Sunday, March 26, 2023

Should harmful speech be restricted? More on Mill and Free Speech

Via Flickr

A while ago, I wrote an extended analysis of John Stuart Mill's defence of free speech. As noted in that piece, Mill is not, as he is sometimes perceived to be, a free speech absolutist. He thinks you can restrict some forms of speech, particularly forms of speech that incite violence, amount to fraud or cause harm.

This limitation on free speech is not surprising. Mill's discussion of free speech comes in the midst of his book On Liberty which is all about the Harm Principle and how a necessary condition for any coercive interference with the behaviour of another person is that the behaviour causes harm to others. What's more, Mill's defence of the Harm Principle is, itself, conditional upon his more general utilitarian philosophy: he thinks a blanket application of the Harm Principle is consistent with the greatest happiness of the greatest number and, furthermore, that even when conduct is harmful to others, coercive interference is justified only if it serves the greatest happiness principle.

But Mill's acceptance of this restriction on free speech raises problems. When, exactly, is speech harmful? If someone reports research findings suggesting that members of certain ethnic minority groups are less likely to do well on standardised tests, or are more likely to engage in crime, are those claims harmful? Can they be censored (or 'cancelled' to use a modern term)? In this article, I want to take up some of these questions.

I will start by discussing an argument put forward by Melina Constantine Bell which would allow for the censorship/restriction of bigoted insults on the grounds that they are harmful and fail to express a proposition or opinion that can be countered with other speech. Bell is adamant that her position does not justify the restriction of offensive speech or speech that expresses a genuine opinion. I will raise some doubts about this and use it as a springboard for discussing some puzzles that arise from using harm as a criterion to restrict speech. In particular, I will argue that (a) it is very easy for merely offensive speech to morph into harmful speech and (b), following Mill, whether speech is harmful is itself a contentious claim and depends on a number of factors that may weigh against censorship. 

1. Bell's Basic Argument

Bell argues that bigoted insults are harmful and, since the harm cannot be redressed or countermanded by more speech, they can be rightfully restricted because of that harmfulness. In defending this view, Bell adopts an interpretation of Mill's harm principle that is broadly consistent with the interpretation I outlined in a previous article. Roughly, she thinks that Mill's harm principle serves as an initial and necessary filter on any attempt to coercively restrict the actions of another person. If an action is harmful to others, then you have a prima facie case for coercively restricting it. But you don't have a sufficient case for restricting it. To get a sufficient case for restricting it, you need to establish that the restriction would be consistent with Mill's underlying utilitarian philosophy, i.e. that it serves the greatest happiness of the greatest number. That is not a foregone conclusion. Restrictive policies may cause more harm than they avert. This is a particular concern when it comes to the restriction of speech, as we will see later on.

This means that Bell appears to adopt the following argument in favour of the coercive restriction of bigoted insults (this is not formally stated in her article; this is my reconstruction):

  • (1) If speech is (a) harmful to others and (b) its coercive restriction is consistent with the greatest happiness principle, then it can be coercively restricted.
  • (2) Bigoted insults are harmful to others.
  • (3) The coercive restriction of bigoted insults is consistent with the greatest happiness principle.
  • (4) Therefore, bigoted insults can be coercively restricted.

Some of the terminology here needs to be explained. A bigoted insult is a term or phrase, grounded in prejudicial attitudes towards certain groups of people, that is intended to insult or cause offence. The N-word is, perhaps, the quintessential example, but any racial, ethnic or gendered slur would potentially count (e.g. 'bitch', 'slut' and so on). Bell also includes a long-ish discussion of sexist and racist jokes in her article, suggesting that they too count as bigoted insults.

When is something harmful to others? Here, Bell adopts Joel Feinberg's famous definition of a harm as a 'setback to life interests'. Everyone has life interests, i.e. stakes in certain things such as their physical health, their career prospects, their financial situation. If conduct undermines and sets back those life interests, it is harmful. This means that harmful conduct is reasonably serious in terms of its effects. It is distinguished from merely offensive conduct which causes some temporary mental distress or disgust, but does not set back life interests. That said, both Feinberg and Bell concede that offence could morph into harm, in certain cases, where the offence experienced is 'severe, prolonged and constantly repeated'. We will return to the problems created by this concession later on.

Finally, what is meant by 'coercive restriction'? Bell includes, obviously, the use of legal punishments (imprisonment, fine etc) but also moral disapprobation. Moral disapprobation can include shunning, excluding, condemning or cancelling a person for what they have done. She includes the latter on the grounds that it involves the use of social power to change behaviour in a way that bypasses dialogue and reasoned debate. I am little bit uneasy about including all forms of moral disapprobation within the realm of coercion. For example, I think strong moral condemnation and criticism is fine and not necessarily coercive. But I agree that certain forms of disapprobation, such as shunning or cancelling, have a coercive nature to them. Furthermore, since a lot of the contemporary debate about free speech concerns attempts to deplatform or cancel people, it makes sense to include them within the terms of the argument being put forward.

That's the argument in its outline form. How can it be defended?

2. How Are Bigoted Insults Harmful?

Bell focuses primarily on premise (2) in her defence of the argument. Her main point is clearly stated:

...bigoted speech causes direct harm to individuals, and plays a central role in maintaining and perpetuating institutional oppression and injustice, it causes not merely subjective unpleasant mental states or offence, but also serious harms that usually remain unrecognized both in law and in public discussion. 
(Bell 2020, p 7)


There are two main claims here. The first is that bigoted insults cause direct harm to the people targeted by them. The second, which may arguably be an extension of the first, is that these insults play a key role in reinforcing systems of institutional oppression and injustice, which are also harmful to the people targeted by the insults.

In defence of the first claim, Bell focuses on the case study of sexist humour and its associated insults to women. She reviews the findings from two papers by Thomas Ford and Julia Woodzicka (this one and this one, if you're interested). These papers find that women often fare worse on cognitive tasks as a result of sexist jokes (compared to control groups), and experience a greater sense of self-objectification after hearing sexist jokes, which is linked to mental health issues. In addition to this, studies find that men exposed to such jokes are more willing to cut funding to women's causes. Away from the humour case study, Bell also reviews evidence from the US suggesting that African Americans do worse on health outcomes because of discriminatory stressors, which is likely to include (Bell is admittedly speculative on this) bigoted insults.

In defence of the second claim, Bell draws heavily from Miranda Fricker's concept of epistemic injustice. Such injustice arises when people that are oppressed by social practices cannot meaningfully share knowledge and ideas that might lead to the reform of such practices. Fricker distinguishes between two forms of epistemic injustice:

Testimonial Injustice: This arises when people's testimony is unfairly deemed less credible because of who they are, e.g. police being less likely to trust the testimony of a black person.
Hermeneutical Injustice: This arises when people lack the concepts they need to interpret and explain their experiences (thus giving them less voice in public conversations and debates), e.g. women lacking the concept of 'sexual harassment' to explain what had been happening to them for centuries.


Bell argues that bigoted insults contribute to epistemic injustice and thereby serve to reinforce systemic oppression. While one insult may not make a big difference, the repeated use of them to target particular groups serves to 'intimidate, stigmatize or exclude' them from social practices and debates. They do not get to participate, in equal terms, in the marketplace of ideas.

If we accept that bigoted insults are harmful -- and this seems plausible, along the lines suggested by Bell -- it does not automatically follow that they ought to be restricted. We still have to clear the hurdle of premise (3) and show that their coercive restriction is consistent with the greatest happiness principle. And, before we do that, we have to recognise the obvious Millian objection. Mill's defence of free speech is all about the critical value of truth and truth-seeking practices, and the problem of imperfect censors. As a general rule of thumb, the Millian position is that bad speech, even harmful speech, should be met with more and better speech, not censorship and restriction.

Bell's response to this is that bigoted insults do not 'communicate a viewpoint that serves as a basis for consideration and discussion'. In other words, you cannot meaningfully respond to them with more and better speech. You cannot challenge their accuracy. You cannot introduce evidence and argument to rebut them. So the standard Millian position does not apply to their use. Bell has a nice thought experiment to illustrate her point:

Suppose a group of white college men yell a racist, sexist insult (e.g. N----- whore!) at a female Black student walking across campus at night. What is she to say to counter that speech? "No, actually, I never have been paid for sex, so 'whore' is inaccurate besides being derogatory to sex workers; and 'N-----' is a word with a history connected to intimidation, terrorism and murder of Black people, and which, because of false beliefs about the intelligence and work ethic of Black people, has operated to deny them recognition of their equality and full personhood". Would that prompt an apology? Would it make her attackers recognize the truth of the words spoken...? The likely effect would be to place the female student in danger of physical violence, and/or invite further insults and mockery. This is not an exchange of ideas...and does not bring anyone closer to truth. 
(Bell 2020, 14)


This strikes me as an effective example and shows that, in some contexts, more speech is not the answer. If we generalise the point and accept that it applies to all uses of bigoted insults, then Bell's argument goes through. Some speech can be restricted, even within Mill's framework, and bigoted insults look like a good candidate for this.

But should we generalise the point? I'm sympathetic to some of this, and I think that many uses of bigoted insults are illegitimate and can be rightly restricted. That said I have some concerns about this argument. I'll discuss two major ones in the next two sections but, before that, I want to briefly mention four other concerns.

First, I worry that a prohibition on bigoted insults would require a careful assessment of the intentions and context in which insulting terms are used. It is, obviously, widely documented that oppressed communities sometimes use bigoted insults in a transgressive way and, presumably, this is permissible. While it will often be obvious when this is happening, sometimes it might not be and there may be borderline cases. In addition to this, sometimes someone may use a term that is insulting, and is taken by others to be insulting, even without directly intending this to be the case. Debates about the use of the 'N-word' strike me as a good illustration of this. The dominant opinion nowadays seems to be that you cannot even mention the word in the context of a debate about how it functions or its history. You can only refer to indirectly through the euphemism 'N-word'. But the arrival at this dominant view has taken time (and varies across cultures). Is it obvious, for instance, that a linguist should be coercively restricted from mentioning the 'N-word' in a public speech about the history of bigoted insults? You might say 'no' but it's not clear to me what Bell's argument would say about this example.

Second, and related to the point just made, whether a word or phrase counts as a bigoted insult, is subject to change over time and, consequently, may be a legitimate object of debate in at least some contexts. For instance, the word 'colored' is now taken be insulting in the US (due to its association with Jim Crow laws) and in the UK (due to its historical use in colonialism). But this wasn't always the case. It was once an acceptable and 'positive' term (hence the NAACP). Also, in some countries, for example South Africa, it is an acceptable term, used to denote those with mixed-race heritage. So someone could use the term without intended insult and it may be worth debating and engaging with them about its history to explain why it may be deemed insulting in some contexts. Coercively restricting the use of the term may actually backfire in those cases if such coercive restriction is deemed overbearing and is poorly explained.

Third, while some phrases are undoubtedly insulting and convey no meaningful propositions or viewpoints, Bell's use of sexist and racist humor as an illustration of the problem with bigoted insults, strikes me as more problematic. It seems to me that jokes can convey worldviews and propositions, albeit often indirectly. Sometimes it is worth engaging with and debating those perspectives. In addition, whether a joke counts as racist or sexist will vary depending on the context in which it is made, the intentions of its maker and, possibly, their identity. For instance, someone may make a joke that is, on its face, sexist, but is used, ironically, to poke fun at sexist attitudes and prejudices. Should we restrict such uses of sexist humour? I don't think so and I think there would be a lot of practical problems in implementing such restrictions.

Fourth, the plausibility of Bell's argument will depend, a lot, on the kinds of coercive restrictions she has in mind. What should happen to the college students in her thought experiment, for example. Should they be reprimanded and shunned? Subjected to bias-sensitivity training? Fined? Expelled? Imprisoned? Some punishments might be consistent with the greatest happiness principle but others would not be. Striking the right balance will be tricky.

Those, as I say, are some concerns with the argument. There are two others I want to explore in more detail.

3. Offensive Speech vs Harmful Speech

The first is the distinction between offensive speech and harmful speech. Bell is adamant that her argument does not apply to merely offensive speech. But what might be an example of offensive speech that is not harmful? I'm not sure, but some examples discussed in Bell's article and prominent in contemporary debates might include claims about race IQ differences, and gender differences in aptitude or ability.

I take it as a given that such claims are offensive to members of the groups targeted by them. If I were told that I was naturally or innately inferior at mathematics, I would be insulted. But would I be harmed?

The distinction between harm and offence is cashed out by Feinberg (simplifying a bit) as the distinction between an objectively determinable setback to life interests (harm) vs a temporary episode of mental distress or inconvenience (offence). Perhaps the argument might be that the claim about inferiority in mathematics causes merely temporary distress. It doesn't actually stop me from pursuing my life interests.

Unless, of course, we follow the argument made above, and acknowledge that the repeated utterance of such claims could create an epistemic social environment in which I and others come to believe in my innate inferiority and so censor or alter my behaviour in a way that does knock back my life interests. So the claim can be both offensive and, eventually and more insidiously, harmful.

Related to this, it doesn't seem right to say that offensive speech is always of merely temporary distress. Take another example. Suppose I am a religious conservative and I am repeatedly exposed to statements such as 'gay sex is good sex' or 'those who believe in God are evil'. I find these statements offensive, given my religious convictions. Maybe this initially causes merely temporary distress. I can get on with my life after hearing them. But it doesn't take too much to imagine a world in which I become fixated on these statements. I constantly hear them being echoed and repeated by my social peers. My religious leaders tell me that the people making these claims are evil and a threat to society. I cannot sleep at night knowing that these people are out there, believing and espousing such things. It makes me shake with rage every time I hear them speak. I sweat. I get anxious. I find it hard to breathe. What started out as mere temporary distress has become life debilitating harm.

Maybe you think this slide from offence to harm is unrealistic. But I'm not so sure that it is. I think, given the right combination of personal characteristics and social conditions, a claim that we might otherwise dismiss as offensive could become harmful in the Feinbergian sense of causing a setback to life interests. What's more, some people might be able to manipulate the social context so that speech that is initially dismissed as merely offensive becomes classified as harmful. Some people, critical of contemporary 'woke' culture, argue that this is exactly what is happening in relation to some kinds of speech. There is an incentive to catastrophise and overinterpret certain speech types so that they get seen as harmful and can be restricted.

But to attribute this tendency to purveyors of 'woke' culture alone is a mistake. If it is possible to transform offensive speech into harmful speech, then anyone looking to restrict certain forms of speech on the grounds of harm -- be they religious conservatives or woke liberals -- has an incentive to do so. It is, consequently, not clear if the line between offensive speech and harmful speech can be maintained.

4. If speech is true, but harmful, what then?

Of course, just because speech is harmful, it does not follow that it should be restricted. Bear in mind premise (1)(b) of the argument stated above. The restriction must also comply with the greatest happiness principle. One reason why Bell favoured the restriction of bigoted insults was because they did not state a point of view or proposition. They could not be countered by more speech. Offensive propositions -- such as the examples given above -- are different. Even if we accept that they are harmful, they still state points of view that can be debated and discussed:

Suppose in a college class, or in a debate in the public square, a white man expresses the view that men are naturally better at mathematics, or that they make better political leaders...Expression of these opinions should be protected, and others should have the opportunity to refute them. Otherwise, both points of view appear equally valid, with two sides in disagreement but with no fact of the matter to be established or recognized. 
(Bell 2020, 17)


This is the standard Millian position. Speech should, ordinarily, be countered with more speech. But, the interesting thing about Bell's defence of this view (and others that I have seen) is that they often focus on cases in which the offensive views are presumed to be false. Countering them with more speech is hoped to expose the falsehood. But what if they actually turned out to be true and, as a result of this truth, continued to be harmful to the people affected by them? 

To be clear, I am not granting that the views given in the quote from Bell are true. My point is simply that, suppose after reasoned debate and argument, most people come to believe that they are true. The fact that they are true doesn't mean they cannot be harmful. It could be true that I have terminal cancer with no prospect of a cure but accepting this truth could harm me. As I have discussed before, there is some evidence to suggest that people with positive illusions about themselves do better when it comes to disease prognosis than those with more realistic/true beliefs.

So what then? What happens when we have a true statement/opinion that is also, clearly, harmful? In that case, we seem to have a tension between the defence of free speech (which is a instrumental one focusing on the value of truth) and the harm principle.

It's worth returning to the original source on this matter. Mill, obviously, argued that pursuit of the truth was consistent with the greatest happiness principle. The only way to arrive at true beliefs was through vigorous and uncensored debate about the merits of those beliefs. Against those who claimed that false beliefs could sometimes be valuable or useful, he argued that there was a circular logic at play. Whether a belief is valuable or useful is, itself, a truth claim that can only be established through the free exchange of ideas:

This mode of thinking makes the justification of restraints on discussion not a question of the truth of doctrines, but of their usefulness; and flatters itself by that means to escape the responsibility of claiming to be an infallible judge of opinions. But those who thus satisfy themselves, do not perceive that the assumption of infallibility is merely shifted from one point to another. The usefulness of an opinion is itself matter of opinion: as disputable, as open to discussion, and requiring discussion as much as the opinion itself. 
(Mill, On Liberty, Ch. 2, p 24)


In other words, there is no real tension between truth and the alleged harmfulness of an opinion because you can only assess harmfulness through the pursuit of truth.

I am not sure how satisfying that is. Isn't it also the case that the harmfulness of a bigoted insult is an opinion whose truth can only be established through debate and the exchange of ideas? Does it follow that bigoted insults, therefore, should not be restricted since their harmfulness is up for debate? Maybe. But if that isn't what should happen in the case of bigoted insults then why should it happen in the case of true-but-harmful opinions?

Of course, there is another way of looking at it. We could accept that an opinion is true, and harmful, and still wonder whether the best way to mitigate its harm is to restrict it. Whether an opinion is harmful depends on its uptake: how those affected or influenced by the opinion respond to it and interpret it in their lives. Consider the following claim (deliberately chosen on the grounds that it is less likely to invoke certain connotations than similar claims):

Claim: Men, on average, do much worse than women on standardised tests. What's more, this outcome does not seem to be attributable to educational advantage or social difference. It appears to be innate.


Let's assume this is true. I am a man. Does this claim harm me? Well, it depends on how I and others interpret it. It concerns men, as a group, not me as an individual. So I could argue that it has little or no bearing on me. All that matters is how I do on the tests, not how men as a group do. Of course, other people might be biased against me, as a member of the male group, and their bias may harm me no matter how positive and upbeat I am, but if they interpret it along the lines I do, the harm to me could be minimised. The harm could be further minimised if they were penalised for interpreting the claim in a way that discriminates against me as an individual.

This might, in most cases, be the appropriate Millian response to the truth-harm dilemma. If a claim is true, but harmful, we still shouldn't censor it, but we should consider ways in which its harmful uptake can be minimised. The free exchange of ideas can continue, and harm can be mitigated.

But what if it is more likely than not, given social circumstances and history, that a true claim will be taken up in a harmful manner? What if it would be really hard to suppress or mitigate harmful uptake? That's the really challenging case for the Millian and the challenge that would arise in any society with historic, institutional bias against groups targeted by harmful-but-true opinions.

1 comment:

  1. It has been asserted that morality can't be legislated. Most indications over many years have supported the assertion. So, if, and only if, continuance of freedom of speech is a fundamental right under the U.S. Constitution, hate speech; bigoted speech and so on are to be tolerated, so long as people subjected to them are not unlawfully discriminated against in some other way. I was never a lawmaker. But, years ago, language I crafted appeared in administrative code in my home state. If we avow things must change we need to be certain of what we avow.