Introduction

One argument against restrictions on hateful speech is that they undermine the democratic legitimacy of the state. Restrictions on speech acts, so this argument goes, to undermine the capacity of citizens to articulate their views in the political arena and therefore prevents them from participating fully in the process of co-authoring laws. Because co-authorship of laws is a necessary condition of the legitimacy of those laws, when this is impeded by restrictions on speech acts, so is the legitimacy of any laws produced. In this paper I critically assess a version of this argument outlined in brief by Dworkin (2009), and more expansively by Weinstein (2017a), which I refer to as the Legitimacy Argument.

An opposing position argues that hate speech itself harms the democratic process, so might justifiably be limited in defence of democracy. According to this position, hate speech serves to undermine the political status of some citizens. Where equal political status is a necessary condition of legitimate policy-making, hate speech can therefore undermine the legitimacy of this process. This is motivated by a belief that the Legitimacy Argument understates both that democratic legitimacy presupposes a level of relational equality between citizens in the political sphere and that hateful acts of expression can, in some circumstances, erode such relationships. This article sides broadly with the second position, and argues that there are reasons to restrict hateful speech on democratic grounds in some instances.

Both proponents and critics of the Legitimacy Argument draw on the idea of effective political voice, or the capacity to shape and influence laws. In this article I argue that both sides are somewhat correct in the way they invoke this, and that both (a subset of) hate speech, and regulations of such speech can have a harmful effect on legitimacy. In cases where hateful speech undermines its targets’ political voice, there are therefore strong pro tanto reasons both to restrict and not to restrict it. Tension between these pro tanto reasons cannot be resolved with reference to a simple principle, or formulation of the kind ‘only speech acts with content of type X in context Y must be restricted by the state’. Instead, they must be weighted against each other on a case-by-case basis. This cuts against the Legitimacy Argument, which argues that the harms of restriction on democratic legitimacy are typically, and in most existing liberal democracies, greater than the harms to democratic legitimacy caused by hateful speech.

To make this argument, I frame the debate in terms of ideal and nonideal theory. I argue that under ideal conditions this tension between strong pro tanto reasons would not arise. Real, nonideal cases are therefore best understood as moral blind alleys, where all courses of action or inaction are harmful, in terms of their effect on democratic legitimacy, as they entail deviation from a democratic ideal. The debate ought therefore to be framed around the possibility of realising the best possible approximation to legitimate policy-making. The failure of the Legitimacy Argument is that it implicitly assumes certain ideal conditions that are absent in the real world. It is possible to envisage a situation where hateful speech could never undermine democratic legitimacy, and if this were the case the Legitimacy Argument would hold. However, such a situation would require a greater degree of political equality than currently exists.

Framing the debate using the language of ideal and nonideal theory has two advantages. First, those who have defended hate speech regulations by invoking political voice have been accused of minimising the effects of hate speech regulations on democratic participation. I believe there is something to this critique, even though I ultimately believe there are cases where restriction is justified. By positioning the choice that states make as a ‘tragic’ one between two flawed options I am better able to acknowledge the harmful effects of regulation on democratic participation than other critics of the Legitimacy Argument. It also allows me to make the argument in favour of restriction on hate speech without denying that an expansive right to freedom of expression is constitutive of democratic legitimacy. As such, it enables me to take on the Legitimacy Argument on its own terms. Second, it helps separate the ‘moving parts’ of the various positions in the debate around the Legitimacy Argument: the requirements of democratic legitimacy; the role of freedom of expression in this process; and the specific impact of hateful speech on it. It shows that both sides have a similar view of democratic legitimacy in mind, and highlights that the disagreement between the two positions is more pronounced on the question of how to apply a democratic ideal in practice, and on the empirical effects of hate speech.

A Note on Definitions and the Scope of the Argument

This article asks the question of how restrictions on hateful speech impacts on legitimacy. Therefore, it does not consider other, unrelated arguments in favour of an unrestricted right to freedom of expression.

I focus on legitimacy in a normative sense. The question I aim to answer is whether states are morally justified in imposing laws, and whether citizens are morally obliged to obey them. I also assume that legitimacy is a matter of degree—that is, that a state might plausibly be described as more or less legitimate, or partially legitimate. There is a separate question around ‘descriptive legitimacy’, which focuses on whether the state enjoys the support of citizens and appears legitimate to them.Footnote 1 Weinstein, for example, seeks to buttress the Legitimacy Argument by pointing out that it is, among ‘normatively appealing’ reasons to defend a right to freedom of expression, more consistent with existing doctrine in (American) law, and subject to a more widespread consensus than the other contenders (Weinstein 2011b, pp. 668–669). I do not consider these arguments here. Instead, I focus on the question of legitimacy as a normative property in the abstract.

Two other sets of terms that I use through the article are ‘hate speech’; and a ‘wide-ranging’ or ‘expansive’ right to freedom of expression. I use the second two terms interchangeably to refer to a right to freedom of expression that includes protection for hateful speech, and is not limited by hate speech laws of the kind that exist in any real, liberal democracies. I do not offer a specific definition of ‘hateful speech’ either, though I use the term throughout. This is because I suggest that there are some speech acts that might justifiably be regulated on democratic grounds because of the effects that they have. Speech acts that might have this effect are really a subset of what is usually defined as hateful speech within certain contexts. Therefore, the speech acts that I believe might be regulated on democratic grounds will, I believe as a matter of fact, be hateful, in that they will be vituperative or threatening, and target a particular group. However, it should not be regulated because of one of these characteristics so I do not believe that all hateful speech may be regulated; only a subset of it that might have certain effects. The reason that I use the language of hate speech is because the Legitimacy Argument is often used to challenge existing hate speech legislation, I believe incorrectly.

The Legitimacy Argument

The Legitimacy Argument is that any restriction on freedom of expression undermines the legitimacy of the state, because to prevent someone from speaking, even hatefully, is to prevent them from participating in the democratic process as they are entitled to. Dworkin presents the starkest version of this view, stating that ‘[i]t is unfair to impose a collective decision on someone who has not been allowed to contribute to the moral environment’ (Dworkin 2009, p. viii). He characterises freedom of expression as part of a political deliberation that occurs ‘upstream’ of the imposition of laws. The argument he makes is that in order for the laws ‘downstream’ at the level of imposition to be legitimate, restrictions cannot be in place ‘upstream’ at the level of political deliberation. So, for example, we cannot impose anti-discrimination laws on bigots ‘downstream’ if they have not been permitted to articulate their bigoted views ‘upstream’ (Dworkin 2009, p. vii). This argument appeals to a broadly liberal worldview whereby the legitimacy of laws is achieved through a process of public justification. When the state places limits on political discourse, it prevents citizens from articulating potential objections, or seeking to challenge proposed laws and provide alternatives. Because of this, any restriction of freedom of speech ‘implies at least a partial suspension of democracy’ (Rawls 2005, p. 354). From this perspective, the legitimacy of laws depends on them being justified to all citizens through a procedure that allows them to raise objections and potential counter-arguments.

Dworkin’s argument is brief, and appears overwrought; it seems implausible to argue that the legitimacy of laws like anti-discrimination laws is completely destroyed if there are limited restrictions placed on hateful speech at some point in the political process (Waldron 2012a, p. 186, b). Weinstein offers a refined version of the Legitimacy Argument in which he concedes that if it were possible to restrict solely the vituperative content of hateful speech acts, such restrictions would not ‘destroy’ or even ‘significantly impair’ normative legitimacy (2017a, pp. 548, 549, n. 90). His worry is that restrictions on hateful speech will also prevent people from articulating the propositional content of their argument, which in his mind would undermine the legitimacy of laws produced (Weinstein 2017a, pp. 564–565). This is because in Weinstein’s view hate speech laws have frequently been applied to punish speakers for non-vituperative expression of bigoted ideas (Weinstein 2017a, pp. 561–562).Footnote 2 Citizens ought to have avenues open to them to advance legitimate interests and raise legitimate concerns to others about laws, and it is this that hate speech laws, or indeed any laws restricting a wide-ranging right to freedom of expression, might inhibit. The opportunity to participate effectively in politics like this is a condition of legitimacy, so when it is restricted the legitimacy of the laws produced is diminished too (Weinstein 2017a, p. 537).

Challenging the Legitimacy Argument

There are two strategies that can be pursued to counter the Legitimacy Argument in favour of hate speech regulation. We might accept that hate speech restrictions harm legitimacy, but defend the legitimacy of particular laws based on moral reasons independent of the process by which the laws were produced. We could also ‘bite the bullet’ and try to argue that a right to freedom of expression in general is not actually required by the accounts of legitimacy that underpin the Legitimacy Argument.Footnote 3 The alternative, which I pursue in this article, is to argue that hate speech might undermine the legitimacy of laws for similar reasons that restrictions on speech do, that is by inhibiting political participation. This is a result of the harmful effects of hate speech that ‘dignitarian’ accounts of hate speech like Waldron’s (2012a) point to, such as the undermining of civic dignity and standing. Pace the first possible response, the indignity suffered by targets of hate speech is not something that needs to be weighted against legitimacy, but is ‘constitutive’ of it (Brown 2015, p. 208). This can be understood as an internal critique of the Legitimacy Argument, given that it accepts both the broad account of legitimacy that Dworkin and Weinstein do, and also that restrictions on hate speech can impinge upon legitimacy.

According to this critique, an ‘aggressive’ free speech regime that protects hate speech would deprive the targets of an adequate political voice (Brown 2017, p. 616). For citizens who are marginalised in this way, political institutions fail the standards of public justification that underpin the Legitimacy Argument because there is no reason that they ought to accept this state of affairs (Brown 2017, pp. 616–617). If this is the case then, to use the language employed by Dworkin and Weinstein, hate speech may, in some contexts undermine the ‘upstream’ mechanisms by which legitimate laws are produced, meaning that targets of hate speech, and the citizenry more broadly, do not have as much reason to treat downstream laws as legitimate (Waldron 2012a, pp. 70–71).

When Dworkin sets out the Legitimacy Argument, he conflates a thicker, more substantive account of political equality—equal political ‘voice’ or influence that depends on equal status—with a thinner protection of core rights or negative liberties. He argues that we have a right both to ‘raise a protest’ and ‘shape the moral environment’ (Dworkin 2009, pp. vii, viii). These two are quite different, as the latter implies that we are entitled to an adequate opportunity to participate in the shaping of laws, whilst the former implies merely the chance to express certain views and ‘sound off’ unimpeded. Similarly, Weinstein invokes a standard of normative legitimacy that treats ‘equal opportunity to participate in the political process’ as a necessary condition of normative legitimacy (Weinstein 2017a, pp. 536–539). He defines the equal opportunity to participate in a thin, formal sense—to ‘raise a protest’ using Dworkin’s terminology (Weinstein 2011a, p. 506). However, he acknowledges that some only see formal rights as necessary to legitimacy because they are constitutive of political voice conceived in the thicker, substantive sense—the right to ‘shape the moral environment’ in Dworkin’s terms (Weinstein 2017a, p. 537). I believe that the most normatively attractive version of the Legitimacy Argument incorporates this thicker, substantive view of equality in the political sphere and the process of co-authoring laws. I take this view because, for me, the attraction of the Legitimacy Argument lies in the centrality of equal respect as a value, and its broad commitment to public justification, which I believe requires equality of this kind between citizens.

There is inevitably a wide range of views as to what substantive political equality entails. To narrow this down a bit, we can draw upon Dworkin’s own view that whilst an effective capacity to shape the moral environment does not entail equality of influence, it does require equality of ‘political standing’ (Dworkin 2011, pp. 388–392).Footnote 4 This requires a distribution of political power that is compatible with ‘the political community’s equal concern and respect for all of its members’ (Dworkin 2011, p. 391).

It is this point that Brown is targeting in his critique of the Legitimacy Argument outlined above: he argues that hate speech can inculcate a lack of respect that is severe enough that citizens ought not, reasonably, be expected to accept it going unchallenged.

It is one thing to point out that the Legitimacy Argument is dependent on a substantive account of political voice, but for challenges like Brown’s to succeed they need to show how hate speech undermines this. Proponents of the Legitimacy Argument tend towards a view of democracy as a process of public justification and deliberation amongst equals; they outline a view of democracy that is in parts deliberative and participatory (Brown 2015, p. 190). This conception of democratic legitimacy depends on citizens having not just formal rights, but an effective political voice, which involves, at the very least, the capacity to influence others (Jarymowicz 2014, pp. 115–116). This political voice depends on citizens treating each other as equals, so means that it is impossible to divorce the idea of civic dignity from effective, or substantive, membership of the deliberative community (Brown 2017, p. 609). Having an adequate voice means more than just being able to say the same things as others; it means that the other participants in that deliberation treat you with a degree of respect, and are prepared to at least countenance being influenced by your view (Heyman 2009, p. 172). This entails a mutual respect for the ‘deliberative capacities’ of other citizens (Cohen 1997a, pp. 92–93). Deploying the Legitimacy Argument to defend freedom of expression as a negative liberty without any reference to the process by which speech acts are rendered more or less effective politically underestimates the extent to which this substantive political equality is inevitably ‘relational’—it depends on how citizens treat each other, and in turn how the state regulates this (Jarymowicz 2014, p. 114; Brown 2015, p. 203).

There are two significant ways that hate speech might undermine these relationships. On an individual level, the targets of hate speech might be less inclined to participate in politics because of a sense that they are not treated with dignity, or in extreme cases because they feel threatened (Brown 2017, pp. 609–610). In this article, I focus, instead, on the claim that hate speech might have a deeper pathological effect on political discourse, because it might plausibly cause people to be taken less seriously in politics when they do decide to participate. The account of legitimisation that I have drawn upon in this article has been one that depends on participants in politics adhering to certain basic deliberative norms, specifically treating others with respect as deliberators. This mutual respect might be undermined if certain groups are routinely treated as if they lack core deliberative capacities, which is a potential effect of hate speech.

The obligation to take others seriously and respect them in deliberation is difficult to pin down. In practice, most people do not comply with this in the way that participatory or deliberative democrats suggest they should at all times. Even the most liberal-minded citizens inevitably seek to influence others in order to secure certain policy outcomes without ever really being prepared to consider their counterpoints on their own merit and be persuaded. It therefore makes sense to think of this in terms of ideal theory, where we define this in the traditional Rawlsian sense of full compliance with the requirements of justice and favourable background conditions (Simmons 2010), and where there is widespread noncompliance that means that the ideal cannot be met. It is also a requirement that is essentially unenforceable—it would be both impossible and illiberal to ‘force’ citizens to respect others, or seek to change deeply bigoted beliefs coercively. Despite this, there are some areas in which they can be promoted by the state. For example, it is possible to enforce norms of civility in some political contexts that might indirectly contribute to a more respectful environment. It is these second-order norms that foster respect that hate speech might come to undermine (Jarymowicz 2014, pp. 117–121).

What sets some hate speech apart as a particularly acute violation of this norm that might warrant coercive restriction by the state is the way that it can affect these second-order norms, and in turn shape social attitudes towards others or cause the target to adjust their behaviour. The particular harmful effect of hate speech is therefore to cause others—not the speaker or the target—to ‘transgress the basic ground rules of deliberation’ in their treatment of the target (Heyman 2009, p. 176). This creates a second-order silencing effect where certain voices are systematically disregarded in a way that violates a view of legitimacy that depends on effective political participation. Hate speech has a ‘disfiguring’ effect on the wider ‘social environment’, and this can rub off on political deliberation (Waldron 2012a, p. 117). If hate speech has this effect it means that it is possible that it harms legitimacy even where it does not have a ‘silencing effect’ on individuals, by causing them to not participate in aspects of politics.Footnote 5 Instead, it suggests that hate speech might mean that citizens are able to participate less effectively as a result of changes in others’ behaviour and political norms.

Where such norms are inculcated, a liberal standard of democratic legitimacy cannot be met, as some citizens who are entitled to participate in the process of authoring and justifying laws lack the de facto capacity to do so, even if there are not formal legal obstacles to this being the case. It is actions that have this effect that might justifiably be subject to restriction from a liberal perspective.

Because my worry is about the second-order effects of hate speech, this argument does not entail purely viewpoint-selective restrictions on hate speech, and the kind of hate speech that might be restricted is therefore not defined entirely according to its content.Footnote 6 Instead, such restrictions would apply to a subset of hateful speech acts, that might cause the silencing effect or restrict others’ political voice.Footnote 7 Though I am unable to provide a full account of these within the scope of this article, factors will include the context and medium of the remarks made, the authority of the speaker, and in particular the nature of any existing injustices and discrimination suffered by the group being targeted.

The ‘Nuanced Principle’ as Necessary Condition

Given the above there are occasions where hate speech restrictions may be justified in order to preserve democratic legitimacy. This is something like Brown’s ‘Nuanced Principle of Democracy’ that ‘legalistic constraints on uses of hate speech are (N-)warranted if they operate for the sake of ensuring that all citizens enjoy real opportunities to participate in public discourse’ (Brown 2015, p. 195). The nature of this claim hinges on what ‘(N-)warranted’ means in this context; for Brown it is a ‘narrow warrant’ that a principle holds until it is ‘overridden or trumped’ by a countervailing argument that also draws upon a normatively salient feature of the case under discussion (3). This means that the state might restrict hate speech on democratic grounds so long as there are not countervailing reasons to protect a more expansive right to freedom of expression on behalf of the speaker of hate speech.

I agree with Brown that if hate speech does undermine democratic voice then this does generate a pro tanto reason to restrict it; indeed, I would strengthen this point and state that hate speech undermining effective political participation is a necessary condition for any case for restriction on democratic grounds. However, there remain strong pro tanto reasons not to interfere in such instances (that I believe Brown underplays), because any restrictions on expression will, by their nature, harm the political process by limiting effective participation, and because of a latent risk of overreach by the state in seeking to enforce these norms (Weinstein 2017b, pp. 770–771).Footnote 8 The presence of the reasons to restrict hateful speech that I have outlined in the previous section do not eliminate the silencing effects of regulations on such speech, that Weinstein pointed to in his defence of the Legitimacy Argument. There are therefore situations where we have good reason to reject both restrictions on hate speech and the absence of restrictions: speakers have legitimate reasons to be wary that their legitimate interests will be disregarded, whilst targets will not be able to fully realise their political voice as a result of marginalisation. The legitimacy of state interference to limit hate speech must therefore be judged on a case-by-case basis that weights these two sets of pro tanto reasons against each other. This contrasts with a view that if a speech act serves to undermine the speech of others then it is required that it be restricted, and the view that even if hate speech harms legitimacy, any restrictions undermine it to a greater extent.Footnote 9

In the final section I orientate this trade-off view by drawing on some of the literature around ideal and nonideal theory. First, it is necessary to lay out some reasons that we ought not to adopt a view that in a set of cases we might assume that we ought always to rule in favour of restriction. In doing so, I want to guard against understating the ways in which restrictions of speech might have a ‘silencing effect’ on the speaker, and the extent to which the options of hate speakers in real-world conditions to advance their views in a more reasonable fashion are sometimes limited. Part of the issue here is that proponents of the Legitimacy Argument are prone to exaggerate the silencing effects of some hate speech laws, eliding any restrictions on speech with complete exclusion from the public sphere (Waldron 2017, p. 700).Footnote 10 However, there is a danger of going too far the other way.

In particular, there are effects to restrictions on expression that extend beyond limits to particular attempts to articulate a political position. Those who are silenced, even when there is good reason, suffer a degree of social and political stigma. Hate speech restrictions inevitably end up doing more than merely preventing someone from using hateful or vituperative speech to make a political point in a given context. This is particularly true if you consider hate speech as I have presented, in the context of an existing political system that does not live up to a democratic ideal. In such a context, disruptive acts such as hateful speech might draw attention to constituencies or groups that do not, currently, have an effective political voice. The various cases of hateful speech by populist (and predominantly right-wing) politicians may be justifiably restricted, but we should not ignore the fact that part of the reason that such groups gain traction in the first place is because part of their constituency feel, rightly, that they have been marginalised in the political process. Of course the targets of hate speech themselves often lack political power, so this argument does not mean that we should argue against restriction in all cases. Nonetheless, it supports the view that the competing reasons on either side of the debate really are closely matched.

More generally, just as the Legitimacy Argument presupposes an idealised view of actual political participation, many of its critics are guilty of having too idealised a view about the other ways that citizens are able to articulate their views when speech is restricted. In this, I side with Weinstein in worrying that bans on hateful speech cannot be targeted at vituperative speech exclusively. Brown, for example, argues that we ought not see the question as a set of trade-offs where the benefits of hate speech bans to legitimacy are framed as ‘potentially offsetting’ the harmful effects of restrictions of speech (Brown 2017, p. 616). For Brown, what matters is that we ensure all have ‘sufficient real opportunity to contribute to political discourse and participate in the formation of public opinion’ (Brown 2017, p. 616). His rejection of the kind of trade-off view I have outlined is, like the Legitimacy Argument, underpinned by a belief that all citizens can have sufficient opportunity to participate politically.Footnote 11 I do not believe that this is always the case; sometimes both an act of hate speech and the act of restricting it will undermine political participation in a way that seriously impinges on democratic legitimacy. Under such circumstances, and in contrast with Brown’s implied threshold view, the best we can hope for is the closest approximation to a state of affairs where all can participate fairly. This kind of judgement requires the weighting of reasons on both sides, and if we take the Legitimacy Argument seriously then the damage done to the democratic process by restricting hate speech will trump the Nuanced Principle in favour of restriction fairly frequently.

Given this indeterminacy, rather than trying to apply a general principle to the relationship between hate speech and democracy, like the Legitimacy Argument and the Nuanced Principle do on different sides of the debate, we should acknowledge the conflicting reasons that arise in any given case and draw on the arguments underpinning these principles to better elucidate them.

Hate Speech Generates a Moral Blind Alley

I have argued that the most attractive version of the Legitimacy Argument, and the most pressing critiques of it, connect legitimacy to effective and not just formal political voice. For this to work, it requires others to comply with certain rules and norms, even where they are not enforceable through coercion or threat of punishment. This places it within the realms of orthodox ideal theory.

This understanding of the application of theories of legitimacy in this context allows the following argument: that under ideal conditions, any restriction on hate speech would harm legitimacy, but under nonideal conditions they might enhance it; this is the position I adopt. This kind of argument has been deployed in favour of the restriction of hate speech in real-world conditions. Brown, for example, notes that the case for legislation against the incitement to hatred is more robust in societies where there are not strong social norms supporting equality amongst citizens (Brown 2008, p. 18), whilst Bonotti argues from a republican perspective that hate speech laws might be necessary to ensure non-domination in a public sphere which is already ‘ethnicised’ (Bonotti 2017, pp. 268–270).Footnote 12 Both of these arguments resemble the case I have made against the Legitimacy Argument, in that they point to the attractiveness of unrestricted freedom of expression, but only under conditions where certain norms are observed that do not currently persist. In this section, I develop this line of criticism using some of the literature around ideal and nonideal theory, specifically the account of moral blind alleys proposed by Schapiro and the use of the Theory of Second Best in economics as an analogy.

The core claim I make is that under ideal conditions restrictions on hate speech harm legitimacy, but hate speech does not. This is because restrictions on hate speech automatically harm legitimacy whilst hate speech alone does not. Limits on speech, by their nature, cut against an inclusive ideal of deliberation by temporarily excluding some perspectives or individuals from the political process. If you accept, as I do, that the hateful aspects of speech cannot be completely isolated from the context of this speech, and that therefore there is no way of effectively silencing just the hateful aspects of speech, there is no way that restrictions on hateful expression will not affect the political participation of the speaker, however minimally. My conjecture is that it is at least possible to imagine a context where hateful comments alone did not have these effects.

Unlike the effects of censorship, which are direct, the harmful effects of hate speech on democracy only occur in a given context where there are existing inequalities and injustices. Hate speech does not undermine deliberative norms by its nature or as an intrinsic by-product of its content because such norms are independent of the content of the views expressed in discussion.Footnote 13 Instead, hate speech can contribute to a moral environment where some citizens lose an effective political voice as a result of the way that democratic norms are reshaped. If the other conditions assumed in the ideal theory of legitimacy persisted, the state would be able to mitigate or compensate for these effects. This is not the case under nonideal conditions where social stigmatisation affects standing in politics in a more profound way. Brown is therefore correct that the targets of hate speech might reject a speech regime that protected speech that diminished their civic status through appeal to an interpersonal standard of justification—that is, that they might reasonably reject the authority of such a regime (Brown 2017, pp. 616–617)—though with the caveat that this argument only applies under nonideal conditions where background societal norms mean that the targets of hateful speech will see their civic status diminished.

Where I differ from Brown is in the weight I attach to ideal theory in providing action-guiding reasons. Brown, like me, criticises the Legitimacy Argument (albeit briefly) for assuming ideal conditions that do not pertain (Brown 2015, pp. 188–189).Footnote 14 However, in this passage, he appears to argue that the reason the Legitimacy Argument fails is because it offers an ideal of democracy but not a viable route to transitioning to this more favourable situation. With this in mind he worries that adhering to a free speech regime now, when these conditions do not pertain and will not come about, means ‘sacrificing’ the targets of hate speech at the moment in pursuit of an unattainable democratic ideal (Brown 2015, p. 189). I believe that ideal theory provides strong pro tanto moral reasons that retain normative force under nonideal conditions independent of the likely realisation of more just background conditions. In the rest of the paper I offer a more nuanced counter to the Legitimacy Argument that incorporates these strong pro tanto reasons, but still resists the idea that states should behave as if ideal conditions pertain when they do not.

Taking this approach to ideal theory means that in cases where an act of hate speech will shape background norms, and as a result adversely affect the way that some citizens are treated in deliberation there are not only competing pro tanto reasons, but the state necessarily brings about an outcome that does not conform to a democratic ideal of co-authorship, whatever it does, or does not do. This is because two principles that are central to citizens having effective political voice conflict: non-interference by the state and maximal inclusion of perspectives on one hand, and equal standing amongst citizens on the other.Footnote 15

Whatever course of action it chooses will incur some unjust cost, in this case by undermining the democratic process. The choice to be made is therefore the option that least undermines a democratic process, i.e. that undermines the political voice of citizens the least. These sort of choices—effectively of the least bad option—are not uncommon under nonideal conditions, and are characterised by Nagel as moral ‘blind alleys’ (Nagel 1979, p. 74). For him, a blind alley is a situation where we must choose either to violate some non-consequentialist constraint, or to cause a worse outcome. Importantly, we have a valid reason to regret either course of action that is not offset by the all-things-considered benefit of taking the alternative option. Schapiro argues that there are blind alleys where either course of action open to a person will cause them to violate a non-consequentialist norm (Schapiro 2003, pp. 332–334). This kind of blind alley can arise where non-conformity by other ethical actors causes a situation where the non-consequentialist duties we have come into conflict. In such cases, two non-consequentialist duties end up conflicting, and one does not take clear priority over another. A subset of these cases is those where there is a single overarching demand of people, say to treat others with equal respect, which produces correlative duties.

This can occur when states seek to enforce a democratic system based on legitimate policy-making that is founded on an underlying value of respect, but requires a range of rights and institutions to work effectively. For state actors blind alleys can and do arise where any possible option entails undermining the political voice of some citizens. Any course of action available to them will incur significant cost, as it will involve violating core political rights, or fundamental democratic principles.

This is the kind of choice faced by states when they consider whether to sanction those who perform acts of hate speech. They must violate the norms associated with preserving deliberation whatever path they choose, because any course of action will undermine the political voice of some. If they restrict a speaker of hate speech they undermine deliberation by preventing that individual from participating in the political process, and being able to scrutinise policies or advance their interests. However, if they do not restrict acts of hate speech some other citizens will be denied their political voice, either through intimidation or marginalisation, or as a by-product of damage to deliberative institutions. Kirshner offers a good general summary of the limitations we need to consider when approaching the question of how to respond to unreasonable and anti-democratic citizens, which is appropriate when we consider whether to regulate ‘hate speech’:

… I contend that the dilemma raised but those who oppose democracy will be more tractable if we treat defensive policies as efforts to augment the democratic character of flawed regimes, instead of as attempts to preserve a moral community or any other idealized status quo (Kirshner 2014, p. 17).

Confronted by this blind alley, it is therefore impossible to posit a sufficient reason for a partially legitimate state to impinge upon speech acts that are not direct invocations of physical harm. There is no way of presenting a principle of the form: ‘speech ought never to be restricted, unless it includes content X in circumstances Y’. Instead, we can posit necessary conditions and pro tanto reasons for interference. In the case of hate speech, where a speech act impinges upon how other citizens participate politically, there is a pro tanto reason to restrict the freedom of expression of the speaker. However, it is not an all-things-considered reason because there is also a cost to silencing citizens in the process of deliberation. Hate speech having the effect of reducing effective political voice is a necessary condition for justified restriction on democratic grounds. However, it is only a necessary condition—each case must be considered individually, to establish the extent to which both restriction and non-restriction cause deviation from the democratic ideal. What is more, there may be cases where this is hard to discern. Effective political voice depends on a range of factors, so assessing the way that restriction and non-restriction of hate speech impacts this is difficult to establish. Different ways that political voice may be undermined are not necessarily commensurable. In this way, the role of an ideal democratic theory is to help characterise the nature of the harms of both courses of (in-)action.Footnote 16

This framing of the argument has two advantages. First, it renders explicit aspects of the democratic case for hate speech regulation in the existing literature. For example, after surveying the effects of hate speech laws, Tsesis concludes that whilst freedom of expression is ‘essential to collective decision making’ (my emphasis), hate speech might cause individuals to justifiably feel threatened, or inculcate discriminatory behaviour that also undermines political autonomy (Tsesis 2009, p. 532). Reconciling these dual claims is central to a democratic argument for hate speech regulation and my framing offers a way of doing this. Second, it diffuses some of the potency of the Legitimacy Argument without denying the central premise that a right to freedom of expression is essential to democratic legitimacy. In responding to Waldron’s critique of the Legitimacy Argument, Dworkin makes the point that we should not defend the censorship of hate speech on the grounds that ‘[w]e can hardly justify a defect in political legitimacy by arguing that it might have been worse’ (Dworkin 2012, p. 341); if you accept my framing of the choice as a moral blind alley, this kind of argument actually becomes quite attractive.

Why Non-interference is Not Always the Best Option

Using the framework of ideal and nonideal theory in this case allows for the argument that there are strong reasons against non-interference, but that these are not all-things-considered reasons for the state not to act. One of the reasons for this is that when confronted with nonideal circumstances, behaving as we would under ideal conditions may actually exacerbate some of the harms that occur. When confronted with a blind alley, the wrong thing to do would be to imagine a world in which the ideal option would be open to us, and to then act, as far as is possible, as if this were the case.

In the case of hate speech, we might have reason to restrict freedom of expression because permitting it causes greater harms to others, or damages the process of the justification of laws more, than restricting them would. This is because, as noted above, the question is how best to achieve the closest approximation to a democratic ideal, rather than the ideal itself, which is unattainable due to structural constraints and widespread noncompliance. When some actors are using acts of expression to undermine the political rights of others, leaving them be just because this best conforms to the ideal will not necessarily bring about the best achievable realisation of political rights.

This line of argument echoes the Theory of Second Best (ToSB) in economics.Footnote 17 ToSB posits that if one is presented with an optimal situation, then introduces some constraint, the second best outcome is not necessarily achieved by leaving all other things the same. In economic terms, the optimal outcome here is pareto optimal, and the types of constraints are those that disrupt this economic equilibrium (Lipsey and Lancaster 1956). When a constraint is introduced that prevents this optimum being met, there is no reason to assume that there will be a favourable outcome achieved by maintaining the other constraints that led to that equilibrium being reached (Lipsey and Lancaster 1956, p. 11). Furthermore, there is no a priori way of judging between possible second best outcomes with reference to the initial, stipulated optimum conditions (Lipsey and Lancaster 1956, p. 12). Indeed, the breaking of paretian rules in some cases might increase overall welfare or utility (Lipsey and Lancaster 1956, p. 31). There is also unlikely to be a ‘simple sufficient condition’ to increase welfare where the optimum outcome cannot be achieved (Lipsey and Lancaster 1956, p. 31).

If this analogy is relevant when thinking of the relationship between freedom of expression and legitimacy, it would therefore support my positions that there is no simple solution to determining how best to configure a right to freedom of expression to secure political voice; and that we ought not automatically assume that it is a maximally expansive right, which protects hate speech, as it would be under ideal conditions. There are two claims that need to hold if the analogy of ToSB is going to be useful to discuss restrictions on hate speech on democratic grounds. First, there needs to be a desirable, best outcome. Here, it is a legitimate democratic system. Second, this must rely on various things being true simultaneously; again, I believe that this is the case.

The first of these is easier to demonstrate. Although there is not a measurable optimum goal, like maximum welfare or resources, as there would be in examples that apply ToSB in economics, we can assume that the desired outcome is one where all people have sufficient opportunity to participate in politics such that laws are justified, as outlined in the previous sections. The main difference is that this is a situation where various non-consequentialist rights are effectively secured, rather than one where a consequentialist metric is maximised, but for the purposes of the analogy it is fair to treat this as an optimum.

The most convincing versions of the Legitimacy Argument also pre-supposes a variety of factors that contribute to political voice. To illustrate this, we can split the way that the legitimacy argument is presented into two versions, that I will call ‘Single Law’ legitimacy and ‘Systemic’ legitimacy. By Systemic legitimacy, I take it to mean that legitimate laws depend on the authority of institutions, which in turn depends on a proper process being followed. According to this version of the Legitimacy Argument, hate speech restrictions harm the legitimacy of laws in general, because they distort the overall process of co-authoring laws.

If we accept this view of legitimacy, the ToSB offers a useful and fairly direct parallel. It is fair to treat the democratic process as a process with many ‘moving parts’. There are various political rights, and institutional arrangements, that need to function for citizens to be able to possess the political voice required such that they can participate in a process of deliberation that can produce legitimate policies by liberal standards. ToSB is relevant because it holds for economic arguments where multiple different kinds of policy instruments are used to achieve a particular goal. This mirrors an ideal situation where the fair level of political voice is simultaneously realised by all citizens. The ideal of deliberation requires not just favourable background conditions, but that citizens conform to certain behavioural norms concerning respectful deliberation. When citizens start to disregard these systematically it unbalances the rest of the deliberative process (Wiens 2016, p. 140).Footnote 18

The Single Law version of the legitimacy argument is a little harder to cast in these terms, but it is still possible. According to this perspective, the legitimacy of a given law depends on people being able to effectively influence the passage of that law. So, for example, Weinstein offers a stylised version of a real-world case where an evangelical wedding photographer is deterred from protesting against the legalisation of same-sex marriage as a result of hate speech laws. She is then subject to laws—which she disagrees with and had planned to protest—that prevent her from denying her services to same-sex couples celebrating weddings (Weinstein 2017a, pp. 567–568).Footnote 19 The point of the legitimacy argument made like this is that the legitimacy of specific laws depends on the process by which they were produced. This version of the Legitimacy Argument might seem to be insulated from my concerns. After all, if we can point to specific instances where a hate speaker has been prevented from participating fully in the formation of a particular law, then other concerns might be put to one side.

This argument falls flat for two reasons. First, the main thrust of the criticisms of the Legitimacy Argument I have outlined is that allowing the hate speaker to speak hatefully in some contexts might prevent others from participating in an analogous way (Stone 2017, p. 691). Second, the example still relies on acts of expression translating into political influence. If, for example, hate speakers were allowed to speak, but were defamed such that others disregarded their opinion in political deliberation, then their right to freedom of expression would have been insufficient for their having a sufficiently effective political voice. Even if we approach the process of legitimising laws on an individual level, and do not consider the idea of systematic or institutional legitimacy, the procedure is still complex, and involves multiple sets of behaviours working in tandem to be effective.

To summarise, the optimum outcome is one where all laws produced are legitimate according to the broadly liberal standards described in this article. This requires a particular process of deliberation and justification; a set of political institutions; and the effective realisation of political rights including a wide-ranging right to freedom of expression are constitutive of this. In practice, not all of these rights can be realised simultaneously, so the process necessary for legitimate law-making cannot be realised as a result of widespread non-compliance. The analogy with the Theory of Second Best shows that when we structure the case like this, the best possible approximation to the ideal of deliberation might be achieved by changing some of the variables away from what they would be in the ideal case. In this way, restrictions on hate speech might be the best way to realise the closest possible approximation to legitimate policy-making, even if, by their nature, they are incompatible with the full realisation of this ideal.

Conclusion

This paper has broadly sided with those who have argued that the restriction of hate speech might actually enhance democratic legitimacy. However, it has done so from a perspective that is in many ways sympathetic to the opposing view: the Legitimacy Argument. I have argued that proponents of this argument are correct about the harms that censorship, even of hateful speech, inflict on the democratic process. Their main error is in underestimating the implications of the fact that the account of legitimacy pre-supposed in their arguments depends on citizens having an effective political voice, which is the effective capacity to shape policies. In particular, they understate the ways that that this capacity is relational; and that the kinds of social relations required for this do not fully pertain in the real world. Protecting hate speech under nonideal conditions like this might restrict the political voice of its targets in a way that it would not if there were more favourable conditions and greater compliance with democratic norms. Where this is the case there are strong pro tanto reasons in favour of both restricting hate speech, and permitting it as part of a wide-ranging right to freedom of expression. These must be weighted against each other on a case-by-case basis to determine the effect of restrictions on legitimacy in the specific context.

This is in some ways a non-conclusion. It suggests that neither the Legitimacy Argument nor its critics have landed a ‘knockout blow’, because neither sides’ arguments eliminate the pro tanto moral force of the reasons on the other side. It is also a frustrating conclusion, because it implies that in certain instances hate speech regulation might enhance legitimacy, but at the same time, it rules out a general principle for determining when this is in terms of the content and context of said speech. I have argued that some restrictions on hate speech are justified with reference to legitimacy, without making the case for particular laws that restrict hateful speech; indeed, according to my framework there are inherent difficulties in trying to establish laws around hate speech at the abstract level. In defence of the project, I would note that the article has only discussed legitimacy as a normative standard. Its goal is to offer a framework to determine whether hate speech restrictions harmed the legitimacy of laws. In line with this, whilst the article does not propose a specific formulation for when to restrict hateful speech, it suggests a framework for understanding the relationship between hate speech and legitimacy in different cases.