2 Case analysis and the dominance argument in CH2

Mill scholars are divided about the burden of proof in CH2. If the expression or discussion of opinions was always harmless, as Dale Miller (Reference Miller2021) suggests, it would straightforwardly follow that their suppression could not be justified by the harm principle. However, if certain assertions or discussions were deemed potentially harmful, there would be a pro tanto reason to silence them. In this case, the proponents of doxastic liberty would have had to demonstrate that the reasons against silencing outweighed the pro tanto reason for it.

At the beginning of chapter 3 (Of Individuality, as One of the Elements of Well-Being), the Mills give the example of an article in the press claiming corn dealers are starving the poor. Imagine a few readers become so outraged that they feel justified in physically attacking corn dealers. Without having read the article, they would never have considered undertaking such attacks. Thus, the article is part of what Mill calls in System of Logic the real cause, the set of antecedents of an event ‘but for which it would not have happened’ (CW 7, p. 328). In any extant interpretation of the Mills’ meaning of ‘harm’, being physically attacked by a mob counts as harm.Footnote ⁷ As the article contributes to a cause that ultimately leads to violence, its author is causally co-responsible for the resulting harm, and his writing falls under the jurisdiction of the harm principle. Therefore, the argument of the Mills cannot be based on the premise that ‘words will never hurt’, as Miller argues. Assuming that the publication of the article is permitted and recognising that it contributes causally to the attacks, the publication must constitute a case of permitted harm. It is thus unsurprising that CH2 contains no argument suggesting that speech is always harmless.Footnote ⁸ It also explains how the Mills characterise the right to express opinions, including harmful ones. They write that apart from extraordinary circumstances in which ‘even opinions lose their immunity’ (CW 18, p. 260, emphasis mine), senders must not be held morally responsible for the harm which results from their messages.Footnote ⁹ The term ‘immunity’ refers to an exemption from liability. I contend that the Mills use the term for a reason. Apart from extraordinary circumstances, individuals are not liable for assertions which contribute to the causation of harm. Extraordinary circumstances occur if an assertion amounts to ‘a positive instigation of some mischievous act’ (CW 18, p. 260).Footnote ¹⁰

The expression of opinions can thus be acknowledged as harm-involving, which then gives rise to the question of their regulation. On Liberty is unequivocal on this point: ‘If any one does an act hurtful to others, there is a prima facie case for punishing him’ (CW 18, p. 224). However, ‘there are often good reasons for not holding him to the responsibility; but these reasons must arise from the special expediencies of the case’ (CW 18, p. 225).

In the light of these remarks, the argument for speaker immunity must be similar to the argument for competition in chapter 5. Damage to the interests of others ‘can alone justify the interference of society’, but this does not mean ‘that therefore it always does justify such interference’ (CW 18, p. 292). The Mills are proposing to exempt a wide range of harm-involving assertions from punishment. The harm principle puts the burden on them to develop a utilitarian argument for this exception based on special expediencies.

The general idea of the utilitarian argument in CH2 is that the potential overall harm caused by the free expression of opinions is more than outweighed by its overall epistemic benefits.

I call the doxastic regime that the Mills defend in CH2 strict speaker immunity.

Some scholars characterise CH2 as an argument against censorship,Footnote ¹² but this is not the best way to put it.Footnote ¹³ The essay addresses a readership that lives in a society in which ‘the holders of power are regularly accountable to the community, that is, to the strongest party therein’ (CW 18, p. 219). This gives rise to the danger that the strongest party uses legislative power to impose ‘its own ideas and practices as rules of conduct on those who dissent from them’ (CW 18, p. 220). But not only that. The majority has its own ‘means of tyrannising’ apart from the ‘acts of the public authorities’ (CW 18, p. 219). The Mills call this ‘social tyranny more formidable than many kinds of political oppression … since … it leaves fewer means of escape, penetrating much more deeply into the details of life, and enslaving the soul itself’ (CW 18, p. 220). The worst blow to individual liberty would be a union of social tyranny and political control. Individuals do not yet suffer the combined invasion from public and magistrate because the ‘majority have not yet learnt to feel the power of the government their power, or its opinions their opinions’ (CW 18, p. 223). But the danger that the majority will soon learn to feel said power is imminent.

In sum, underlying CH2 are reflections about the constitution of Victorian society. The Mills make what I call a supposition of overwhelming conformity (or simply overwhelming conformity), that

Only after introducing overwhelming conformity do the Mills develop a kind of general epistemological case analysis which is intended to demonstrate that ‘silencing the expression of an opinion’ amounts to the ‘peculiar evil’ of ‘robbing the human race’ (CW 18, p. 229).

By structuring CH2 as a case analysis, the Mills give the impression that they have developed an a priori proof of almost mathematical cogency of everyone’s unrestricted right to express opinions.

The Mills distinguish three cases: A set of opinions O can be true, it can be false or it can be partly true and partly false. As fallibilists, the Mills assume that we can never be certain about the objective truth value of O. And here is the nub of the case analysis. It is supposed to show that regardless of the objective truth of a set of opinions O, it does more epistemic harm than epistemic good to suppress the expression of O. This appears to be uncontroversial for the first case, when O is true. But even if O is false or contains just an iota of truth, critical engagement with O in open discussion promotes on balance the epistemic ends of truth and understanding. In analogy to the game theoretical term ‘dominant strategy’, I call this the dominance argument: namely, regardless of the objective truth value of O, it is always better not to restrict the expression of O.

The dominance argument concludes that any restriction of the expression of an opinion is wrong because it deprives us of an opportunity to exchange error for truth or the opportunity to obtain a ‘clearer perception and livelier impression of truth’ or both.

Utilitarianism ‘holds that actions are right in proportion as they tend to promote happiness’ (CW 10, p. 210). From a utilitarian point of view, the dominance argument in itself cannot provide a complete justification of strict speaker immunity. The Mills claim that liberty of thought and discussion tends to produce truth and understanding, but they don’t explain in any detail how epistemic progress promotes happiness.Footnote ¹⁴ They merely state that all useful opinions are true (the ‘truth of an opinion is part of its utility’, CW 18, p. 233) and that no false opinion is useful (‘no belief which is contrary to truth can be really useful’, CW 18, p. 234). They do not propose that all true opinions are useful or that no true opinion is harmful. It is by no means guaranteed that more knowledge produces more happiness. Why do the Mills fail to discuss this point? The reason is that CH2 is not about knowledge in general but about the truth and justification of received opinions. I will come back to this important issue in §4.

From a utilitarian viewpoint another surprising aspect of the case analysis is that it does not refer to epistemic consequences but to epistemic opportunities. Silencing deprives us of opportunities to discuss opinions, but whether this entails an overall loss depends on the costs of having such opportunities, the size of the possible gain and the likelihood of obtaining its benefits in comparison to alternative options. Thus, the loss of an opportunity is not necessarily bad. Take the example of the flat earth doctrine (Veber Reference Veber2021). Even if wrong, engagement with it can improve one’s argumentative skills in general and one’s understanding of sciences and the natural world in particular. But utilitarians must not only examine whether an option taken by itself has positive value; they must also ask which alternative is best. It must be shown that dealing with flat earth theory is the locally or globally best approach for the achievement of given epistemic aims. Only then would it be epistemically harmful to suppress it by legal prohibition or moral coercion. Of course, no one seriously proposes to silence the flat earth doctrine. But this has to do with the perceived harmlessness of the view. Silencing its expression would violate the harm principle and thus be wrong.

In a footnote, the Mills limit ‘the fullest liberty of professing and discussing’ to assertions that one holds ‘as a matter of ethical conviction’ (CW 18, p. 228).

This conviction requirement is an important filter against epistemic pollution. Not being allowed to say what one truly and deeply believes – no matter how foolish, perverse or immoral this may seem to others – is a state of unfreedom and arguably detrimental to one’s ‘mental well-being’ (CW 18, p. 257). Speaker immunity isn’t meant to protect the production of truth-indifferent, manipulative messaging. Liberal democracies need not tolerate the attempts of autocratic regimes or domestic groups to undermine their institutions by the mass production of pseudo-news, hate speech or mad narratives. Quite the reverse, they must protect the public sphere from pollution resulting from the automated or human-made generation of fictitious or mendacious content on behalf of sinister domestic or foreign interests.

However, the conviction requirement is not a very strong filter. It demands that senders honestly articulate their subjective degree of belief in a proposition, without requiring them to offer anything in terms of evidence or reasons. Thus, true believers have the perfect freedom to generate and express wild assertions which must be checked and, as the case may be, cleared away by others. Without an internalisation of costs, there will be an oversupply of wild assertions. Even if, in rare cases, some of them turn out to be true (or clearing away falsities has the benefit of better understanding), it will very likely involve high epistemic opportunity costs. The resources needed could have been used more advantageously by engaging with more serious or better substantiated claims.Footnote ¹⁵

In short, even if the case analysis were valid (and we were indeed losing a valuable epistemic opportunity by any silencing), it would not have shown that all silencing obstructs epistemic progress. Therefore, the dominance argument fails to support the strong version of strict speaker immunity.

3 Challenges to the infallibility argument

At the beginning of CH2, the Mills introduce the famous infallibility thesis:

This thesis supports the following reasoning: We improve our beliefs by systematically considering reasons for and against them, but we can never attain absolute certainty. At some point, we might decide that particular claims do not deserve further attention. For instance, based on our current understanding of physics, we deem perpetual motion machines to be impossible. Therefore, patent offices do not examine applications for such devices. In a sense, the patent office refuses them a hearing, but as it is not prohibited to publicly argue for the possibility of such machines or to try to construct one, it is still possible to make the case. Society must keep the opportunity for proposing any conceivable claim open because all permanent silencing of discussion would be an assumption of infallibility.

Like most people, I find this reasoning persuasive. However, I doubt that the infallibility thesis provides sufficient support for the strong version of strict speaker immunity. My point is similar to the following counterargument which the Mills themselves raise and try to rebut (CW 18, pp. 230–233).

The counterargument denies that all silencing assumes infallibility. Society might only claim to be sufficiently certain that the silenced assertions are wrong and that their propagation does serious harm. Note that by ‘silencing’, the Mills mean ‘measures intended to obstruct the propagation of opinions’. It may be legal or even socially accepted to hold or discuss certain silenced opinions ‘in the narrow circles of thinking’ people’ (CW 18, p. 241). Thus, the reason for strict speaker immunity cannot be that any silencing would categorically refuse the silenced opinion any hearing whatsoever. The overwhelming majority may just want to avoid the widespread diffusion of pernicious opinions (CW 18, p. 241).

Accordingly, the rejoinder holds that

Thus, even if it were generally true that speaker immunity is beneficial, society could introduce local restrictions of doxastic liberty when the further propagation of a very likely false claim would be highly dangerous or harmful. For instance, genocide denial is illegal in certain jurisdictions. Such legislation is incompatible with strict speaker immunity; but as it is usually allowed to criticise the evidence or to offer counter-interpretations in academic debate in such jurisdictions, the prohibition of genocide denial does not necessarily assume infallibility on the part of the legislator.

The Mills respond to the counterargument that there is

I call this the greatest difference claim. The Mills maintain that doxastic restrictions undermine the basis of knowledge and understanding and are therefore self-defeating.

Without the freedom to discuss things properly, we may overlook relevant reasons. This is the general – and generally plausible – idea behind the assurance claim.Footnote ¹⁶ However, the problem with the claim is its unhealthy dose of hyperbole. Let’s take a closer look. According to the second part of the claim, we are not warranted to assume a proposition to be true without complete liberty. This is clearly incorrect. Through perception (or reasoning), ‘a being with human faculties’ can undoubtedly obtain rational assurance of being right without discussion. For instance, the fact that A heard B make a friendly remark justifies the belief that B did indeed make a friendly remark. One would be justified in acting on this assumption without further discussion. Of course, perception is not infallible, and the liberty of contradicting helps us correct misperceptions. If B is generally unfriendly, A might suspect that they misheard or misinterpreted B’s remark, in which case they might wish to discuss the matter with C. Nevertheless, perception is a reliable source of justified beliefs in a very broad range of cases.

The first part of the assurance claim states that we are justified in assuming a proposition to be true due to complete liberty. However, this also appears to be problematic. After all, the justification for assuming the truth of a proposition is not the liberty to contradict it, but the evidence or reasons in favour of it. The justification for assuming the truth of an observation lies in the right kind of causal relationship between the observers and the observed object rather than in some liberty right. Doxastic liberty is only relevant in terms of assurance insofar as it permits discussions which work as reason-revealing mechanisms. Moreover, someone’s liberty to contradict and disprove our opinion does not guarantee that they will exercise this right and, indeed, confirm or disprove our opinion. Complete liberty therefore provides no sufficient justification for the assumption of truth. Last but not least, even if all participants exercised their complete liberty to confirm or disprove a proposition, we would not necessarily be justified in assuming its truth. If all discussants lack competence, the discussion may well fail as a mechanism for revealing relevant reasons (Fantl Reference Fantl2018).

Complete liberty therefore provides neither a necessary nor a sufficient justification for the assumption of truth. I conclude that the assurance argument does not give support to the greatest difference claim. To take up the former example, if a government is convinced that there is strong evidence that some historical event constituted a genocide and if it is convinced that genocide denial is harmful, it may suppress its propagation without necessarily assuming infallibility or undermining the basis of knowledge.

4 The power-critical account: dissident speaker immunity

As mentioned in §2, the Mills open CH2 with the supposition of overwhelming conformity but then seem to shift their focus to a kind of a priori argument for the strong version of strict speaker immunity.

In this section, I will bring together the two strands of argument and contend that the purpose of strict speaker immunity is to provide protection against excessive conformity.

Closer inspection of the case analysis lends considerable weight to this interpretation. The Mills distinguish between ‘received’ and ‘silenced’ opinions, which appear in pairs. They set the truth value of received opinion as false in the first branch (CW 18, pp. 229–243), true in the second branch (CW 18, pp. 243–252), and partially true in the third branch (CW 18, pp. 252–257). A silenced opinion negates a received opinion or parts thereof. Silencing is the response of the overwhelming majority in defence of received opinions. The most important received opinions are ontological and normative commitments constitutive of a society’s basic power structures; these fundamental intellectual commitments constitute what I call a hegemonic ideology. I take this to be the primary meaning of the term ‘received opinion’ because it is in line with what the Mills say about social tyranny.Footnote ¹⁷ Received opinions in the primary sense aren’t just represented in law and social practice but also inculcated in individual minds and bodies; they occupy space in the ‘inward domain of consciousness’ (CW 18, p. 225; see also CW 18, p. 282), and they shape or dominate how we feel, how we perceive our bodies, how we navigate public spaces, how we frame our life plan and who we socialise with. They are crucial for how individuals see themselves and the world, how they explain events and structures, and what they deem to be valuable or essential in life. Thus, received opinions provide essential parts of each individual’s script for how to be, what to think and what to desire. This is why they matter considerably for human well-being.Footnote ¹⁸

A good example for a received opinion in its primary meaning is the belief that the legal subordination of the female sex to the male is justified (CW 21, p. 261). Such subordination is connected to a multitude of institutions, social practices and engrained beliefs that complement and support it; however, it also contradicts other received opinions, which in Subjection of Women the Mills call the ‘principles of the modern world’ (CW 21, p. 323).Footnote ¹⁹ The hegemonic ideology of Victorian society (1869) is thus reflectively unstable. I will say more about reflective stability in §5.

How precisely do the Mills conceive the silencing of opposing opinions? They point out that historically silencing has often been aimed at the total or temporary obliteration of a doctrine (CW 18, pp. 236–239). But silencing can also be less violent and radical. Those who are interested in maintaining ‘all prevailing opinions outwardly undisturbed’ find a ‘state of things very satisfactory’ in which oppositional views ‘never blaze out far and wide, but continue to smoulder in the narrow circles of thinking’ people (CW 18, p. 241). In this less radical sense, silencing consists in restricting the publicity of certain doctrines (CW 18, p. 241), even when they are legal and perhaps socially accepted in certain arenas. Such marginalisation stabilises power ‘without the unpleasant process of fining or imprisoning anybody’ or absolutely interdicting ‘the exercise of reason by dissentients’ (CW 18, p. 241).

The Mills also highlight those responsible for the silencing: the state and the public. A state is a collective agent with political authority over the population within its territory; the public is the term the Mills use to describe an imagined community of like-minded people who share certain norms, values and beliefs. The agents of moral coercion are socially encouraged and approved self-appointed enforcers who punish perceived deviations from received opinions. They police the doxastic space by signalling to like-minded people that a person with ‘unprevailing opinions’ is bad (has problematic views or a non-conformist lifestyle), cannot be trusted and must therefore be banned as far as possible from important positions and opportunities in society. This induces people to disguise their unprevailing opinions ‘or abstain from any active effort for their diffusion’ (CW 18, p. 241).

In short, silencing involves actions by the powers of society (the state and the people) that aim to restrict the publicity of dissident thoughts through punishment, such as legal penalties and moral coercion, in defence of established beliefs.

Towards the end of CH2, the Mills observe that proponents of received opinions praise certain forms of verbal aggression as authentic expressions of ‘honest zeal and righteous indignation’ (CW 18, p. 259) but criticise them as vulgar if used by their opponents. This is an example of the ‘worst offence’, as they call it, when the powers of society stigmatise their opponents as ‘bad and immoral men’ (CW 18, p. 259). To help balance power asymmetry in the doxastic realm, the Mills propose turning the tables by granting dissenters more discretion in the choice of their rhetorical means.

Would this include contemporary practices, like no-platforming or de-platforming? Some commentators argue that they are incompatible with the spirit and letter of On Liberty in general, citing CH2 as evidence (Veber Reference Veber2021). According to the power-critical account, however, the answer depends on the distribution of power in society.

Imagine a group of Victorian women shouting down established academics or politicians in power. This would be wrongful silencing if they were self-appointed enforcers acting in defence of received opinions about the place of women in society. However, if they were proponents of minority views, say feminists, and were disturbing a lecture about the innate inferiority of women, then their intervention would not technically constitute silencing in the sense of CH2. As a general rule, dissenters lack the means to effectively stigmatise a speaker who belongs to a hegemonic group. Although disruptive actions may still be considered rude, foolish or immoral, it is the dissidents who enjoy speaker immunity in a society marked by overwhelming conformity, not the proponents of received opinions.Footnote ²⁰ The Mills were aware that power is a matter of degree and that power relations in society change. With growing influence, the minority’s privilege to denounce or disrupt their opponents disappears. From a Millian perspective, those who wield a critical amount of stigmatising power must not declare that their opponents are ‘bad and immoral men’, whose opinions are beyond the pale and must not be heard. Moreover, power is relative. People who have little power in society at large may be very powerful within narrow circles, and in general they must not use this power to silence dissent. The imagined Victorian feminists are permitted to shout down proponents of the hegemonic ideology, but they are not permitted to silence each other in discussions between free and equals.

In sum, the power-critical account states that CH2 successfully justifies strict immunity for dissident speakers through the combination of the following four reasons:

No harm: The perlocutionary effect of different forms of harmful speech – such as genocidal language, negative stereotypes, objectification and slurs – requires power on the part of the speaker. By definition, dissenting opinions are largely powerless in a society with an overwhelming majority. Aside from causing displeasure, dissidents cannot directly harm members of the majority through the expression of opinions.Footnote ²¹

Epistemic benefit: As dissident opinions are by definition less prevalent in cultural practices and public discourse, they are more likely to be epistemically productive. They raise ideas that challenge prevailing views, thus promoting knowledge and understanding. In contrast, hegemonic opinions are well-known and deeply ingrained in the minds and characters of the overwhelming majority and are represented in society’s basic institutions. Consequently, it is less likely that their expression will provoke ‘a real discussion, descending to foundations, and not resting satisfied with vague and general assertions’ (CW 21, pp. 275–276).

Rationality: By definition, dissident opinions are unpopular with the overwhelming majority. As popular opinions are mostly shaped in one form or the other by socialisation beginning in earliest childhood, they are typically ‘rooted in the feelings’ (CW 21, p. 261). Dissident opinions, in contrast, can only spread if they overcome the resistance of acquired emotional attitudes by the power of reasoning. It may be easier for the overwhelming majority to avoid cognitive dissonance through silencing an epistemic vanguard than by changing their acquired behaviour.

No authority: The harm principle is supposed to regulate ‘the dealings of society with the individual in the way of compulsion and control’ (CW 18, p. 223). Self-appointed enforcers and officials are formally or informally authorised to imprison people or to exclude them ‘from the means of earning their bread’ (CW 18, p. 241).Footnote ²² In contrast, dissidents cannot punish the members of the overwhelming majority ‘in the form of legal penalties, or the moral coercion of public opinion’ (CW 18, p. 223). The harm principle is not meant to regulate the behaviour of dissenters.

In short, the members of the overwhelming majority have the means and the motivation to suppress dissident assertions that are generally harmless and epistemically productive when expressed and discussed. Thus, society has an overriding epistemic interest in granting moral and legal immunity to dissident speakers.

5 Dispensing with the supposition of overwhelming conformity: a robust principle of speaker immunity

Supposing overwhelming conformity, the Mills encourage us to be thankful towards dissidents: ‘If there are any persons who contest a received opinion, or who will do so if law or opinion will let them, let us thank them for it, open our minds to listen to them’ (CW 18, p. 252). Citizens of a society with received racist opinions, should thank anti-racist dissidents for showing them the truth. Likewise, citizens of a non-racist society should thank racist dissidents for deepening their understanding of their non-racist doctrines. Some may view the idea of thanking racists for sharing their views as bordering on moral perversion. However, if the argument of CH2 applies, there are no trade-offs regarding speaker immunity and no epistemic risks to consider. Apart from the direct epistemic benefits which result from the discussion of dissenting views, there are important indirect effects. Generally, it is good to know what others think about a topic, even if their views appear to be wrong or immoral or both. Knowing about others’ convictions enables us to predict their behaviour more accurately and provides important information for our own decisions about how to deal with them. If there were no penalties for expressing immoral views, the need to hide them through semantic and pragmatic machinations would be reduced. There would be no need for dog whistles and fig leaves (Saul Reference Saul2024). Obversely, other members of society would save the time and effort that would otherwise be needed to decipher such coded language. Speaker immunity thus tends to reduce the mental load of concealment and unmasking.

Without overwhelming conformity, however, the dominance argument can no longer be used to justify the rigid prohibition of all silencing of dissident opinions. The interplay of power, opinion and epistemic progress becomes more complex. Different scenarios of disunity are conceivable. To name just two, the population could be divided into two or more factions of comparable strength competing for state power (pluralist division), or the state could be under the effective control of a minority that owns the means of production, including the mass media (oligarchic dominance). I shall focus on pluralist division in the following.

Can the argument of CH2 be adapted to this alternative scenario? Recall that the Mills write in chapter 1 that liberty ‘as a principle, has no application to any state of things anterior to the time when mankind have become capable of being improved by free and equal discussion’ (CW 18, p. 224). It is based on ‘utility in the largest sense, grounded on the permanent interests of man as a progressive being’ (CW 18, p. 224). The crucial questions then are (a) whether in principle societies characterised by pluralist division are capable of improvement through discussion and (b) whether they are necessarily on a reliable and irreversible path of improvement through discussion. Some interpret CH2 in this way: If ‘wrong opinions gradually yield to fact and argument’ (CW 18, p. 231), progress is assured. This looks like naïve optimism about the epistemic outcomes of discussions in the real world (Ahlstrom-Vij Reference Ahlstrom-Vij, Fricker, Graham, Henderson and Pedersen2020, p. 408). But the Mills are not naïve. They recognise the possibility that the ‘spirit of liberty, in so far as it resists such attempts [of improvement], may ally itself locally and temporarily with the opponents of improvement’ (CW 18, p. 272). Moreover, they don’t deny that discussions can be unproductive or epistemically harmful (CW 18, p. 257), nor are they naïve about the biases of discussants. Nonetheless, there is a strong Millian presumption for a robust principle of speaker immunity in ‘societies being capable of being improved by free and equal discussion’ (CW 18, p. 224), as liberty ‘is the only unfailing and permanent source of improvement’ (CW 18, p. 272).

In a first step, I will explain the content of the robust principle by outlining its three necessary conditions; in a second step, I will examine how the Mills justify their claim that the principle promotes progress more effectively than its alternative, the use of force for the attainment of progressive goals.

According to the robust principle, speaker immunity must not be restricted unless three conditions are fulfilled.

1. (1) Any restriction of the problematic view X must be compatible with the fullest liberty to express and discuss X among competent people. The Mills use the assurance argument to support the claim that all silencing is wrong. Apart from the problems highlighted in §§2 and 3, the argument is flawed because the Mills themselves do not assume that all silencing necessarily suppresses any discussion of certain topics (see §4). Nevertheless, important lessons can be learned from the (failure of the) argument. The correct conclusion is not that all forms of silencing are necessarily wrong, but that silencing any discussion of a particular claim is wrong. An enlightened state may penalise the expression of certain problematic views (like genocide denial) in a wide range of social situations, provided that it permits open discussion among competent people (like historians, contemporary witnesses and so forth). Understood in this way, the assurance argument provides the first necessary condition for a justified restriction of speaker immunity.

2. (2) Any restriction of the problematic view X must be based on strong reasons for presuming dissident opinions are incorrect. The restrictions should be based on beliefs ‘which have reached the point of being uncontested’ (CW 18, p. 250) among competent people. Otherwise, there is the danger that ‘progress, which ought to superadd, for the most part only substitutes, one partial and incomplete truth for another’ (CW 18, p. 252).

3. (3) The compliance costs of any restriction of the problematic view X must not exceed a certain threshold. In the Principles of Political Economy, Mill mentions impatient reformers

Not being allowed to say (or hear it said) what a person truly and deeply believes is arguably detrimental to their well-being. If a sizeable part of the population is not permitted to express what they feel to be true, resentment, disobedience and even backsliding may ensue. Therefore, no exception to speaker immunity should be allowed without substantial evidence that (a) expressing doctrines perceived as immoral or wrong poses a vital threat to valuable parts of the institutional order or causes a serious harm to others, that (b) non-authoritative means of influence have been exhausted,Footnote ²⁴ and that (c) coercion is unlikely to be counterproductive. The robust principle takes (a)–(c) as the third necessary condition for a justified restriction of speaker immunity. Taken together, the three necessary conditions are sufficient to justify the suppression of opinions.Footnote ²⁵

In short, a proponent of speech restrictions must prove that competent people agree the opinion is untenable and that propagating it would pose a substantial risk to individual rights or institutional stability. At the same time, it must keep spaces for rational deliberation open, must have tried other means of influence without success and must have reasons to believe that coercion will have no counterproductive effects.

We can now see how a robust principle of speaker immunity is superior to its alternative, the use of moral coercion or legal penalties to force progressive legislation on an unwilling population. This reasoning draws on Millian reflections on the nature of an ideal society and how to attain it. While it may ‘be easier and shorter to get possession of the government’, it is preferable to get possession ‘of the intellects and dispositions of the public’ (CW 3, p. 799). In Considerations, Mill argues that the value of a society depends on ‘the qualities of the human beings composing the society’ (CW 19, p. 389). A robust principle of speaker immunity is essential for achieving an ideal society. This is because, according to the Mills, the ‘mental and moral, like the muscular powers, are improved only by being used’ (CW 18, p. 262). In order to enhance their mental faculties, citizens must constantly be pushed beyond the boundaries of unthinking belief or doxastic indifference. A despot ‘of superhuman mental activity’ would transform society into a mass of ‘mentally passive people’ (CW 19, p. 400). Thus, the best form of government cannot be progressive despotism, but rather ‘one in which the whole people participate’ (CW 19, p. 412), providing them with the opportunity to develop individually and collectively. After all, it is better to have a population that is satisfied with and convinced of progressive principles than one that is simply forced to obey them.

If the core idea of CH2 is sound, then progressive principles gain acceptance in the population primarily through critical engagement, which is burdensome even for members of an overwhelming majority. Recall that the Mills ask them to overcome their aversion and be grateful for the dissenters’ service to the community.Footnote ²⁶ They must sacrifice certain pleasures of power, such as employing ‘unmeasured vituperation’ against dissenters or stigmatising ‘those who hold the contrary opinion as bad and immoral men’ (CW 18, p. 259).Footnote ²⁷

In a pluralist society, engaging with dissenting opinions is arguably even more challenging. Different groups compete for control of political power. It is possible that the other group(s) will use their power to suppress one’s own group. Engaging with their positions in a calm and honest manner, as the ‘real morality of public discussion’ requires, can be perceived as a dangerous concession of political ground and influence.

Therefore, it is crucial that the factions in pluralist societies recognise the robust principle of speaker immunity, meaning they should only suppress assertions in strictly circumscribed cases. Otherwise, there is an acute danger that changing majorities will use state power to force progressive or reactionary doxastic regimes on the population, as the case may be. This would not result in steady progress based on reason, but rather a cycle of partisanship.

On Liberty predicts that, if all goes well, there will be a ‘gradual narrowing of the bounds of diversity of opinion’ (CW 18, p. 250) with a robust principle of speaker immunity in place. In the best form of society, all or almost all people fully understand the meaning of and embrace the reasons for a reflectively stable hegemonic ideology. Such a society would be evolutionarily stable because reactionary doctrines could not successfully penetrate a population whose members all or almost all embrace a hegemonic ideology in reflective equilibrium.Footnote ²⁸