to the harms its victims suffer, given that they are often subject to other injustices too (Brown, 2015, pp. 56-58;Simpson, 2019, p. 90).
One charge often levelled against hate speech laws (even when they are justified) is that they may be too vague, subjective, and overbroad, or that they may assign enforcing authorities too much discretion in their implementation (Strossen, 2018), including the power to silence marginalised groups (2018, pp. 81-88). Partly for this reason, some have defended the idea of counterspeech, an alternative to formal regulation. Counterspeech involves victims of hate speech or other citizens speaking back to hate-mongers in order to assert their rights and equal civic status, whether through marches and demonstrations, or on social media. Corey Brettschneider (2012) argues that the state itself should engage in counterspeech via its public officials and by supporting civil society associations that engage in it. Yet counterspeech may be difficult when the hate speaker is anonymous or much more powerful than the victim (Brown, 2015, p. 261), or ineffective given vulnerable minorities' marginalized status.
When it comes to the regulation of hate speech, empirical research which examines its causes may be particularly relevant. These include intergroup bias (Leyens et al., 2007 p. 140;Roussos & Dovidio, 2020) as well as the shock and trauma that can result from events such as financial crises, pandemics, emergencies, disasters, wars, and terrorist attacks (e.g., Benesch, 2014;Chetty and Alathur, 2018;Hinton, 2010

| PORNOGRAPHY
The regulation of pornography is a longstanding issue for free speech scholars but the explosion of online pornography in the past two decades has made it more topical and urgent. 4 A reasonable normative definition is that 'pornography is sexually explicit material designed to produce sexual arousal in consumers that is bad in a certain way' (West, 2012a; see also Watson, 2010, pp. 535-6).
Early debates on pornography often adopted a legal moralist (Devlin, 1968) position, regarding it as a threat to society's 'moral fabric', or conceptualised it through the categories of the 'offensive' or the 'obscene'. Contemporary feminist theorists focus instead on the harm pornography causes to women. Four categories of harm can usefully be distinguished. First, pornography is said to be responsible for sexual violence and other harms (e.g., harassment) perpetrated against women, a view pioneered by the feminist writer Catharine MacKinnon (e.g., 1993; see also Dworkin, 1993;Scoccia, 1996). This view seems to be supported by recent meta-analyses of the relationship between pornography and sexual violence, which have found a positive correlation (Hald et al., 2010;Wright et al., 2016). However, not all studies reach the same conclusion. More importantly, correlation does not entail causation: perhaps some men are more likely to view violent pornography because they are already predisposed to harm women, rather than vice versa. Thus a more plausible claim is a probabilistic one: pornography may increase the likelihood of violence and other harms against women, though it is hardly the only cause (Eaton, 2007, pp. 696-7).
In some religiously conservative societies, for example, little pornography circulates but harm against women may be rife. As we argued in relation to hate speech, it is important that philosophers working in this area pay attention to developments in the social scientific literature on the link between pornography and sexual violence (and other harms), as that may well have implications for their normative arguments.
Second, pornography is said to harm the women involved in its production. For example, women may be tricked, trafficked, or blackmailed into pornography, or harmed via so-called 'revenge porn'. Furthermore, even when pornographic actors apparently endorse their choice, many of them may be manipulated and/or not be fully aware of what they are becoming involved in, and this plausibly erodes the moral force of the consent principle. We also need to consider issues such as the availability of alternative economic options, consent manufactured by patriarchy, and the burden of stigma for women who have worked in the pornography industry.
Third, it is often argued that pornography subordinates women in society more generally. MacKinnon pioneered this claim; one dimension of it is that the circulation of pornography bears some causal responsibility for the myriad harms society visits on women including harassment, discrimination, inequality of opportunity, domination, lack of political voice, and sexist attitudes more generally. While many other feminists have followed MacKinnon in arguing that pornography especially contributes to reinforcing gender inequality, some have contested this claim.
Cynthia Stark (1997), for example, argues that Christian doctrine contains an equally damaging view of women as servants of men. MacKinnon (1993) was the first to make the conceptual claim that pornography constitutes the harm of subordinating women, and thus is a form of conduct akin to hate speech. In the much-discussed Anti-Pornography Civil Rights Ordinance that she drafted with Andrea Dworkin, pornography was defined as 'the graphic sexually explicit of subordination of women'. 5 (The Ordinance was struck down by US courts on the grounds that censoring a message of subordination violated First Amendment viewpoint neutrality). The subordination view has been developed with considerable sophistication by recent philosophers. On Langton's (1993Langton's ( , 2009) influential speech act theory, pornography is an illocutionary act of subordination that ranks women as inferior to men, by portraying them in a demeaning way, deprived of critical rights and powers, and one that objectifies them, by treating them merely as means for men's pleasure.
Langton's claim has been much discussed. It is not clear, for example, that pornography (always) includes the intention to subordinate. Furthermore, it is unclear whether and how the producers of pornography possess the requisite speaker authority necessary to derank women compared, for example, with legislators who have formal authority (e.g., lawmakers in apartheid South Africa who deprived black citizens of the vote). Perhaps systemic and 6 of 13 -BONOTTI AND SEGLOW embedded patriarchal norms are sufficient to confer this authority or perhaps, as Maitra (2012) has suggested in the context of racist hate speech, pornography's authority as a speech act comes from men and women's acquiescence to it, i.e., their general failure to speak back. As an alternative to Langton's view, Mary Kate McGowan (2019) has recently reformulated the subordination hypothesis in a philosophically sophisticated way through the idea of 'conversational exercitives', implicit norm-enacting moves in speech practices that change what is salient or permissible in that practice. Thus, for example, if male co-workers have a conversation in which one degrades a female colleague, that makes it permissible to degrade female colleagues in their future conversational practice.
Conversational exercitives do not require intention or authority in the way Langton's illocutionary subordination does. However, McGowan's argument is liable to the objection that if the harmful actions rendered permissible do not ensue, then no subordination has occurred, which seems counterintuitive.
Closely related to subordination, a fourth harm is the way pornography silences women, again a claim pioneered by MacKinnon (1992). Silencing may occur in a straightforward causal sense where pornography helps create a hostile environment that makes women reluctant to speak out e.g., about sexual harassment. Beyond that, Hornsby and Langton (1998;cf. Langton, 2009;Maitra, 2009;West, 2003) maintain that silencing occurs through disabling women's speech in an illocutionary sense by interfering with their communicative intentions, for example their intention to say no to sex. Widespread pornography consumption by men means that women's speech often misfires, failing to gain uptake, a phenomenon that may also occur with victims of racist hate speech (West, 2012b). McGowan (2009McGowan ( , 2019 argues that women's rejection of men's sexual advances also requires that men accept women's authority to refuse them, and pornography's objectification of women may erode the recognition of that authority, even if uptake is secured. Silencing can occur too if men believe women's refusals are insincere or fail to recognise their true feelings; on McGowan's norm-governed theory, pornography makes these types of actions permissible and therefore more likely (2019, pp. 150-5).
Despite these harms, some liberals maintain that pornography is a form of speech that seeks to alter the moral environment in some way and therefore merits protection (Dworkin, 1993), a version of the democratic view; that by censoring pornography governments fail to respect citizens' individual autonomy (e.g., Dworkin, 1981; but see Langton, 1990 for a critique); or that governments cannot reliably distinguish harmful pornography that requires censorship from other kinds.

| NEW CHALLENGES
In recent years new issues, many driven by rapid developments in social media, have entered current debates on free speech. Here we briefly mention three.

| No-platforming
Universities have been deeply involved in free speech battles in recent years, reflecting wider culture wars (Leaker, 2020;Titley, 2020). Faculty with objectionable views, decolonised curricula, and safe spaces for 'snowflake' students faced with views that, according to their critics, challenge their identities (Fish, 2019, pp. 65-108;Corlett, 2018;cf. Cross & Richardson-Self, 2020) have all been foci of debate. Significant too is no-platforming, that is, barring controversial individuals such as far-right politicians and professional provocateurs, from speaking at university campuses. No-platforming is highly controversial. In their influential book The Coddling of the American Mind, for instance, Lukianoff and Haidt (2018) argue that no-platforming can be traced back to a failure, in Western liberal democracies, to instil certain forms of psychological resilience in post-Millennial generations. By contrast, some scholars have defended the practice by arguing that university inquiry is different from public debate as it demands systematic, rigorous, impartial, and evidence-based deliberation, rather than the expression of any BONOTTI AND SEGLOW -7 of 13 viewpoint no matter how prejudiced or ill-informed (Post, 2017;cf. Fish, 2019, pp. 68-72). In defending a version of this argument, Simpson and Srinivasan (2018; see also Simpson, 2020) emphasize how academics routinely engage in content discrimination, by reviewing research, appointing staff on academic merit, setting the curriculum, grading students' work, and so on, thus de facto sanctioning inequality of status among competing ideas in order to guarantee academic rigour and underwrite universities' institutional independence. For the same reasons, they argue, universities can also legitimately no-platform outside speakers who fail to respect academic norms.
Other writers challenge this perspective by distinguishing between universities' core research and teaching activities, where content restrictions are permissible, and broader public debate on campus, where contentious free speech should be guaranteed and therefore no-platforming disallowed. Chemerinsky and Gillman (2017) take a more absolutist position of this kind, while Ben-Porath (2017) calls for free and open exchange tempered by some pragmatic reforms. However, Simpson rejects this distinction, arguing that 'the communicative climate of the campus at large [should be] characterised by similar kinds of rigour, thoughtfulness, and deference to academic expertise as the lecture theatre or faculty research seminar (2020, p. 299). Likewise, Levy (2019) argues that, given the prestige of speaking platforms, potential speakers who would undermine universities' epistemic authoritycentral to their mission -may legitimately be excluded (e.g., Holocaust deniers or climate change sceptics).

| Fake news
The widespread use of personalised social media to deliver and access information on current affairs has amplified the emergence of fake news, disrupting some of the classic philosophical justifications for free speech. While there is an ongoing debate about precisely what constitutes fake news (Aikin & Talisse, 2018;Gelfert, 2018;Rini, 2017), two key elements are that its producers normally intend deliberately to mislead their audiences (but cf. Michaelson et al., 2019) and that it displays the simulacra of journalistic accuracy while de facto violating its standards. Besides this, there is disagreement as to whether fake news is driven by political (as opposed to e.g., economic) goals (Rini, 2017, p. E45) and whether its effectiveness depends more on flawed institutional structures or on individual epistemic and cognitive weaknesses (Chambers, 2021;Rini, 2017).
Fake news challenges the Millian view that robust free speech is the route to better justified opinions. Indeed, fake news stories' social circulation may contribute to their widespread endorsement rather than demise (cf. Schauer, 2012, p. 137). This may be exacerbated by people's cognitive biases (Gelfert, 2018, pp. 111-2;Chambers, 2021;cf. Moles, 2007); by their lack of the epistemic virtue of 'accuracy', which involves 'making a goodwill effort to find out what is true or to acquire true belief' (Chambers, 2021, p. 159); and by a social media infrastructure that facilitates fake news's rapid diffusion, with uncertain norms over whether sharing messages implies endorsing them (Rini, 2017). Similarly, by exploiting cognitive biases, fake news seems to disrespect individual autonomy. It is also likely to thwart rather than promote rational debate and to undermine trust in the democratic process and democracy's 'truth-tracking potential' (Chambers, 2021, p. 153).
Vigorous counterspeech against its misinformation is one response to fake news. Yet, once fake news has shifted the discursive norms on a topic and entrenched itself in public opinion, pre-emptive measures may be necessary. These may include seeking to empower citizens to discern fake stories more readily (Lepoutre, 2019) and holding those responsible for untrue stories accountable (Rini, 2017, p. E55), for example by flagging fake news sites with a low 'reputation score' on social media, thereby reducing their credibility.

| Online public shaming
Online public shaming occurs when individuals are publicly censured on social media for allegedly reprehensible actions, often resulting in visible humiliation. Like pornography and hate speech, it is speech as conduct. While the 8 of 13 -BONOTTI AND SEGLOW mobilisation of public shame against egregious wrongdoers, such as racists, may seem quite acceptable, the issue is not always so clear, and evaluating its moral permissibility involves examining various normative considerations. Parr (2020a, 2020b) outline some criteria aimed at guiding this assessment (2020b, pp. 1001-6).
These include proportionality, which requires taking into account the negative consequences of online public shaming, only employing it to enforce 'morally authoritative social norms' (Billingham & Parr, 2020a, p. 373) and condemn serious harm; considering the norm violator's genuine liability for her behaviour (2020b, pp. 1001-3); only using shaming if it is no more harmful than alternative options; respecting the privacy of norm violators; not abusing them or attacking them based on their ascriptive characteristics; and aiming for their reintegration back into society. On their view, online shaming is a means to hold shamers accountable for their actions. Jacquet (2015) has likewise pointed to the productive potential of shame, as a means of militating for institutional reform directed at those with power and status, such as high-profile tax avoiders or individuals holding prestigious positions with sexist views. However, there can be backlash effects against shamers who, like whistleblowers, can make themselves more vulnerable (Adkins, 2019).
By contrast, according to Aitchison and Meckled-Garcia (2021), online public shaming is an informal mode of punishment that targets the reputation of norm violators and excludes them from the community of norm-abiding citizens. As a form of punishment, public shaming is problematic since it is not accompanied by the right to due process that is central to liberal democracy, including the right of reply. Moreover, as Nussbaum (2015) argues, shaming (like humiliating punishments) also fails to respect the equal dignity of its targets by signalling that they are second-class citizens.

| CONCLUSION
Free speech absolutists tend to cleave to the value of individual liberty and focus on the content of speech. Those more willing to regulate it tend to see speech as a form of conduct and, at least in debates around hate speech and pornography, are concerned with threats to individuals' status, a more egalitarian view. While regulating free speech does normally entail a restriction of individual freedom, liberals should also be concerned with citizens' equal status, and the value of a life free from domination, oppression and subordination. Worries about campus hate speech, online shaming and the distorting effects of fake news can also be seen as motivated by a concern to protect individuals' equal ability to participate in public debate (cf. Bejan, 2017). Thus the balance between the libertarian and egalitarian dimensions of liberalism may be central to future developments in the free speech literature. And that debate may partly hinge on how we understand 'speech', and on whether and why we think speech is special (see Brison, 1998b;Kendrick, 2017Kendrick, , 2018Schauer, 1983Schauer, , 2015Simpson, 2016). Perhaps some understandings of what speech is, and why it is special, may justify assigning greater importance to its libertarian than to its egalitarian dimension, and vice versa. But that is a question that we leave for another day. 1 We use free speech interchangeably with freedom of expression in this article.

ENDNOTES
2 It should be noted that Schauer sees truth as only one component in the calculus of overall utility and criticises Mill's view. 3 Marmor (2018) claims that the right to speak and the right to hear are two quite separate rights.