Sydney Law Review
The Case against Free Speech
Free societies employ a variety of institutions in which speech is heavily regulated on the basis of its content in order to promote other desirable ends, including discovery of the truth. I illustrate this with the case of courts and rules of evidence. Of course, three differences between courts and the polity at large might seem to counsel against extending that approach more widely. First, the courtroom has an official and somewhat reliable (as well as reviewable) arbiter of the epistemic merits, while the polity may not. Second, no other nonepistemic values of speech are at stake in the courtroom, whereas they are in the polity. Third, the courtroom’s jurisdiction is temporally limited in a way the polity’s may not be. I argue that only the first of these — the ‘Problem of the Epistemic Arbiter’ as I call it — poses a serious worry about speech regulation outside select institutions like courts. I also argue for viewing ‘freedom of speech’ like ‘freedom of action’: speech, like everything else human beings do, can be benign or harmful, constructive or pernicious. Thus, the central question in free speech jurisprudence should really be how to regulate speech effectively — to minimise its very real harms, without undue cost to its positive values. In particular, I argue against autonomy-based defences of a robust free speech principle. I conclude that the central issue in free speech jurisprudence is not about speech, but about institutional competence. I offer some reasons — from the Marxist ‘left’ and the public choice ‘right’ — for being sceptical that capitalist democracies have the requisite competence and make some suggestive remarks about how these defects might be remedied.
One major accomplishment of the post-Enlightenment revolutions in moral and political thought that began in the 18th and 19th centuries is that the value of free speech is now widely taken for granted on all ends of the political spectrum in the capitalist democracies. This consensus, I will argue, has now gone badly awry, even by Enlightenment standards. All things considered, much, perhaps most, speech, in fact, has little or no positive value. So the idea that its free expression is prima facie a good thing should be rejected. And since the only good reasons in favour of a legal regime of generally free expression pertain to the epistemic reliability of regulators of speech, we should focus on how to increase their reliability, rather than assume, as so much of popular and even some philosophical discourse does, that unfettered speech has inherent value. If much of what I will henceforth call ‘non-mundane’ speech were never expressed, little of actual value would be lost to the world — or so I will argue.
That my topic is ‘non-mundane’ speech bears emphasising at the start. The category of what I will call ‘mundane speech’ is quite central to human life, and almost never noticed by the law, even in obviously unfree societies. Mundane speech is the kind of speech that facilitates our discharge of the ordinary and unnoted business of daily life. It is the kind of speech that allows us to arrange to meet our friends at a particular restaurant at 8 pm, or the speech that gets our kids to finish their homework. Most speech is mundane speech, and most societies, even unfree ones, do not bother with mundane speech. Non-mundane speech — speech about matters of political and moral urgency, speech that purports to be of aesthetic value, speech that purports to help us understand the truth about matters of societal importance, speech that is thought central to self-formation and the good life — is what is really at issue in debates about the regulation of speech.
My claim is that most speech of this kind in fact has little or no net positive epistemic value (that is, value for helping us discover the truth) and not enough non-epistemic value (either for the speaker or listeners) to justify its expression, regardless of the costs to social welfare. In a slogan: most non-mundane speech people engage in is largely worthless, and the world would be better off were it not expressed. The internet is the final confirmation of this truth, I shall suggest, though the plausibility of this claim predates that technological innovation.
I propose to get to that conclusion by starting far from it; namely, by calling attention to the fact that the Western liberal democracies are rife with institutions that view massive restrictions on speech as essential to realising the ends of free societies. In universities and schools, for example, no one thinks the classroom should be turned over to unregulated expression of opinion, without regard to cognitive value, civility, or pedagogical purpose. But I shall focus here on just one central institution: the law courts. In courts, the idea that the ‘unfettered interchange of ideas’ (to quote a typical formulation of the United States (‘US’) Supreme Court) has any value is rejected from the start. Moreover, it is rejected without considerable controversy; indeed, without attracting much critical notice or comment. Let me explain.
If speech were actually ‘free’ in the courts — that is, unrestricted by state power — then there would be almost no need for most rules of evidence. A judge or judges would be summoned as the official arbiters, triers of fact (if different) would be assembled, and then the parties, or their advocates, would say anything and everything, adduce any and all evidence that they believe might help their case. There might, of course, be rules that constitute, as US free speech doctrine has it, ‘time, place and manner’ regulations, but beyond that, advocates could say anything, call any witnesses, offer any evidence, engage in any rhetorical trope that suits their cause.
The preceding is not the rule in any liberal democracy, but I shall focus on the American case for two reasons: first, it is the one perhaps most notable, or notorious, for investing a lay jury in almost all cases, both civil and criminal, with figuring out the truth about what transpired; and second, in its free speech jurisprudence, the US accords the widest legal latitude to speech of any Western democracy. Nazis, racist sociopaths, and misogynistic pornographers are all constitutionally protected members of the fabled ‘marketplace of ideas’ in American society at large. Or as Adrienne Stone and Simon Evans aptly put it in a review of Australian free speech jurisprudence, but referring, in this instance, to the exceptional character of American law: ‘No other constitutional system of freedom of expression confers so much protection on ... unpleasant, caustic, insulting, and vulgar forms of speech’.
Most common law jurisdictions, indeed most legal systems in the advanced liberal democracies, assign fact-finding to professional jurists in many (if not all) cases. But the US is notable in this context for investing discovery of the truth almost exclusively in the hands of ordinary people — the same people, it bears emphasising, who are supposed to discover the truth about the great political issues of the moment, whether it is climate change, or universal health care, or the most just tax policy. How the US approaches free speech in the context of the jury trial is, I suggest, revealing about free speech and its value. And the central fact about the rules of evidence in the US is that they are predicated on distrust of lay juries. Ironically, this coexists with a popular culture that celebrates the ideal of a ‘jury of one’s peers’, but the reality is that the evidentiary rules reflect, again and again, doubts about the competence of ordinary people as triers of fact.
More precisely, the rules reflect doubts about the epistemic capacities of lay jurors; that is, their ability to arrive at the truth unless the speech they are exposed to is carefully controlled. The rules are predicated, in other words, on what the philosopher Alvin Goldman has called ‘epistemic paternalism’. Paternalistic rules substitute the rule-maker’s judgement for what is in the interest of the subject for the subject’s own judgement on that score. Epistemically paternalistic rules substitute the rule-maker’s judgement about what would be in the epistemic interests of the subject — that is, his or her interest in discovery of the truth — for the subject’s own judgement. The American rules of evidence are deeply paternalistic in this epistemic sense, and, as a result, deeply hostile to free speech.
It is basic to the Federal Rules of Evidence in the US, for example, that evidence that might make more or less probable the existence of some material fact (that is, ‘relevant’ evidence) can nonetheless be excluded by the judge if the risk that it will confuse or prejudice the jury is too great (that is, render them less capable of rendering an epistemically reliable verdict). So, for example, very gruesome crime photos might be excluded; so too facts about a defendant’s criminal history; and so too evidence that a defendant in a homicide settled a civil wrongful death action related to the same incident. All these pieces of evidence are relevant to assessing the truth of what happened, but in each case the rules of evidence take the view that they can be excluded since lay jurors might draw improper inferences from them. For example, they might let disgust at the gruesomeness of the crime outweigh their sober weighing of the evidence against the defendant; or they might fail to understand that the standards of proof are different in the civil and criminal context, and so misinterpret the probative value of a civil settlement in a criminal prosecution.
The distrust of jurors in the US is most apparent in our baroquely complicated hearsay rules. Officially, the rules prohibit the use of out-of-court statements to prove the truth of what they assert — the US legal system does not trust laypeople to adequately discount the probative value of evidence not subjected to ordinary cross-examination and other putative trial safeguards. But in reality, hearsay statements can come into evidence under one of about 30 different exceptions (though they are not all called ‘exceptions’, for reasons that need not concern us here), exceptions all predicated on the idea that this particular kind of out-of-court statement has some alleged circumstantial guarantee of reliability. So, for example, an out-of-court declarant who is in a state of excitation is thought to be reliable (a somewhat bizarre supposition, though it is probably right that those in a state of excitation will not lie); so too one who is talking to her doctor about her medical condition; so too one describing how he feels at that very moment; so too one who implicates himself in a crime, and so on. The only reason the hearsay rules are so complex in the US is that lay people are the main triers of fact, and, on the one hand, the legal system does not trust ordinary people to weigh hearsay evidence properly; yet, on the other hand, we all know that hearsay evidence is often useful and probative. The US rules strike a somewhat odd, but revealing, compromise: judges are charged with being ‘gatekeepers’ for what the ‘ordinary’ people on a jury can hear. And, with a few exceptions, the judge’s main charge is to see that the ordinary folk hear only what will help them discover the truth, and nothing that will impair their epistemic task.
Expert testimony is another area where the US approach betrays its distrust of jurors and assigns the judge an increasingly complicated ‘gatekeeping’ function. For many years, most US courts followed what was known as ‘the Frye rule’ (after a New York case from the 1920s), according to which putatively ‘scientific’ evidence could be admitted if it was ‘generally accepted’ in the scientific community as reliable. Frye itself involved lie detectors tests, which are still not admissible in federal courts in the US. The Frye standard, in other words, was a proxy criterion for scientific reliability: it did not require courts to assess the quality of the science on which a putative expert would rely, but it did require that such evidence have been ‘generally accepted’ by scientific experts. In 1993 in Daubert v Merrell Dow Pharmaceuticals, the US Supreme Court made clear that the Frye standard was not the relevant standard under the Federal Rules of Evidence. Rather, judges were to admit scientific evidence if it was produced by scientifically reliable methods. The Court even cited, as an aid to the lower courts, the philosopher of science Karl Popper for the idea that genuine scientific theories should be ‘falsifiable’. The late US Chief Justice William Rehnquist, a member of the conservative wing of the Court, remarked in dissent, ‘I defer to no one in my confidence in federal judges; but I am at a loss to know what is meant when it is said that the scientific status of a theory depends on its “falsifiability,” and I suspect some of them will be, too.’
The admission is telling on several fronts. The late Chief Justice’s reasonable concern indicates that the arbiters of worthy speech in the courts may not, in fact, be competent to adjudicate the epistemic value of the speech they are charged with evaluating. The worry is confirmed by the confusions in the Supreme Court’s Daubert opinion itself, since it equates Karl Popper’s view with that of the philosopher Carl Hempel’s, even though they are actually competing views about the nature of science! That is, the US Supreme Court declares that one mark of the scientific reliability of evidence is that it is ‘falsifiable’ or ‘verifiable’, citing, respectively, Popper and Hempel. Unnoticed is that Popper and Hempel’s views conflict, that falsifiability and verifiability are not the same thing. Popper was a diehard empiricist and follower of David Hume: he accepted that induction from past experience to conclusions about laws of nature was irrational. Even if every swan we have ever seen was white, it wouldn’t follow that all swans were white, unless one assumed that the future must be like the past. But our only evidence that the future must be like the past involves the very same inference at issue in the swan case. Popper concludes that, since inductive inference is not rational, all we can hope for in science is claims that are falsifiable, that is, claims for which we can imagine possible evidence that would contradict the generalisations.
Popper’s view had the unhappy consequence of rendering many uncontroversially scientific claims ‘non-scientific’. Take a claim typical in astronomy, like ‘there exists a black hole’. This cannot be a scientific hypothesis according to Popper since it is not falsifiable. No amount of evidence could falsify a claim about the existence of an X without an inductive inference of precisely the kind forbidden by the Humean argument against induction. In part because of these worries, other philosophers of science, like Hempel, proposed a different criterion of scientific or cognitive content, a kind of empirical verifiability. They proposed the possibility that the hypothesis could be supported (at least partially) by empirical evidence via a logic of confirmation — the kind of evidence deemed rationally insufficient on the Hume/Popper view. That Hempel’s view, the opposite of Popper’s, was conflated with it in a court opinion specifying the criteria for the admissibility of scientific evidence might seem a good argument for more freedom in the admissibility of evidence.
In reality, though, almost nothing has turned on this conflation. The main practical consequence of Daubert in the US has been the creation of an industry of lawyers and technical advisors devoted to litigating the scientific reliability of putative experts. Since the standard for admissibility for scientific evidence is, in effect, that it be good scientific evidence, and since in any case presenting complicated causal questions, the admissibility of the scientific evidence is often decisive, many pre-trial motions are now devoted to adjudicating whether each party’s science is really good science.
But why impose such barriers on ‘scientific’ evidence in the first place? Why not permit a ‘marketplace of ideas’ in which any alleged expert can take the stand and make his or her best case, and then the jury sorts it out? Again, anyone familiar with American political culture can imagine a good answer: because in a ‘free market’ of ideas, in which any paid hack or shill can argue anything, you end up with a majority of the population dubious of Darwin’s theory of evolution by natural selection and sceptical that human activities have any effect on global warming.
To sum up the discussion so far: ordinary people in the US are charged with making findings about the truth of what transpired in disputed incidents that come before the civil and criminal courts. No one thinks they should be exposed to a freewheeling marketplace of ideas, an unregulated and unrestricted presentation of evidence and arguments; instead, the basically untrustworthy laypeople are subjected to the paternalistic care of a judge, whose job it is to decide what they can safely hear that might actually facilitate their correct findings of fact, allowing for their cognitive infirmities and other biases. There is no free speech in the courtroom, and (almost) no one thinks there should be.
By contrast, when these same laypeople are asked to choose a President, someone who will decide American tax policy, whether to go to war, the correct approach to climate change, and who should get healthcare, the basic constitutional posture in the US is that everyone (whether person or corporate entity) should be able to say almost anything, and without any meaningful restrictions on the advantages that accrue to those with wealth and access to the major media of communication. Can we explain why the public sphere should be a free-for-all of distortion and misinformation, as it too often is in the US, while the juridical sphere, where matters of life and death, freedom and incarceration, wealth and penury, are decided, is not?
Obviously, the two cases are different in lots of ways, but I think it will be useful to be clear about what those differences really are and why they might matter. There are, it seems, three primary differences between the courtroom and the polity at large relevant to our purposes. First, the courtroom has an official and somewhat reliable (and, importantly, reviewable) arbiter of the epistemic merits, while the polity may not. Second, no other non-epistemic values of speech are at stake in the courtroom, whereas they are in the polity. Third, the courtroom’s jurisdiction is temporally limited in a way the polity’s may not be. I shall refer to these three crucial differences between the courtroom and the public sphere, in shorthand, as ‘the Problem of the Epistemic Arbiter’, ‘the Problem of Non-Epistemically Valuable Speech’, and ‘the Problem of Time’. I shall take these up in reverse order, which also corresponds to an ascending order of significance.
A functioning legal system must be sensitive to the amount of time required to adjudicate disputes, and the American rules of evidence are not unusual in acknowledging temporal considerations as a proper ground for excluding otherwise probative evidence. But functioning polities also operate under time constraints, with major decisions about war and tax policy — decisions that affect far more momentous matters, like the well-being of tens of millions — being taken sometimes in a matter of months, not unlike complicated trials. More importantly, how much time should be accorded to a decision should be proportional to the moral magnitude of the decision’s outcome. Three months to litigate a complicated securities fraud case might seem just right or excessive, but what is at stake may only be the liberty of a handful of individuals and millions of dollars. Yet in less than 18 months after the 11 September 2001 (‘9/11’) terrorism attacks in the US, the US Government ‘decided’ to launch a war of aggression against Iraq, which killed and wounded at least several hundred thousand people, and turned several million people into refugees. Is it really true that temporal considerations were irrelevant in the latter case? Since what is at stake — the risks of great benefits and catastrophic costs — are often (probably more often) as great or greater in the context of political decisions than ones in courts of law, should we really think that the limited temporal horizon has special importance only in the juridical context? Such a conclusion strikes me as incredible.
It is often said — in an optimistic, but not necessarily realistic, spirit — that the ‘truth wins out’ over the long haul. That is usually complemented by John Maynard Keynes’ correct observation that over the long haul, we are all dead. Courts and societies should be sensitive to both points. Certain truths, such as truths about the causal structure of the world, will win out, because they cannot be ignored forever: no amount of argument will change the fact that the locked door will causally block your exit from the room. And no amount of propaganda will alter the causal reality of climate change. But if there are other truths — for example, about whether Nazi Germany or democratic Australia best promotes the well-being of its population — we have fewer guarantees about how the long haul will go. To be sure, non-Aryans might cause great difficulty for their Nazi overlords, but the Nazis had a solution to that problem. More realistically, the capitalist overlords of nominally democratic societies have their own, generally less final, solutions to the risk of causal disruption by the disgruntled masses. But we might all be dead, or immiserated, before the latter injustices make themselves known. That is another reason to think that temporal considerations are as relevant in the polity as in the courtroom.
Yet even if one thought the temporal pressures on courts rather than polities were morally more significant, that does not change the fact that reasons of epistemic paternalism are, quite explicitly, the primary reasons why American courts regulate what it is the triers of fact may hear. If we are to find important disanalogies between the courtroom and the polity when it comes to the regulation of speech, they must be sought elsewhere.
The massive restrictions on speech in the courtroom impose no limits on the kind of speech often thought morally and legally important not because of its epistemic value, but because of its non-epistemic value to the speaker or society. The speaker restricted in the courtroom can still vent on a street corner, or on a blog, or in an opinion piece in the local newspaper. So, too, regardless of hearsay rules and evidential prejudice, she may, outside of the courtroom, write poetry or upload videos of herself on YouTube. Courts may impose massive restrictions on speech for epistemic reasons, but those restrictions take away almost nothing from the two other most important values associated with freedom of speech: effective democratic self-government, and the various non-epistemic values of speech for the speaker and society — broadly eudaemonic and autonomy-based values.
This might seem a decisive difference between courtroom and polity when it comes to the regulation of speech, but the matter seems to me more complicated. First, and most obviously, the contribution of speech to valuable democratic selfgovernment itself depends on epistemic considerations. The promulgation of falsehoods, innuendo, and lies may undermine democratic self-government, not promote it. And second, while the eudaemonic value of expression is often apparent, that hardly settles the question of the relative priority of values the law should promote when it regulates expression. No one thinks the eudaemonic value of sexual orgasm, for example, settles the question of the moral or legal propriety of involuntary sexual contact, or public masturbation, or prostitution, so it is surprising that anyone would think the eudaemonic value of expression should be dispositive as to its value or the propriety of legal regulation. Arguments for free speech based on autonomy can sometimes be predicated dogmatically, but implausibly, on the priority of autonomy over all other interests and values. Unfortunately, the most plausible metaphysics and psychology warrants the conclusion that people are not autonomous; autonomy interests of persons are, I will argue, just certain kinds of eudaemonic interests. We should think of ‘free speech’ like ‘free bodily movement’: it has a lot of utility, but only within limits. Once the so-called ‘autonomy’ interests of speakers and listeners are put into the mix of competing eudaemonic considerations, the balance to be struck among them does not obviously favour free speech — or so I shall argue.
Let us take these points up in turn.
Speech can contribute to democratic self-government, but it can equally well contribute to fascism, genocide, and even less egregious kinds of injustice. Hitler was a very effective speaker, and so too were the radio journalists in Rwanda exhorting their comrades to finish off their Tutsi neighbours. In the US, purveyors of misinformation and ignorance like the TV personality Sean Hannity, the radio commentator Rush Limbaugh or the popular website operator Matt Drudge have tens of millions of followers and no doubt have some influence on how their audience votes. But only someone who thought the popular will had intrinsic value regardless of its basis or its content could possibly think a polity ruled by their fictions and half-truths justified a free speech regime. If democratic selfgovernment always led straight to moral catastrophes, who in their right mind would suggest that that form of government has any value? The often suppressed premise in all arguments for free speech based on the value of democratic selfgovernment is that a polity so governed will not lead to such moral abominations. But that premise, as far as I can see, is just wishful thinking, not a realistic assessment of how a regime of unfettered speech can influence a polity.
The American experience is instructive on this score, as a few salient examples will show. There is no meaningful controversy among scientists about the fact that human activities are causing potentially catastrophic changes to the world’s climate. The Intergovernmental Panel on Climate Change, for example, declared in 2007 that ‘[m]ost of the observed increase in global average temperatures since the mid-20th century is very likely [that is, more than 90% likely] due to the observed increase in anthropogenic greenhouse gas concentrations.’ Meta-analyses of the scientific literature, as well as surveys of climate scientists, confirm that roughly 98% of all researchers concur. By contrast, a 2008 poll found that only 47% of the American public had the true belief that global warming was due to human activity. The fact that only a minority of Americans had a true belief about global warming was due almost entirely to delusions among Republicans, the right-wing party in the US. In 2008, just 49% of Republicans said there was even solid evidence that the earth’s temperature was even rising (down from 62% in 2007!), while only 27% of Republicans said that global warming is caused by human activity. The massive public ignorance has certainly made it easier for the US to take no meaningful steps to address climate change.
In March 2003, the US, under the Bush Administration, launched an illegal war of aggression against Iraq; hundreds of thousands died as a result, and millions more have been displaced. The war was facilitated, domestically, by rampant false beliefs about connections between Iraq under the dictator Saddam Hussein and the terrorist group, al Qaeda, which carried out the 9/11 attacks on the US, among other crimes. Two months before the war began, a poll found that 68% of Americans held the false belief that Iraq was involved in the 9/11 attacks, with 13% even claiming that ‘conclusive’ evidence of Iraq’s involvement had been found. This was presumably a direct result of the Bush Administration’s strategy of linking Saddam Hussein’s Iraq to Osama bin Laden’s al Qaeda — occasionally explicitly, often implicitly, but generally unmistakably. By summer and early spring 2003, 45 to 52% of Americans said that they believed — again falsely — that the US had ‘found clear evidence in Iraq that Saddam Hussein was working closely with the al-Qaeda [sic] terrorist organization’. Researchers examining this incidence of false belief found, among other things, that watching the right-wing Fox news network (part of the Murdoch media empire) was ‘the most consistently significant predictor of misperceptions.’ For example, those who primarily watched Fox were twice as likely to believe that links between Hussein and al Qaeda had been discovered. By contrast, those who generally watched or listened to public TV or public radio were 3.5 times less likely to believe that links to al Qaeda had been discovered.
Two related considerations might, however, be adduced in favour of unfettered expression to democratic self-government, notwithstanding the triumph of falsehood it sometimes makes possible. First, there is the idea, which we may associate with Friedrich Hayek, that unfettered speech by the masses is an important device for aggregating widely dispersed ‘information’. Second, there is the worry that governing elites can hardly be trusted to decide which opinions deserve to be part of the marketplace of ideas in a democracy. The second worry — a very serious (perhaps decisive) one, I hasten to add — is actually just an instance of the general worry about identifying a reliable Epistemic Arbiter, one who sorts inputs into deliberation based on their epistemic value, and not irrelevant considerations; as such, I shall postpone it to the final part of this article. The Hayekian point, by contrast, seems — at least initially — to be more specific to the value of free speech for democratic self-government, and so merits further consideration now.
The Hayekian objection can be parsed in two, depending on the kind of widely dispersed ‘information’ thought to be at issue. On the one hand, information about what I will call ‘basic needs’ — needs for food, shelter, and the basics for survival — is, indeed, very widely dispersed and needs to be freely available to avert catastrophes. As Amartya Sen’s research has shown, famines do not occur where there is freedom of the speech and press, since governments and fellow citizens are put on notice when people are starving. On the other hand, also widely dispersed in a population are individuals’ conceptions of their ‘interests’ apart from their basic needs — for example, their preferences, wants, and other values. Hayek himself held the implausible view (what I will call the ‘subjectivist’ view) that people are reliable arbiters of their interests, whether basic or non-basic. The subjectivist view has had legions of philosophical critics — from Plato and Aristotle in antiquity, to Hegel, Marx, Mill, Marcuse, Brandt, and Railton in the modern era — and has had almost no defenders apart from doctrinaire free-market proponents. People, to be sure, usually know what they want, but they often do not know what they need, or whether their wants serve their needs, and sometimes even how to get what they want or need. Everyone recognises that addicts or those brainwashed by religious or political cults will ignore even their basic needs, and so will fail miserably in meeting their interests. But overwhelming evidence from cognitive science also shows that even so-called ‘ordinary’ people make systematic mistakes in the instrumental reasoning required to meet both basic and non-basic interests. The philosopher Sarah Conly, in her recent defence of coercive paternalism, gives a useful summary of what are, by now, familiar empirical findings:
We are ... unduly influenced by the particular description used in the presentation of our options (more likely to choose a medical procedure with a 20 percent chance of success than one described as having an 80 percent chance of failure); unduly prone to think that we ourselves are less likely than others to suffer misfortune, even of something entirely random, like lightning; prone to miscalculate the value of a thing depending upon whether we do or don’t yet own it; prone to assuming things that have one superficial characteristic in common also have similarities throughout (commonly known as stereotyping).
Conly correctly sees the connection between this line of research and earlier critiques of human rational agency; as she puts it:
[w]e have already revised our view of human agency, following Marx, Freud, and the philosophical insights of feminism. What we see now, in light of contemporary psychology and behavioral economics, is that some further revision is necessary.
The import of these earlier lines of critique, however, was not simply that we do a poor job at instrumental reasoning about how to realise our basic and non-basic interests, but that our very conception of our interests, including our non-basic interests, is often an artefact of irrational social, economic and psychological forces; forces that, in addition, also can impede the realisation of even our basic needs.
That stronger conclusion also wins support from contemporary cognitive science. As the philosopher and cognitive scientist Jesse Prinz has argued, emotional responses drive our evaluative judgements, yet our dispositions to have particular emotional responses are artefacts of biology, as well as familial and cultural influences, over which we have little or any autonomous control. Without a doubt, people identify with their values, and so regulation that infringes on those values affects people’s eudaemonic well-being. But it can hardly be considered an infringement on their autonomy to regulate expression of those values, given that those values do not themselves result from autonomous choices.
Strikingly, the recent cognitive science literature complements the critique developed in the last century by Frankfurt School theorists like Max Horkheimer, Theodor Adorno, and Herbert Marcuse. They described the ways in which capitalist societies generate ‘wants’ and ‘interests’ in the population, thus actively shaping their values and desires. So, for example, Horkheimer and Adorno called attention, in particular, to the way in which market economies turn cultural products — novels, films, even music — into commodities that are designed to deliver predictable and ephemeral satisfactions, while also moulding their consumers’ sense of what has value. In an era when we are most of us now numb to the endless recycling of cinematic plots and emotions — no movie is a success, after all, if it does not lead to at least two or three sequels with the same name followed by the appropriate Roman numeral — consider how prescient Horkheimer and Adorno were to observe in the early 1940s that the ‘[c]ulture [industry] ... infect[s] everything with sameness’, noting that the ‘standardization and mass production’ of cultural products — such as ‘hit songs, stars, and soap operas’ — requires that they ‘conform to types recurring cyclically as rigid invariants’ so that the ‘details become interchangeable.’ ‘Its element is repetition’, as Horkheimer and Adorno write, a slogan that should be emblazoned in the sky over Hollywood and Bollywood.
The nightmare world of the Frankfurt School theorists — which, alas, is our world — is one in which people’s conception of their non-basic interests tracks what the market can deliver precisely because the market has nurtured people to have precisely those interests. (Sometimes people even confuse non-basic interests for basic ones.) There may be a natural tendency of persons to care about how they appear to others — Rousseau’s amour-propre captures one aspect of this. But there is no ‘natural’ interest in being extremely thin, or having an expensive wristwatch, or having teeth with a certain glistening whiteness: the aspects of appearance that count are artefacts of culture that, in capitalist societies, means the market. Yet many people conceive of their non-basic interests in ways that are responsive to the massive indoctrination by the capitalist advertising industry, which bombards human beings, from their childhood, with images and messages designed to determine that they will want what profiteers can deliver. This depraved cycle of artificial need and pointless consumption flourishes thanks to our fetish for ‘free speech’, which, in the US, extends even to substantial protections for commercial speech.
To sum up: the Hayekian argument in favour of ‘free speech’ in the service of democratic self-government — namely, that ‘information’ is widely dispersed and can only be heard in a regime with robust protections for free speech — is only partly successful. People are usually rather good at expressing their basic needs for food, shelter, and safety — at least in the extreme cases. But they do less well at the instrumental reasoning required to meet their basic needs, at least over the long term, and do extremely poorly when it comes to forming their non-basic interests, where they are largely hostage to extraneous forces, not autonomous selfdirection. The Enlightenment was often predicated on faith in the capacity of human beings for rational self-governance, and arguments for free speech based on democratic self-government, as well as Hayekian arguments, are torch bearers of this Enlightenment faith. The Enlightenment also valued the realisation of certain ends, such as the well-being and flourishing of all persons. Frankfurt School theorists like Marcuse share that Enlightenment ambition, but emphasise, correctly I think, that under current conditions it is not necessarily compatible with a libertarian approach to speech. As Marcuse famously put it: ‘society cannot be indiscriminate [in its tolerance of speech] where the pacification of existence, where freedom and happiness themselves are at stake’. As heirs of the Enlightenment, we should ask both when does free speech contribute to rational self-government and when does free speech contribute to human flourishing? Both those questions have, I suggest, been neglected in theoretical work over the last century.
If free speech’s contribution to democratic self-government also depends on its epistemic value — on the extent to which it is not simply an instrument of falsehood and misunderstanding, including self-misunderstandings — the same cannot generally be said, however, about the other values often ascribed to freedom of expression, namely, its contribution to individual well-being (which I will call, for short, its ‘eudaemonic’ value) and to the autonomy interests of speakers and listeners. In this section, I shall argue that the only interests at stake are eudaemonic in character, and that all so-called autonomy interests should really be treated as just eudaemonic ones.
I have already given some reasons in the prior sections for thinking that neither speakers nor listeners are actually ‘autonomous’; that they are, instead, mostly artefacts of social, economic, and psychological forces beyond their control, mere vessels through which the various prejudices of their communities or their personal histories pass. But here I want to make a stronger claim; namely, that the Kantian/Christian rhetoric about autonomy is a fiction: we are not autonomous beings, and so there are no ‘autonomy’ interests that could trump all other considerations. What we call ‘autonomy’ interests are simply our eudaemonic interests in speaking and listening as we desire.
At this point, we need to consider in some more detail the ‘autonomy’based defences of free speech to see what is at stake in such accounts. Their core idea is that persons are, in some sense, rationally ‘selfgoverning’ or ‘self-directing’, and that respect for that fact means that speech cannot be regulated on the ground that autonomous persons might act badly or cause harm on the basis of speech to which they are exposed. ‘Jews are parasites that destroy social well-being’ or ‘The poor deserve their fate, they don’t work hard enough’ or ‘Atheists are godless and immoral people who should be imprisoned’ should all be part of protected speech, because respect for the autonomy or the autonomous interests of persons requires that we allow the listeners to assess such claims (it also requires that we let the speakers articulate such claims).
In what sense are persons ‘autonomous’? Here I follow philosopher Susan Brison, who has documented many of the senses of ‘autonomy’ invoked in the literature defending worthless speech. As she notes, ‘autonomy’ has something to do with ‘self-government’ and ‘self-rule’. That might mean, as Scanlon puts it, that ‘[a]n autonomous person cannot accept without independent consideration the judgment of others as to what he should believe or what he should do.’ Or it might mean, as Nagel puts it, that ‘[t]he sovereignty of each person’s reason over his own beliefs and values requires that he be permitted to express them, expose them to the reactions of others, and defend them against objections.’ But can people decide to ‘accept’ without ‘the judgment of others’ what to believe? And is ‘each person’s reason’ really ‘sovereig[n] ... over his own beliefs and values’? I think the answers to both questions are ‘no’, and that the autonomy-based rationales for freedom of speech are predicated on a fiction.
There is a longstanding debate in modern philosophy of the past several hundred years about whether the idea of our freedom or autonomy can be reconciled with a scientific picture of how the world works. The classic form of the problem — the one that exercised, for example, Kant — was based on the supposition that all physical matter was governed by the deterministic laws of Newtonian mechanics, and thus everything that happens must happen; since we ourselves are composed of law-governed matter, so too for everything we do. Kant thought the only way to rescue genuine autonomy was to suppose that the will could also operate outside the law-governed realm of natural phenomena. How that was possible remains mysterious, as even Kant recognised, since his official position was only that if anyone is really autonomous, their actions would necessarily have to have as their source a will that stands outside the ordinary causal order of nature. But that leaves open the possibility that no one ever acts autonomously or responsibly.
The dominant view among philosophers — who, as a group, continue to be very fond of morality and freedom — is that our putative autonomy and moral responsibility is fully compatible with our will being causally determined. Of course, we now know that Newtonian mechanics is false at the quantum level (a fact that does not do much to help non-quantum humans!), but the anxiety about our autonomy has found new sources: in the influence of socio-economic forces, of the unconscious, of our emotions, and of the neurophysiology of the brain. But the central ‘compatibilist’ idea — the idea that our autonomy is compatible with causal determination — has always been that it suffices for autonomy that our choices be caused in the right kinds of ways: for example, by the desires or feelings we ‘identify with’, or based on a conception of the good that we accept (for whatever reason), and so on.
Thomas Nagel memorably described this compatibilist response as ‘even less plausible than’ the Kantian kind: ‘All such accounts fail to allay the feeling that, looked at from far enough outside, agents are helpless and not responsible’. Nietzsche famously remarked that ‘a thought comes when “it” wants, not when “I” want’, a claim that is phenomenologically indisputable: but that means all the thoughts that precede our actions, including our speech, have causal determinants that are unknown to us. Galen Strawson, arguing self-consciously in a Nietzschean vein, notes that everything we say and do is surely traceable to our ‘character’; to that amalgamation of antecedent cognitive, affective and conative states that comprises our sense of who we are — and those states are, of course, heavily affected by the ‘judgments’ of others. But, as Strawson notes, it is obvious that we are not responsible for our ‘character’; indeed, even if we try to modify our character, we are led to do so by our pre-existing cognitive, affective, and conative states for which we are not responsible, so even changes in our character originate from causal forces for which we bear no responsibility. We are, in short, not responsible for who we are, since we are the products of vastly complicated causal networks that extend well beyond us; but who we are determines what we do, including what we say. If there is any room for an ideal of ‘autonomy’ in a realistic picture of the human situation, it will be highly revisionary of the Christian/Kantian ideals that undergird contemporary ‘autonomy’ defences of free speech. That, in any case, is what I will assume for the remainder of this article, with one caveat.
The caveat is this: there is a metaphysically thinner notion of ‘autonomy’, also associated with Kant, according to which autonomy consists simply in the capacity ‘to use one’s intelligence without the guidance of another’. I will call this, obviously anachronistically (and a bit ironically), ‘Nietzschean autonomy’, since it is silent on questions of ultimate causal determination, and makes a distinction only between those who ‘think for themselves’ and those who simply think what others tell them to think. The basic idea here is clear enough: there is a difference between Jones who weighs evidence and makes up his mind no matter what his neighbours think, and Smith who simply defers to the opinion of the community. It may be, at the end of the day, that Jones’s Kantian ‘enlightenment’ is fully explicable in terms of the causal determinants of Jones’s beliefs and actions, but it is still true that Jones is not (to use Nietzsche’s term again) a ‘herd animal’, while Smith is. And I think we can, and should, agree with Nietzsche — and, as it happens, with Mill — that such enlightened individuals contribute to well-being. This does not, however, change the basic argument, since Nietzschean autonomy derives its value not from a kind of impossible exercise of autonomous choice, but from the value of those whose cognitive and affective capacities allow them to transcend the blinkered horizons of the masses.
Let us return, however, to the main line of argument. Speakers and listeners plainly take themselves to have something like autonomy interests, but as best I can tell these are really just certain kinds of ‘eudaemonic’ interests, that is, the interests of speakers and listeners in satisfying their own conception of what they want to say or hear, no matter the actual causal determinants of those interests. (If the speakers and listeners are capable of Nietzschean Autonomy, then the eudaemonic values promoted extend beyond the speaker or listener.) To be sure, humans are discursive animals, creatures who live in the domain of meanings and reasons and inferences, however imperfectly they traverse the territory (and however imperfect that territory really is). And because of our basic discursivity, our eudaemonic stake in being able to express our views is substantial. (The internet now confirms that latter point, alas, to the point of excruciating tedium.) But these eudaemonic interests in expressing ourselves have to be balanced against the costs of bad and worthless and harmful speech, such as those attendant upon our ‘brave new’ cyberworld of 24-hour often irresponsible, dishonest and twisted invective. Henceforth, I will treat the so-called ‘autonomy interests’ of individuals as just another kind of eudaemonic interest, one that certainly deserves substantial weight, but is just, ultimately, one consideration among many.
How might the eudaemonic interests in speech fare in such an assessment? Cyberspace provides a useful case study, because it makes vivid the character of both mundane and non-mundane speech as it really exists in the world, and so also makes vivid what is at stake in the regulation of speech with an eye to its eudaemonic value. Much of the speech on the internet is simply the cyber-version of the traditional mass media, so I want to put that to one side for the moment, and much of it is plainly mundane; for example, speech about gardening, cooking, home repair, dining out, and the like. With respect to the non-mundane speech that is really distinctive of the internet, let us be candid that it falls primarily into three categories:
(1) endless varieties of pornography, that is, the sexual depiction, in both images and words, of, inter alia, girls and women (and, to a lesser extent, boys and men) naked, having vaginal and anal intercourse, performing or receiving oral sex, and/or being bound, whipped, ejaculated upon, and otherwise humiliated;
(2) insult, abuse, and invective based on ideology, opinion, gender, race, ethnicity; and
(3) the expression of opinions about the moral, political, cultural, and aesthetic issues of the day, most of which simply regurgitates trite pabulum and the various au courant prejudices of whatever culture in which the speaker originates.
The last category might have some epistemic value, and sometimes the second might too. But all three can justify their existence primarily in either eudaemonic terms (in the case of pornography, the eudaemonic value at issue is primarily, though not exclusively, to the consumer; in the case of invective, primarily, though not exclusively, to the producer), or in terms of the putative autonomy interests of speakers and listeners — though realistically understood, these are just a particular kind of eudaemonic interest, as I have argued. Please do not misunderstand me: there is much that is no doubt wonderful and valuable about some pornography and some invective, but that they have taken over so much of the internet is obvious confirmation of the anti-Enlightenment thought that creatures like us are not primarily rational and are primarily instrumentalities of our very powerful drives, drives for sexual gratification, cruelty, domination, and so on. This is a fact about persons that serious social and legal policy should acknowledge, not wish away.
What might we conclude about the eudaemonic value of speech from the evidence of speech on the internet, assuming I have correctly characterised it? Three points stand out. First, much unmediated non-mundane speech has little epistemic value, though it has some. Second, the undeniable eudaemonic value of this speech comes at various costs — direct costs to the eudaemonic value for those harmed by pornography and invective, for example, and indirect costs to those harmed by the epistemic distortions that follow from pornography, invective and the endless repetition of silly opinions. Examples of the latter categories would include those victimised by sexual or ‘hate’ crimes after cyber-incitement; those subject to cyber-harassment of all kinds; and those harmed by social policies that can trace their origins to the massive orgy of ignorance and falsehood that is so much of the unmediated internet. Here, in particular, we should be alert to the effects of ‘group polarisation’, in which likeminded individuals who interact only with each other end up gravitating towards extremes. Discussion among like-minded individuals magnifies the effect, but if individuals with the same attitude are simply exposed to others with the same attitude, that alone can produce movement to the extreme. Various explanations for the phenomenon have been proposed: for example, that individuals are sufficiently sensitive to social comparisons that they will opt for the most extreme position that they deem to be dominant in their group to make sure they are included; or that like-minded individuals tend to come up with the best arguments for the position to which they are already disposed, but those arguments simply reinforce that antecedent position. Whatever the explanation, the phenomenon is real, and operates in alarming ways among communities of, for example, online misogynists and paedophiles. Third, and finally, the eudaemonic interests of speakers and listeners also have to be evaluated in light of the costs just noted, since, as argued earlier, what is really at issue is the ability of speakers and listeners to satisfy their conception of what they want to say and hear. Whether those eudaemonic considerations outweigh the harms to well-being of speech is surely an open question.
I do not want to overstate the points. The eudaemonic value of expression outside cyberspace requires more sustained consideration. And the internet certainly makes accessible plenty of real information and knowledge, but mostly in virtue of making access easier to the traditional mediated sources of information, such as reputable newspapers around the globe, and scholarly and other scientific research. The self-congratulatory rhetoric of bloggers and tweeters notwithstanding, the unmediated blather that is so much of cyberspace has added little net value to the world. It is in desperate need of an epistemic arbiter.
But what about John Stuart Mill, more than one reader is probably thinking? How can these kinds of sceptical considerations about free speech be reconciled with the arguments of Mill, the patron saint of liberty in the modern era, at least in the capitalist democracies? Alas, most readers have apparently not read Mill, and so do not realise that a case against free speech is, even by Millian lights, hardly out of bounds.
Mill, himself, was quite clear that his arguments for liberty presupposed certain background conditions among speakers and listeners, especially education and maturity — without such conditions being satisfied, liberty would be unlikely to maximise utility after all. In other words, speech per se has no value, its value resides in its consequences, and the consequences of speech given unfavourable antecedent conditions could be quite dire. Liberal capitalist democracies, of course, tend to assume, in a self-congratulatory spirit, that they of course meet the Millian conditions, but it is far from obvious that they do. Herbert Marcuse’s famous (or infamous) 1965 attack on excessive tolerance of harmful speech was essentially Millian in spirit: he denied that socio-economic conditions were such that Millian liberty of speech would actually maximise utility, he did not repudiate Mill’s basic rationale for free speech. Indeed, Marcuse was simply following Mill’s own explicit caveats about his argument:
[T]his doctrine [the Harm Principle as a restriction on government regulation of liberty] is meant to apply only to human beings in the maturity of their faculties ... For the same reason, we may leave out of consideration those backward states of society in which the race itself may be considered as in its nonage ... Despotism is a legitimate mode of government in dealing with barbarians, provided the end be their improvement, and the means justified by actually effecting that end. 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. Until then, there is nothing for them but implicit obedience to an Akbar or a Charlemagne, if they are so fortunate as to find one.
Partisans of Mill typically ignore this caveat, attributing it to British racism about its Indian colony (though it surely bears noting he thought that medieval Europe was in a comparable condition). But Mill’s perhaps parochial misperception of the facts on the ground are irrelevant; the really crucial point is that Mill himself, with this caveat, clearly acknowledged that liberty would only be utility-maximising under certain conditions. Mill does not support freedom of speech, simpliciter; he supports it when, given the background conditions, it actually makes human beings better off. Superficial readers ignore this, but it is quite clearly Mill’s own view (as Marcuse seems to have well understood).
That this is Mill’s view is not surprising if we pay attention to the particular arguments he offers for liberty of expression. Recall that Mill believes that discovering the truth (or believing what is true in the right kind of way) contributes to overall utility, and that a largely unregulated ‘marketplace of ideas’ (as it has come to be called) is most likely to secure the discovery of truth (or believing what is true in the right kind of way). Mill’s commitment to the so-called ‘marketplace’ is, however, based on three claims about truth and our knowledge of it. First, Mill thinks we are not justified in assuming that we are infallible: we may be wrong, and that is a reason to permit dissident opinions, which may well be true. Second, even to the extent our beliefs are partially true, we are more likely to appreciate the whole truth to the extent we are exposed to different beliefs that, themselves, may capture other parts of the truth. Third, and finally, even to the extent our present beliefs are wholly true, we are more likely to hold them for the right kinds of reasons, and thus more reliably, to the extent we must confront other opinions, even those that are false.
For this line of argument to justify freedom of expression, the expression in question must be related to the truth or our knowledge of it, and certain background conditions must obtain, that is, those exposed to speech must be able to evaluate it (hence Mill’s exception for children and the Indian subcontinent, which would be better off with ‘benevolent despotism’!). As the philosopher David Brink has emphasised, Mill assumed that valuable speech had to assist our deliberative capacities, our abilities as putatively rational agents to weigh evidence, reason logically, and draw appropriate conclusions. Much non-mundane speech has no plausible claim to enriching our deliberative capacities; Brink focuses on ‘hate speech’, but the point extends more widely. Mill also assumed, contrary to the evidence we now have, that people are largely capable of rational deliberation and evaluation of evidence; but if they are not, then the Millian argument no longer applies, as even he would have to agree.
Mill’s fallibilism also involves certain nuances that should not be forgotten. ‘All silencing of discussion is an assumption of infallibility’, he famously asserts, but that is false, as the example of speech restrictions in the courtroom demonstrates. We often silence discussion in contexts where we both have reason to assign a high level of credence to what we believe and where there will still remain opportunities for critical discussion and thus discovery of the truth — outside this courtroom, or this classroom, or even this electoral cycle. But let us also remember that even Mill did not actually accept the thesis about our fallibility in its strongest form. For Mill held that there is no reason to have a ‘free market’ of ideas and arguments in the case of mathematics (geometry in particular), since ‘there is nothing at all to be said on the wrong side of the question [in the case of geometry]. The peculiarity of the evidence of mathematical truths is that all the argument is on one side.’ This is all the more striking a posture in light of the fact that Mill is a radical empiricist, and so denies that there is any a priori knowledge: even logical and mathematical truths are a posteriori, vindicated by inductive generalisations based on past experience. On Mill’s view, then, there simply would not be any epistemic case for making room for the expression of opinions on which there is no contrary point of view that could make any contribution to the truth.
This last point is particularly important when it comes to some of the most contentious issues about speech regulation in free societies. In the US, for example, there was considerable controversy in the 1970s when ‘civil liberties’ advocates came to the defence of the right of American Nazis to march in a Chicago suburb with a large number of Holocaust survivors and their relatives. The civil liberties advocates took the position that ‘unpopular’ speech deserves protection. The Nazi speech was, fortunately, ‘unpopular’, but it was also false and harmful. The American Civil Liberties Union (‘ACLU’), which defended the Nazis, lost members over this incident, but that hardly proves the ACLU was wrong. The ACLU was surely correct in thinking that ‘unpopularity’ is not an epistemically reliable indicator of falsehood or harmfulness. But that does not mean that Nazi speech is not both false and harmful: as even a radical empiricist like John Stuart Mill could acknowledge, doing a simple inductive inference over the horrible experiences of the 20th century would support a confident conclusion that Nazi speech is worthless, that Nazis simply have nothing to say that is worth hearing. (In addition, their speech is understandably harmful to the Holocaust survivors exposed to it, though not only to them.)
That last observation brings us to a final ambiguity in Mill’s position. Liberty can be limited when it causes ‘harm’ to others on Mill’s view. But what counts as a harm, beyond the obvious cases of physical violence? Psychological harm is real harm, as anyone familiar with someone suffering from mental illness will recognise. Yet we certainly do not, as Mill understood, want to treat simple ‘offense to moral sensibilities’ as a case of harm to which the law might respond, since it would make a mockery of the value of liberty. But there is a partial solution to that worry. Harm to someone’s psyche does not warrant legal regulation, we can say, when either: (1) the harm derives entirely from violating the harmed person’s beliefs about how others should behave; or (2) the harm derives entirely from attacks on the beliefs the harmed person holds about the world. The first category deprives moralistic busybodies of a claim to protection; the second deprives dogmatists of all stripes from having a claim. But the characteristic harm associated with hate speech, for example, raises a colourable concern, since the harm results not from violating the target’s sense of how others should behave or the target’s beliefs about the world, but from the threat to the target’s sense of his or her own well-being and worth. Even a target who thought there was a moral right for Nazis to march and speak would still be harmed in the relevant sense. Pavlov’s dogs developed instinctive responses to the sound of bells; is it so surprising that survivors of Nazi concentration camps might suffer nausea, nightmares, and anxiety attacks when confronted with Nazis marching down the street? Mill, without the benefit of the empirical evidence of the 20th-century’s horrors, could hardly be criticised for not thinking about such cases. But if the ‘harm principle’ is still thought to be a legitimate constraint on state power, then it must surely accord due weight to the harm to the mental lives of its citizens, not just their physical well-being.
I conclude that Millian considerations do not block the sceptical considerations about free speech adduced so far, which brings us to the final issue — the Problem of the Epistemic Arbiter.
With respect to the non-epistemic values of speech, then, the challenge to its unregulated expression comes down to one issue: can we confidently develop a mechanism for the regulation of such speech with regard to maximising its epistemic (and other) value and minimising its harm to the well-being of others, including the eudaemonic interests of speakers and listeners? The best case for unfettered expression as a contributor to democracy is that any attempts to impose fetters are as likely to undermine as promote Enlightenment values of democratic self-government and the flourishing of human beings. The best case for unfettered expression in virtue of its value to the speaker is that any attempt to limit such expression with regard to other values — including the well-being of other people, as well as the well-being of the polity — is as likely to minimise overall well-being as promote it.
Courts and classrooms have epistemic arbiters, judges and teachers, respectively, who try to remedy the deficiencies of any regime of genuinely free speech. But to the extent judges and teachers fail in their role as epistemic arbiters, some of the goods associated with free speech — such as discovery of the truth or self-realisation — can happen elsewhere. There is only one serious argument against regulation of speech for all its pernicious effects and that is the worry that in society writ large we do not have a reliable epistemic arbiter, and, moreover, any attempt to designate one runs the risk of sacrificing all the other goods associated with free speech insofar as the arbiter is unreliable or makes too many errors.
That concern is the central theme of American free speech jurisprudence: ‘we’, it is said, do not trust the Government to decide what speech is worth hearing. Not trusting the Government is hardly an unreasonable posture, especially in a plutocracy like the US. But the reality, of course, is that even in America we trust the Government to do all kinds of rather alarming things, such as hiring and arming individuals to police the rest of us, even investing in them the power to invade our homes, put us behind bars, or shoot us dead if necessary. Empowering the State to decide that we can’t say certain things seems rather more trivial by comparison.
Or does it? One might think the crucial difference is this: as largely irrational but still discursive creatures, human beings can be influenced about what to do by what others say. Sometimes that influence may qualify as ‘rational’, sometimes irrational, but to the extent that state power has to be responsive to the will of its subjects, however indirectly, then the ability to speak freely is crucial to mobilising action, and thus affecting the direction of state power. When it comes to people’s basic needs, moreover, their ability to express them seems crucial, as Sen has shown, to insuring that such needs be met. If the State abuses our trust in the way it manages the police, a regime of free speech can, in principle, check that abuse, and hold the Government to account. When it comes to trusting the State to regulate speech, we are being asked to trust the State in a domain where there is no meaningful remedy if that trust is abused, apart from violent resistance or revolution.
These considerations should be taken seriously, though tempered by the realisation that many democratic societies, with robust free speech cultures, from Germany to Canada to Australia, employ such arbiters, and not primarily with regard to epistemic considerations, but rather moral ones, such as the protection of dignity or equality. The crucial point is that the worry at issue is only that the State may suppress or regulate speech that has some kind of value, not that any speech is presumptively valuable. Speaking, like everything else human beings do, can be for good or ill, benign or harmful, constructive or pernicious. Even if the ‘unpopularity’ of speech is not an epistemically reliable indicator of either its falsity or harmfulness, that hardly warrants the irresponsible libertarian position that we should tolerate the damage to truth and to the well-being of the victims of bad speech as necessary costs to be borne on behalf of insuring that possibly true, non-harmful, and otherwise valuable speech might be heard, even though unpopular. The real question is: can we, as free societies, regulate speech to minimise its very real harms, both to the well-being of persons and to the pursuit of welfare-enhancing social policies in the polity at large, without undue cost to the other values of speech?
If the key question in free speech jurisprudence is how to insure competent regulation of bad speech, given that much, maybe even most, non-mundane speech may turn out to be bad speech, then the question is not about free speech at all, but about political institutions. For the recurring worry in free speech jurisprudence is that state actors, even in democratic societies, will suppress speech not because it has little or no value, but for reasons unrelated to its epistemic or social value. In Marxist theory, the worry is that the State is just an instrumentality of the ruling class, and thus it will repress speech with an eye to the interests of those with money and property. At the other end of the political spectrum, public choice theory worries that the State will do the bidding of whatever well-funded ‘interest’ group can capture its regulatory and legislative processes. The difference between the two is one of degree, rather than kind: public choice theory typically does a good job (at least in an American context, I am agnostic about elsewhere) explaining legal outcomes against the backdrop of the ideological parameters set by the interests of the ruling classes; Marxist theory typically does a better job explaining why the only ‘interest’ groups that compete meaningfully are ones that do not threaten the basic perquisites of capitalist elites. But if both are right, then looking to the State to be epistemically reliable arbiters of speech, with an eye to maximising social welfare, is a dangerous illusion. Capitalist democracies, given their pathologies as diagnosed by the Left and the Right, cannot be trusted to do the job.
I have considerable sympathy for these worries, especially in the American context with which I am most familiar. But since the problem at issue is not that the expression of non-mundane speech is prima facie valuable and worth protecting, we should focus on the real problems. First, can democratic institutions be reformed to make them more reliable at regulating harmful speech? Second, perhaps capitalist democracies are simply incapable of regulating speech to maximise human well-being, in which case the problem is not one of democratic theory, but of the kind of socio-economic order we should have?
Before turning to these questions, however, let me take up a threshold objection — namely, that all the arguments against freedom for bad speech canvassed so far are really just arguments against democracy itself, so a worry about reforming democracy is already beside the point.
If the real problem with false and pernicious speech in democratic societies is that it leads some benighted individuals to vote for those who will carry forth harmful agendas — think of the strange ‘Tea Party’ phenomenon in the US or, more recently, the election of Donald Trump — then isn’t this really an argument against democracy? Democratic procedures, however, also have dignitary values that are not touched by the case against free speech so far. Indeed, the thrust of the arguments here are largely concerned with making democratic procedures more conducive to well-being, not supplanting them. Democratic procedures, at least when functioning, register (however imperfectly) what people want or desire, and doing so has both a kind of dignitary value as well as eudaemonic value for persons.
The case against free speech so far would only be a case against democracy if we believed, with Plato, that there was such a thing as expertise about each individual’s good, and that there were some political process apart from democracy well-suited to realising that good. (I am assuming, already with a nod to liberal democracy, that realising the individual’s good is the relevant desideratum. That assumption is generally not contested in bourgeois political theory, and I adhere to local academic convention here.) I am sceptical about both claims.
I assume — though I do not argue for it here — that what is good for a person depends, at some level, on the person’s wants and desires, even if heavily laundered by information. So one reason to regulate speech in a democratic society is to effect the laundering of wants and desires necessary for yielding meaningful information for the individual about his or her good. But I also assume — though
I do not argue for it here — that not everything that is bad for a person requires laundering individual wants and desires: we can be confident already that we know a lot about what is bad for a person. One might say I assume a kind of Platonism about ‘the bad’ is true, even though Platonism about the good is false. And if that were right, then that would put constraints on free speech, even allowing for the independent value of democracy with respect to the good. Let me explain.
Even if what constitutes a good life is hostage to individual (or socially determined) variation in wants, ‘Platonism’ about the bad for individuals is still quite plausible. I use the label ‘Platonism’ ironically, since the claim is not that what is bad for persons is a supra-sensible truth, only that certain things are bad for humans ‘objectively’, in the sense that we do not need evidence of laundered wants and desires to know that they are bad. More accurately, it might be called ‘Objective Humeanism about the bad’. On a ‘Humean’ view, goodness and badness are relative to humans, to creatures like us. It is a fact about the nature of creatures like us that our satisfactions are rather various. But it is also a fact, or so I will suggest, that there are natural limits on our well-being; natural limits that, as it were, set the threshold for bad speech. These limits should not seem very controversial: being killed, mutilated, raped, enslaved, starved, immiserated, humiliated, degraded, enervated, stupefied, and so on are not good things for creatures like us. Sometimes it might be valuable for others if ‘Humean Bads’ (as I shall call them) are inflicted, but it would not be crazy to think that Humean Bads are, prima facie, bad for those subject to them. (The opposite of Humean Bads does not generate Humean Goods — or not many of them — since goodness is more demanding than just the absence of the bad.) The case against free speech assumes that there are such things as Humean Bads and it also assumes that it is a benchmark for democracy that it is less likely to produce Humean Bads, which is probably true. (That is an inductive inference over a rather limited bit of evidence — roughly the last one hundred years. If the inference is wrong, then democracy may go by the boards too.) Bad speech leads to Humean Bads, we might say, and thus, even if there are fewer reasons to regulate speech for the sake of producing a single good, there may be a lot of reason to regulate speech to prevent the realisation of Humean Bads. Of course, there may also be reasons to regulate speech for the sake of the good, conceived pluralistically, since no one is wellserved by living in a bubble of misinformation about the means to realising their goods. But that point does not depend on whether or not there are Humean Bads in my sense.
So if the objection is not against democracy per se, and if the real problem, as I argued earlier, is not that the expression of non-mundane speech is prima facie valuable and worth protecting, then we need to return to the core issues given the challenge posed by public choice and Marxist critiques of democracy. First, can democratic procedures be reformed to make them more reliable at regulating harmful speech? Second, perhaps capitalist democracies are simply incapable of regulating speech to maximise human well-being, in which case the problem is not one of democratic theory, but of the kind of socio-economic order we should have? Those are huge questions, but let me conclude with some suggestive remarks, even if they fall well short of resolving the issues.
On the first issue: recently, the philosopher Paul Woodruff and the political theorist John McCormick have drawn attention to some of the novel democratic devices of the ancient Greeks. To protect against the malevolent influence of existing wealth and entrenched interests, the Greeks employed some radical procedures, most notably, lotteries for governing offices and popular ‘tribunals’ — the latter of which had the power to review, and sometimes reject, legislative and executive decisions. The lotteries also typically excluded from the eligible pool wealthy citizens, to insure representation from the lower classes. The beauty of this kind of lottery is that it can nullify some of the pernicious influence of both ruling elites and so-called ‘interest’ groups, since those invested with power to, for example, regulate speech or review existing speech regulations, are unknown before they win the lottery. The lottery device and popular tribunals that review legislative and executive action do not necessarily negate the pernicious effects of ideological indoctrination, however. It would also be necessary to insure that once the lottery ‘winners’ are known, they are also insulated from outside influence and post-service rewards. Some of the worries associated with lay juries noted at the start could also recur in the context of popular tribunals. So an effective lottery and popular tribunal system would face many obstacles, quite apart from being enacted! But should we not ask whether radical democratic procedures like these might unsettle the pathologies of capitalist democracies diagnosed by the Left and the Right? Perhaps so. And perhaps conjoined with the work of an independent, and non-politicised, judiciary — something that many free societies already have — democratic societies might do better at dealing with the pernicious aspects of speech.
The second question raises even harder questions. What must a society be like such that free speech is actually welfare-enhancing? Mill’s strictures of education and maturity are hardly much help, especially given his own complacency that those strictures had relatively little bearing even on his own benighted readers in the 1860s! The preceding discussion in this article has hinted at some of the relevant factors, for example: 
(1) the epistemic quality of the major media, since they bear primary responsibility for beliefs that are widely accepted in the society at large;
(2) the effect that common cognitive biases have on decision-making;
(3) the effect that state sector or private propaganda (‘consumerism’ in the latter case) have on human conceptions of their basic and non-basic interests; and
(4) the extent to which citizens are able to evaluate the epistemic authority of different sources of information and analysis, since almost everything we believe, about climate change or traffic conditions on our commute to work, depends on relations of epistemic authority.
One of the main problems in the US right now is a complete breakdown in the ability to assess epistemic authority: so, for example, that the National Academy of Sciences endorses a view is not thought to be relevant by some substantial portion of the population.
How do we address these factors, such that they are all finely tuned to making speech welfare-enhancing? The poor epistemic quality of the media seems partly a function of private ownership (for example, the Murdoch empire) and partly a function of market incentives (for example, ‘pandering’ to the lowest common denominator). Although some public media (for example, the Public Broadcasting Corporation in the US, the BBC in England) have rather good track records, others, shall we say, do not (for example, Pravda circa 1975). Private sector propaganda seems endemic to capitalism, unless we are to enact dramatic restrictions on commercial speech, both for epistemic and ethical reasons (perhaps we should?). The inability of millions of people to assess epistemic authority sensibly seems mostly an artefact of private sector propaganda. Could these problems be rectified within the framework of the capitalist democracies? I do not know. But let me conclude by suggesting that these are the issues that we heirs of the Enlightenment should be examining, and that it is long past time to abandon the implausible idea that ‘free speech’ simpliciter is an obvious force for further enlightenment and human well-being.
[∗] Karl N Llewellyn Professor of Jurisprudence and Director of the Center for Law, Philosophy & Human Values, University of Chicago, Chicago, United States. A very early version of this article was originally presented as the annual Julius Stone Address in Jurisprudence at the University of Sydney, Sydney, Australia on 14 August 2013. My sincere thanks to Leslie Green, Martha Nussbaum, and Frederick Schauer for very helpful comments on the version presented there, comments provided on short notice no less; they are absolved from all heresies and mistakes that remain, despite their best efforts. At Sydney, for very helpful comments and questions at the public lecture and/or the Sydney Law School seminar afterwards, I am grateful to Rosalind Dixon, Tom Dougherty, Helen Irving, Martin Krygier, Wojciech Sadurski, Michael Sevel, Adrienne Stone, and Kevin Walton. I am also grateful to Gabriel Broughton, University of Chicago Law School Class of 2013, for excellent research assistance and excellent substantive comments on the article, and to Sarah Conly for helpful feedback on a later version. Written feedback from David Strauss on an even later draft helped me reconceive the project in certain important ways. The current version benefited from a Work-in-Progress luncheon at the University of Chicago Law School (and especially from comments by William Hubbard, Richard McAdams, Eric Posner, and Geoffrey Stone); from a workshop at the Human Rights Program at the University of Palermo (Palermo, Italy) especially the comments of Pierre Brunet and Bruno Celano; from a workshop at the University of Miami Law School (Miami, United States), especially the comments of Michael Froomkin; and from a useful discussion with Robert Post. I also received helpful written feedback from the philosopher Brian O’Connor on that version. My thinking about all these issues was greatly enriched by the opportunity to teach a workshop on philosophy of free speech with the philosopher Robert Mark Simpson at Chicago in the Winter and Spring quarters of 2015. I also thank the Alumni Faculty Fund of the University of Chicago Law School for financial support and Noorjit Sidhu, University of Chicago Law School Class of 2017, for research assistance with the final version.
 In the useful categorisation suggested by Cohen, I will be defending a kind of ‘minimalist’ view (which focuses ‘on the magnitude of the evil those protections [for speech] prevent rather than the magnitude of the good they protect’) and rejecting the ‘[m]aximalist’ view that ‘the transcendent value of expression guarantees that it trumps’ all cost: Joshua Cohen, ‘Freedom of Expression’ (1993) 22(3) Philosophy & Public Affairs 207, 210, 218. Cohen’s own view, which has minimalist and maximalist elements, may not be structurally that dissimilar from where I end up, though I have doubts about how he conceives the ‘fundamental interests — expressive, deliberative, and informational’ at stake, and doubts about how he assesses the costs of bad speech and the remedies for it: at 211. I shall return to those issues below. Cohen’s position is also affected by his allegiance to the later Rawls’s ‘political liberalism’, but I will bracket my doubts about that for purposes of the discussion here. On this issue, Enoch is particularly good: David Enoch, ‘Against Public Reason’ in David Sobel, Peter Vallentyne and Steven Wall (eds), Oxford Studies in Political Philosophy (Oxford University Press, 2015) vol 1, 112.
 See Frederick Schauer, ‘The Boundaries of the First Amendment: A Preliminary Exploration of Constitutional Salience’ (2004) 117(6) Harvard Law Review 1765.
 For an earlier foray into this general topic, on which I occasionally draw here, see Brian Leiter, ‘Cleaning Cyber-Cesspools: Google and Free Speech’ in Saul Levmore and Martha C Nussbaum (eds), The Offensive Internet: Speech, Privacy and Reputation (Harvard University Press, 2010) 155.
 See generally James Cox and Alvin Goldman, ‘Speech, Truth, and the Free Market for Ideas’ (1996) 2(1) Legal Theory 1, 12–16, for a useful catalogue of such institutions.
 Arizona Free Enterprise Club’s Freedom Club PAC v Bennett, 564 US 721, 750 (2011).
 Grayned v City of Rockford,  USSC 158; 408 US 104, 115 (1972).
 See Adrienne Stone and Simon Evans, ‘Australia: Freedom of Speech and Insult in the High Court of Australia’ (2006) 4(4) International Journal of Constitutional Law 677.
 Public speech in the US has considerably more narrow latitude in practice than in many other Western democracies, but this is not a result of government regulation of that speech, but of certain pathologies flowing from corporate control of the major media.
 Stone and Evans, above n 7, 686.
 For discussion and references, see Alvin I Goldman, ‘Epistemic Paternalism: Communication Control in Law and Society’ (1991) 88(3) The Journal of Philosophy 113, 115 cited in Brian Leiter, ‘The Epistemology of Admissibility: Why Even Good Philosophy of Science Would Not Make for Good Philosophy of Evidence’  (4) Brigham Young University Law Review 803, 814.
 For different formulations of paternalism, see Gerald Dworkin, ‘Defining Paternalism’ in Christian Coons and Michael Weber (eds), Paternalism: Theory and Practice (Cambridge University Press, 2013) 25, 28–31. For my purposes here, I am relying on something like his ‘Definition A’.
 The interest of jurors in discovering the truth is an interest the system imputes to the jurors, to be sure, though some evidence suggests they take it seriously. See, eg, John Gastil et al, The Jury and Democracy: How Jury Deliberation Promotes Civic Engagement and Political Participation (Oxford University Press, 2010).
 Federal Rules of Evidence r 403 (2016 ed, as amended to 1 December 2015) <https://www.rulesofevidence.org/article-iv/rule-403/>.
 For doubts about the probative value of these, see Olin Guy Wellborn III, ‘Demeanor’ (1991) 76(5) Cornell Law Review 1075; Jeremy Blumenthal, ‘A Wipe of the Hands, A Lick of the Lips: The Validity of Demeanor Evidence in Assessing Witness Credibility’ (1993) 72(4) Nebraska Law Review 1157; Frederick Schauer, ‘Can Bad Science Be Good Evidence? Neuroscience, Lie Detection and Beyond’ (2010) 95(6) Cornell Law Review 1191.
 Frye v United States 293 F 1013 (DC Cir, 1923) (‘Frye’).
 293 F 1013 (DC Cir, 1923).
 But see Frederick Schauer, ‘The Decline of “The Record”: A Comment on Posner’ (2013) 51(1) Duquesne Law Review 51, 53–4 (2013).
  USSC 99; 509 US 579 (1993) (‘Daubert’).
 Ibid 593.
 Ibid 600.
  USSC 99; 509 US 579, 593 (1993).
  USSC 99; 509 US 579 (1993).
 See the discussion in this article below at 417–19.
 The exception has been advocates of ‘free proof’, but their views have not prevailed in any modern legal system. See Mirjan Damaska, ‘Free Proof and its Detractors’ (1995) 43(3) The American Journal of Comparative Law 343.
 Post, in his nuanced discussion of the tension between disciplinary expertise in the academy (which depends precisely on the kind of discrimination between ideas based on their content) and US free speech doctrine (which generally forbids content-based discrimination), locates a resolution in the ideal of democratic legitimation that he sees as the core value of the United States Constitution amend I (‘First Amendment’): Robert C Post, Democracy, Expertise, and Academic Freedom: A First Amendment Jurisprudence for the Modern State (Yale University Press, 2012) 13–14. And as he notes later, in considering the ways in which content-based discrimination is central to education, ‘[t]he classroom is not a location in which the value of democratic legitimation is at stake’: at 70. The courtroom has somewhat more claim to be such a location, though not generally in Post’s primary sense, which pertains to how public opinion is formed. But Post’s analysis is explicitly aimed at reforming current doctrine; democratic legitimacy is not presently the animating value in free speech jurisprudence. I am agnostic on whether it should be, but if something like Post’s view prevailed, this would strengthen a number of the criticisms of free speech I develop in this article. But see the discussion of democracy below at 417–23, 436–8.
 I thus disagree with Stone, who suggests that ‘there is no similar [temporal] constraint’ in the context of public debate: Geoffrey Stone, ‘The Rules of Evidence and the Rules of Public Debate’  University of Chicago Legal Forum 127, 129. I am inclined to agree with him regarding his particular example — publicising marital infidelity by a candidate for office — but there are other cases where temporal considerations loom very large in public debate given what is at stake.
 Ibid 142–3. Stone acknowledges this point, but underestimates its import by framing the question too narrowly: eg, if character evidence is inadmissible in the courtroom to prove propensity to act in accordance with character, should it be inadmissible in the political process? Stone argues, plausibly, that the cases are relevantly different since, for example, the propensity argument in a criminal trial is made to show that the defendant committed the crime in question, whereas in an election, evidence of bad character is not offered to establish any particular misconduct by the candidate. But the relevant analogical question is not about particular kinds of inadmissible evidence, but about the general principle justifying such admissibility decisions: namely, taking account of the epistemic infirmities of jurors or voters in deciding what information they may hear.
 What I am calling ‘eudaemonic’ values emphasise the ways in which speech contributes to well-being, for example, by enabling self-fulfilment or in virtue of the inherent value of self-expression — see, eg, C Edwin Baker, Human Liberty and Freedom of Speech (Oxford University Press, 1989); Martin H Redish, ‘The Value of Free Speech’ (1982) 130(3) University of Pennsylvania Law Review 591. Autonomy-based views sometimes have a consequentialist structure, where autonomy is treated as a kind of ‘good’ that free speech can promote — see the classic discussion in Susan J Brison, ‘The Autonomy Defense of Free Speech’ (1998) 108(2) Ethics 312 — but some autonomy-based views treat autonomy as a categorical constraint on regulation of speech. This is true, in differing ways, of Thomas Scanlon, ‘A Theory of Freedom of Expression’ (1972) 1(2) Philosophy and Public Affairs 204; David A Strauss, ‘Persuasion, Autonomy, and Freedom of Expression’ (1991) 91(2) Columbia Law Review 334; and, most recently, Seana Valentine Shiffrin, Speech Matters: On Lying, Morality, and the Law (Princeton: Princeton University Press, 2014). Cohen’s ‘interests’ theory, above n 1, is a consequentialist version of the autonomy view, since it incorporates a willingness to measure costs of speech. Strauss too incorporates a calculative element. I argue below in favour of collapsing autonomy views into eudaemonic views.
 Justice Kennedy in Citizens United v Federal Election Commission, 558 US 310, 339 (2010) notes that, ‘[s]peech is an essential mechanism of democracy, for it is the means to hold officials accountable to the people ... The right of citizens to inquire, to hear, to speak, and to use information to reach consensus is a precondition to enlightened self-government and a necessary means to protect it.’ (But reaching consensus is not valuable per se, without regard to the epistemic merits of the consensus in question). See also Vincent Blasi, ‘The Checking Value in First Amendment Theory’  (3) American Bar Foundation Research Journal 521.
 I have learned much from the comprehensive treatment in Brison, above n 29, 332, which distinguishes six different kinds of ‘autonomy’ that have been invoked. As she aptly notes, some of these accounts are basically consequentialist in structure and so admit balancing of ‘autonomy’ against other considerations.
 I single out the Far Right end of the political spectrum in the US for the obvious reasons — its existence depends disproportionately on ignorance and falsehoods, as is obvious to everyone outside the parochial boundaries of US political discussions. The structure of the worry, however, is not proprietary to the Right, especially in other countries with greater diversity of political debate. In the US, for example, the ‘liberal’ end of the political spectrum has been taken over by fictions about the harms of genetically-modified foods. But that is trivial by comparison to the fictions that grip the ‘Right’.
 The suppressed premise can be weakened to say: a polity governed by a free speech regime is less likely to lead to such abominations than the alternatives. That may be true, though the evidential question is complex. In any case, this weakening effectively smuggles in the question I think we need to treat as separate: namely, whether competent regulation of speech is possible. I return to that issue, in Part IVB below.
 The phenomenon is certainly not limited to the US. The British public, for example, is also massively misinformed about a range of issues: see Jonathan Paige, ‘British Public Wrong about Nearly Everything, Survey Shows’, The Independent (online), 9 July 2013 <http://www.independent.co.uk/news/uk/home-news/british-public-wrong-about-nearly-everything-survey-shows-8697821.html> . Britain, of course, is also hostage to a Murdoch empire of misinformation and propaganda masquerading as news.
 Working Group I of the Intergovernmental Panel on Climate Change (‘IPCC’), Summary for Policymakers (2007) IPCC, 3 <https://www.ipcc.ch/pdf/assessment-report/ar4/wg1/ar4-wg1-spm.pdf> (emphasis in original).
 Ibid 10 (emphasis in original).
 This consensus is reflected in the scientific literature. Overwhelming support among scientists (over 98%) was found in both a 2005 analysis of 928 abstracts published in refereed scientific journals between 1993 and 2003 (Naomi Oreskes, ‘The Scientific Consensus on Climate Change’ (2004) 306(5702) Science 1686 ) and then a 2013 analysis of 11 944 peer-reviewed articles on climate change published between 1991 and 2011 (John Cook et al, ‘Quantifying the Consensus on Anthropogenic Global Warming in the Scientific Literature’ (2013) 8(2) Environmental Research Letters 1, 3). The same conclusion has been reached by the National Academy of Sciences, the American Meteorological Society, the American Geophysical Union, and the American Association for the Advancement of Science: see Oreskes at 1686. Likewise, surveys of climate scientists have found near-unanimous agreement (97–98%) among publishing climate experts: see William R L Anderegg et al, ‘Expert Credibility in Climate Change’(2010) 107(27) Proceedings of the National Academy of Sciences of the United States of America 12107; Peter T Doran and Maggie Kendall Zimmerman, ‘Examining the Scientific Consensus on Climate Change’ (2009) 90(3) Eos, Transactions American Geophysical Union 22.
 See Pew Research Center, A Deeper Partisan Divide Over Global Warming (8 May 2008) People Press, 1 <http://www.people-press.org/files/legacy-pdf/417.pdf> .
 What the precise causal contribution of false belief is to US inaction is a harder question, but it seems unlikely to help!
 Steven Kull, Clay Ramsay and Evan Lewis, ‘Misperceptions, the Media, and the Iraq War’ (2003/2004) 118(4) Political Science Quarterly 569, 572.
 Excellent accounts are given in Richard M Pious, Why Presidents Fail: White House Decision Making from Eisenhower to Bush II (Rowman & Littlefield, 2008) 222–3; Chaim Kaufmann, ‘Threat Inflation and the Failure of the Marketplace of Ideas: The Selling of the Iraq War’ (2004), 29(1) International Security 5, 16–19.
 Kull, Ramsay and Lewis above n 40, 572.
 Ibid 589.
 Ibid 589–90. Even the consumers of public TV and radio news still had false beliefs, it bears noting. There are, of course, many other topics on which the American public has been ill-served epistemically by the robust protections for free speech, including false speech. There is no scientific controversy, for example, about the theory of evolution by natural selection — as the US National Academy of Sciences correctly put it, ‘[n]o other biological concept has been more extensively tested and more thoroughly corroborated than the evolutionary history of organisms’: National Academy of Sciences, Committee on Science and Creationism, Science and Creationism: A View from the National Academy of Sciences (National Academies Press, 1984) 22. Yet, in 2012, 46% of Americans believed God created humans in their familiar form in the last 10,000 years. Frank Newport, In U.S., 46% Hold Creationist View of Human Origins (1 June 2012) Gallup <http://www.gallup.com/poll/155003/Hold-Creationist-View-Human-Origins.aspx> . Only 15% of the population held the correct view that human beings evolved without divine intervention. False beliefs about biology have pernicious effects on science education, but usually do not kill people. False beliefs about Iraq’s involvement with 9/11 or about climate change did and will contribute to killing people. Robust protections for freedom of speech in the US facilitated these catastrophic errors that undermine the putative value of democratic selfgovernment.
 I take no position yet on whether regulation would be worse than a regime of unfettered expression along the dimensions noted — that is just the Problem of the Epistemic Arbiter, to which I return in Part IV below.
 Friedrich Hayek, The Road to Serfdom (University of Chicago Press, 1944).
 It may also provide a reason for non-democratic regimes to favour more free speech, as a way of aggregating information relevant to successful governance. (Thanks to David Strauss for pointing this out.)
 Amartya Sen, Development as Freedom (Oxford University Press, 1999).
 Hayek probably also held the view that, even if subjectivism is false, people are better arbiters of their interests than the alternative arbiters — but that just raises, again the Problem of the Epistemic Arbiter, to which I will return in Part IV below.
 Sarah Conly, Against Autonomy: Justifying Coercive Paternalism (Cambridge University Press, 2013).
 Ibid 21–2. The philosopher J D Trout draws a similar conclusion from a review of the psychological research:
We severely underestimate our health risks, from HIV to heart disease and cancer, and so don’t take adequate precautions (the ‘optimistic bias’). We discount the future value of resources, and so radically undersave for a variety of important and foreseeable prospects, ranging from the costs of college education and health care to retirement (the ‘discounting bias’). These biases of reason and emotion are in no way exotic; they afflict normal people under normal stresses. Their effects are both routine and expensive. Because they are allowed to go uncorrected, people unnecessarily suffer disease and poverty.
J D Trout, ‘Paternalism and Cognitive Bias’ (2005) 24(4) Law and Philosophy 393, 393.
 Conly, above n 50, 7.
 See, eg, Jon Elster, Sour Grapes: Studies in the Subversion of Rationality (Cambridge University Press, 1983).
 Jesse J Prinz, The Emotional Construction of Morals (Oxford University Press, 2007).
 See also the discussion in Brian Leiter, ‘Moralities are a Sign-Language of the Affects’ (2013) 30(1– 2) Social Philosophy and Policy 237.
 I return to the issue of ‘autonomy’ in Part IIIB below.
 Max Horkheimer and Theodor W Adorno, Dialectic of Enlightenment: Philosophical Fragments (Gunzelin Schmid Noerr (ed), Edmund Jephcott trans, Stanford University Press, 2002) [trans of: Gesammelte Schriften: Dialektik der Aufklärung und Schriften 1940–1950 (first published 1944)]; Herbert Marcuse, One-Dimensional Man: Studies in the Ideology of Advanced Industrial Society (Beacon Press, 2nd ed, 1991).
 Everyone knows, I take it, that totalitarian societies, whether capitalist or not, try to do the same thing; the Frankfurt School theorists rightly call attention to the more decentralised version in the capitalist democracies.
 Horkheimer and Adorno, above n 57, 94. Totalitarian countries have produced their own aesthetic monstrosities, but that has no bearing on Horkheimer and Adorno’s correct descriptive observation about cultural products in capitalist societies. The only point at issue here is that subjects’ conceptions of their non-basic interests are heavily moulded by socio-economic forces.
 Ibid 95.
 Ibid 98.
 Ibid 108.
 Nothing remotely comparable obtained prior to the emergence of national and international markets for cultural products.
 To be truer to Hayek, I should note that his view was that the price mechanism was better at communicating information about preferences in a population than any alternative, regardless of how those preferences were determined and regardless of what they were preferences for. Preferences can be expressed through speech, but most fundamentally for Hayek, they are expressed through what people buy (or try to buy). And when elites manufacture false beliefs (eg, about climate change or health care), they will lead people to have preferences (noted by the price mechanism) that can, in fact, run counter to their basic needs. So the Hayekin argument in favour of free speech, based on the idea of widely dispersed information, must come with several caveats related to the other reasons adduced in this article for scepticism about the value of speech. (I am grateful to William Hubbard for guidance on these issues.)
 Herbert Marcuse, ‘Repressive Tolerance’ in Robert Paul Wolff, Barrington Moore Jr and Herbert Marcuse, A Critique of Pure Tolerance (Beacon Press, 1969) 88.
 Here I have been strongly influenced by the illuminating analysis and critique in Brison, above n 29, and by discussion of this article with Brison and Bob Simpson. Brison is mainly concerned to show that ‘autonomy’ defences of free speech do not rule out the regulation of ‘hate speech’. I find her persuasive on that point, but my target is bigger: I want to show that even the nonconsequentialist ‘autonomy’ defences do not provide a good reason to rule out the content-based regulation of speech on other grounds.
 Brison, above n 29, 323.
 Scanlon, above n 29, 216.
 Thomas Nagel, ‘Personal Rights and Public Space’ (1995) 24(2) Philosophy & Public Affairs 83, 96.
 Scanlon, above n 29, 216.
 Nagel, above n 69, 96.
 For a more precise formal statement of the argument, see Peter van Inwagen, ‘An Argument for Incompatibilism’ in Gary Watson (ed), Free Wild (Oxford University Press, 2nd ed, 2003).
 Brison, above n 29, 331 notes the continuing resonance of such views with contemporary ‘autonomy’ defenses of free speech.
 Thomas Nagel, The View from Nowhere (Oxford University Press, 1986) 113.
 Friedrich Nietzsche, Beyond Good and Evil (Rolf-Peter Horstman and Judith Norman (eds), Judith Norman trans, Cambridge University Press, 2002) 17 [trans of Jenseits von Gut und Böse (first published 1886)]. See also Brian Leiter, ‘Nietzsche’s Theory of the Will’ (2007) 7(7) Philosopher’s Imprint 1, also reprinted in Ken Gemes and Simon May (eds), Nietzsche on Freedom and Autonomy (Oxford University Press, 2009) 107.
 Galen Strawson, ‘The Impossibility of Moral Responsibility’ (1994) 75(1/2) Philosophical Studies: An International Journal for Philosophy in the Analytic Tradition 5.
 We might distinguish, for example, between kinds of causal determination and their sources. For the contours of one alternative, see Donald Rutherford, ‘Freedom as a Philosophical Ideal: Nietzsche and his Antecedents’ (2011) 54(5) Inquiry 512, and see the further discussion in the text.
 Like most philosophical disputes, this one cannot be easily resolved, at least if ‘easily’ means securing agreement among all disputants. But disputants here have far too many ulterior motives to make it productive to pursue the topic at length in the context of an article about free speech.
 Immanuel Kant, What is Enlightenment? (first published 1784, Modern Library, 1949).
 Ibid 132.
 So, for example, Strauss, in a quite sober version of an autonomy account, says that regulating speech that might ‘persuade’ someone to a bad end is objectionable because it ‘interfere[s] with a person’s control over her own reasoning processes’: Strauss, above n 29, 354. Strictly speaking, of course, people do not ‘control’ their reasoning processes: what one believes is not a matter of volition, and the best evidence from cognitive science suggests that most ‘thinking’ is unconscious: see, eg, David M Rosenthal, Consciousness and Mind (Oxford University Press, 2005); Robert Mark Simpson, ‘Intellectual Agency and Responsibility for Belief in Free Speech Theory’ (2013) 19(3) Legal Theory 307. What Strauss has in mind, however, is not really control of reasoning, but rather the worry that reasoning might be distorted by nonrational forces. Thus, his ‘persuasion principle’ — that the State should not regulate speech that might persuade people in harmful directions — is limited to speech that involves rational persuasion. Even on Strauss’s account, ‘the persuasion principle can be overridden if the consequences of permitting the speech are sufficiently harmful’: Strauss, above n 29, 360. But what is ‘rational’ persuasion, the putative limit on regulation before we get to rationally persuasive speech that is ‘sufficiently harmful’? Strauss says only that non-rational persuasion involves ‘false information’ and tries to produce ‘an ill-considered reaction’: Strauss, above n 29, 335. This clearly cannot suffice, however, to demarcate kinds of persuasion for a variety of reasons. Only on the assumption that rational persuasion requires that the premises taken to justify a true conclusion must themselves be true does the first restriction follow. But why does rationality mandate that? On an instrumental conception of reason, and even assuming true belief is the desired end, it is perfectly rational to be led to true belief via falsehoods. But on an instrumental conception of rationality, even true belief is an optional outcome: it depends on what our ends really are. If human happiness depends on false belief (as it probably does), then it is instrumentally rational to get people to believe happiness-inducing false claims, and rational to do so by presenting them with false information, if that is necessary to induce the happiness-inducing states. That suggests that the ‘persuasion principle’ requires some substantive conception of rationality. (Notice that commitment to the logical validity of inferences will not elide the problems already noted with a merely instrumental conception of rationality.) Unfortunately, there are no plausible substantive conceptions of rationality, though it is easy to see how Strauss’s constraints make sense if one shares Kantian intuitions. Indeed, his second constraint — namely, speech that induces ‘ill-considered action’ — is transparently parasitic on a substantive conception of reason that remains unspecified. (Action is only ‘ill-considered’ relative to either an instrumental or substantive conception of rational considerations.)
 Even Karl Popper, proponent of the ‘open society’, had serious doubts about how the medium of television would affect social well-being if its content were not carefully regulated: see Karl Popper, The Lessons of This Century: With Two Talks on Freedom and the Democratic State (Routledge, 1997) ch 7 (‘Television Corrupts Mankind. It is Like War’).
 This conclusion is not, I suspect, inconsistent with Cohen’s specification of the three ‘interests’ in free speech: namely, the ‘expressive interest ... in articulating thoughts, attitudes, and feelings on matters of personal or broader human concern’: Cohen, above n 1, 224 (emphasis in original); the ‘deliberative interest’ in figuring out ‘what is best’ or ‘genuinely worthwhile’: at 228; and the interest ‘in securing reliable information about the conditions required for pursuing one’s aims and aspirations’: at 229. Each of these are interests whose satisfactions contribute to eudaemonic well-being, though both the ‘deliberative’ and ‘informational’ interest must clearly be checked by consideration of human irrationality and influences that defeat successful deliberation. Cohen is insufficiently sensitive to this, I fear, because his putative ‘Fact of Reasonable Persuasion’— the claim that, ‘[p]eople have the capacity to change their minds when they hear reasons presented, and sometimes they exercise that capacity ... But for the Fact of Reasonable Persuasion, more speech would be a diversion rather than a remedy’: at 232–3) — is actually not responsive to the dangers associated with speech, which is that most of the time people are not reasonably persuaded of what is true, just, fair, or decent. ‘More speech’ is only a remedy for (some) bad speech in a world in which people usually, not merely ‘sometimes’, exercise the capacity of changing their views in response to reasons. That is not our world.
 One reason to do so is that the traditional media, even in their online forms, tend to regulate speech quite heavily.
 This is rather difficult to measure, but according to one study, roughly 30% of all data transferred across the internet is pornographic: ‘Porn Websites Get More Visitors Each Month than Amazon and Twitter Combined’, The Huffington Post (online), 5 May 2013 <http://www.huffingtonpost.
com/2013/05/03/internet-porn-stats_n_3187682.html>. Note that I am putting pornography into the category of non-mundane speech because of its apparent centrality to the self-formation and self-expression of hundreds of millions of people.
 On those harmed by pornography, see Neil M Malamuth, Tamara Addison and Mary Koss, ‘Pornography and Sexual Aggression: Are There Reliable Effects and Can We Understand Them?’ (2000) Annual Review of Sex Research 11(1) 26. On the harms of invective, see Jeremy Waldron, Harm in Hate Speech (Harvard University Press, 2014). All three kinds of costs that Cohen distinguished are at issue, though I have doubts about the distinction between Cohen’s ‘indirect costs’ of speech in which ‘the injury results from the expression causing (by persuasion, suggestion or providing information) someone to do something harmful’ and speech that ‘constitute[s] a degraded, sickening, embarrassing, humiliating, obstructively moralistic, hypercommercialized, hostile, or demeaning environment’: Cohen, above n 1, 231–2. I also do not think the differences matter: the only question is whether there is a reliable causal connection.
 See Leiter, ‘Cleaning Cyber-Cesspools’, above n 3.
 David G Myers and Helmut Lamm, ‘The Group Polarization Phenomenon’ (1976) 83(4) Psychological Bulletin 602.
 Daniel J Isenberg, ‘Group Polarization: A Critical Review and Meta-Analysis’ (1986) 50(6) Journal of Personality and Social Psychology 1141.
 See Roger Brown, Social Psychology (The Free Press, 1965) 701.
 Dean G Pruitt, ‘Choice Shift in Group Discussion: An Introductory Review’ (1971) 20(3) Journal of Personality and Social Psychology 339, 340.
 Matt D O’Brien and Stephen D Webster, ‘The Construction and Preliminary Validation of the Internet Behaviours and Attitudes Questionnaire (IBAQ)’ (2007) 19(3) Sexual Abuse 237; Ethel Quayle and Max Taylor, ‘Paedophiles, Pornography and the Internet: Assessment Issues’, (2002) 32(7) British Journal of Social Work 863. Regarding misogyny and group polarisation, see Patricia Bou-Franch and Pilar Garcés-Conejos Blitvich, ‘Gender Ideology and Social Identity Processes in Online Language Aggression against Women’ (2014) 2(2) Journal of Language Aggression and Conflict 226; Azy Barak, ‘Sexual Harassment on the Internet’ (2005) 23(1) Social Science Computer Review 77.
 Here the single German word wissenschaftlich is closer to the mark than the English, which demands that the ‘scholarly’ and the ‘scientific’ be distinguished. A Wissenschaft is a disciplined and epistemically reliable method for investigating some subject. Although some humanities subjects have collapsed as Wissenschaften since the 1970s—English and Comparative Literature are the most notorious examples, though they may be recovering—many others remain intact. Of course, a Wissenschaft may produce falsehoods — think of behaviourism in psychology in the 20th century, and neoclassical macroeconomics in the last half of the 20th century — but it still involves methodological strictures that are meant to block extemporaneous and ungrounded opining of the kind one associates with, for example, followers of Leo Strauss (‘Straussian’), postmodernists, members of the Ayn Rand cult, and so on.
 For example, bloggers made a mess of the Bush draft dodging story, missing the forest for the trees. See Leiter, ‘Cleaning Cyber-Cesspools’ above n 3. Michael Froomkin presses on me that I am underestimating the positive aspects of the internet, such as access to entertainment from movies to art to fiction; the ways in which the internet has lowered the costs of communication and increased access; and the ‘wisdom of crowds’ that the internet can embody. I agree with Froomkin on the first point. On the second, I agree that the internet has lowered the cost of communication, but it is far from clear that has been good, since it has lowered the cost (and increased visibility) for ignoramuses, fools, and propagandists of all stripes. On the third, there are occasions where crowds have ‘wisdom’, but the internet, to date, has not been very good at mobilising it — the crowds are as likely to be wrong as right, foolish as wise, ignorant as well-informed. And because of group polarisation effects, the internet is ill-suited to correcting these tendencies.
 John Stuart Mill, On Liberty (Elizabeth Rapaport (ed), Hackett Publishing, 1978). See also Conly, above n 50, 52 ff. for scepticism about Mill’s ideas about the limits on this caveat.
 Even a defender of the American free speech status quo like Geoffrey Stone concedes that when it comes ‘to maintaining an effective, fair, and open political process’, the choice is ‘not only whether we should trust the government to regulate the press, but whether we should trust the press to define our political process’: Stone, above n 27, 147–8.
 Marcuse, above n 65.
 Mill, above n 95, at 9–10.
 See, eg, Michael McConnell’s complaint that my critique of religious accommodations marks me as the ‘Anti-Mill’: Michael W McConnell, ‘Why Protect Religious Freedom?’ (2013) 123(3) The Yale Law Journal 770, 794.
 Mill, above n 95, 104.
 See Frederick Schauer, ‘Social Epistemology, Holocaust Denial, and the Post-Millian Calculus’ in Michael Herz and Peter Molnar (eds), The Content and Context of Hate Speech: Rethinking Regulation and Responses (Cambridge University Press, 2012).
 Mill, above n 95, 10.
 David O Brink, ‘Millian Principles, Freedom of Expression, and Hate Speech’ (2001) 7(2) Legal Theory 119. See also David O Brink, Mill’s Progressive Principles (Oxford University Press, 2013) ch 7.
 Mill, above n 95, 88.
 See Tom Campbell and Wojciech Sadurski (eds), Freedom of Communication (Dartmouth Publishing Company, 1994).
 Mill, above n 95, 104.
 See Frederick Schauer, ‘The Phenomenology of Speech and Harm’ (1993) 103(4) Ethics 635.
 Mill, above n 95.
 See Ronald Dworkin, Taking Rights Seriously (Harvard University Press, 1977).
 See Cohen, above n 1, on ‘direct’ harm; see also Waldron, above n 86.
 Cf Frederick Schauer, ‘Facts and the First Amendment’ (2010) 57(4) UCLA Law Review 897.
 Stone, who has also explored analogies between the rules of evidence and public debate, arrives at a similar bottom line: he worries that ‘there is great danger in authorizing government to involve itself’ in the political process by regulating information given ‘the possible effect of partisanship affecting the process at every level’: Stone, above n 27, 140. Or, as he succinctly put it to me in conversation: ‘You worry about cesspools. I worry about the Stasi’.
 Frederick Schauer, Free Speech: A Philosophical Enquiry (Cambridge University Press, 1982) provides the most sustained argument for this conclusion, though conditional as well on governmental competence. See also Stone, above n 27.
 I should admit that I view ‘rational’ as basically a term of commendation, and thus without cognitive content (except in stipulative, technical contexts, for example, modus ponens supports a rational inference). See Allan Gibbard, Wise Choices, Apt Feelings: A Theory of Normative Judgement (Harvard University Press, 1990). Instrumental ‘rationality’ dominates our thinking about what is ‘rational’ for reasons Hume would have well-appreciated: it is obviously the default practical norm for creatures like us.
 Sen, above n 48.
 On the German and Canadian approaches, and their differences with the US, see the useful discussion in Ronald J Krotoszynski Jr, The First Amendment in Cross-Cultural Perspective: A Comparative Legal Analysis of the Freedom of Speech (New York University Press, 2006) chs 3–4.
 Consider by way of example: no ‘talking head’ ever appears in the national media in the US who suggests that the way to secure medical and retirement benefits for the elderly and infirm well into the future is to confiscate the fortunes of the Koch Brothers and the Walton children and similar beneficiaries of family good fortune. Perhaps that is a bad idea, but it is not even discussed or broached, because of the ideological parameters in which public debate takes place.
 It goes without saying, I would hope, that non-democratic societies cannot be trusted either!
 The same is, of course, true of speech itself, as should be clear from the preceding discussion. But expression may have another consequence; namely, that it contributes to legitimising the outcomes of political processes, even when they do not go the way of all speakers. The legitimation function of expression is a value independent of the eudaemonic value to the speaker, but not independent of the eudaemonic value to society as a whole: and to the extent it is a good thing, it is contingent on the kind of social order that is legitimated. (Thanks to Sarah Conly for raising this issue.)
 Here and elsewhere I have made a nod to Marx’s idea — one shared, in different ways, by materialist anthropology (eg, Marvin Harris), archaeology (eg, Ian Morris) and history (eg, Eric Hobsbawm) — that the economic circumstances in which theorising occurs influences, in ways not always apparent, the ideas that seem ‘obvious’ or ‘acceptable’. Normative moral and political philosophy, dependent as it always is on ‘intuitions’ about the good and the right, is especially vulnerable to this concern. I flag the generic problem here not because I have a suitable corrective for the problem — no one does — but as a warning for readers in very different circumstances, present or future, who may find all this puzzling.
 Non- or pseudo-democratic capitalist societies like China and Russia also do a bad job regulating speech, as do other autocratic and non- or pseudo-capitalist regimes, like many of those on the Arabian peninsula. I assume that is not in serious dispute.
 See Paul Woodruff, First Democracy: The Challenge of an Ancient Idea (Oxford University Press, 2005); John P McCormick, Machiavellian Democracy (Cambridge University Press, 2011).
 I have been greatly helped here by smart comments from Tom Dougherty, though he should not be presumed to agree with the discussion in the text.
 An epistemic authority is a source who tells us what to believe, and we believe it just because the source told us to do so. As Kuhn noted decades ago, almost all knowledge of science is acquired by means of epistemic authorities: Thomas S Kuhn, The Structure of Scientific Revolutions (University of Chicago Press, 3rd ed, 1996).