No-Platforming: Different Kinds of Speech, Different Kinds of Justifications

The caricature of the debate sketched above contains a grain of truth: much of the current debate over no-platforming can be depicted as pitting defenders of reason—heirs of the Enlightenment, as they paint themselves—against those who emphasize other values. In this section, I very briefly review this debate. As well as being brief, the review is idiosyncratic because I use the term no-platforming more broadly than is usual.

No- platforming, in the narrowest sense, refers to the (attempted) refusal of a platform to hate speech, especially the speech of the antidemocratic far right. Originally, this practice was justified on the basis that those who want to destroy democratic institutions do not have a right to access these institutions. Today, in the age of illiberal democracy, this justification has less bite since the far right now often (and with some justification) argues that democratic institutions are compatible with its goals. In any case, the targets of no-platforming campaigns are now broader than they once were. Typically, speakers are now refused a platform on the grounds that their claims constitute hate speech, and hate speech is harmful (O'Keefe Reference O'Keefe2016; Heinze Reference Heinze2018). No-platforming is now often justified on the basis that the groups that the far right attacks (people of color, transwomen, and so on) are among the most vulnerable already and have a right not to be subjected to further attack and potential exclusion.

The platform that is being denied when someone is no-platformed, as the term is standardly used, is usually the stage in a university context. I use no-platforming more broadly, to refer to two other platforms from which people might be barred. First, people may be prevented from expressing certain views online. Comments on articles in online newspapers or on blogposts might not be accepted (premoderated) or deleted if they cross a line into what is considered hate speech, for instance. Second, there may be campaigns that aim at excluding certain people or block certain views from being expressed in the opinion pages of newspapers, in interviews with media organizations, and so on. This, too, is naturally seen as a kind of (attempted) no-platforming.

My use of the term no-platforming is broad not only in the range of platforms I have in mind but also in the range of opinions. In a narrow sense of the term, no-platforming targets hate speech. I have in mind not only—or even mainly—hate speech but unreasonable minority views about matters of great public interest. Beyond hate speech, paradigm cases of unreasonable minority views include the expression of climate change skepticism or of a purported link between vaccines and autism.

Typically, though far from exclusively, proponents of no-platforming in the narrower sense, focused on hate speech, argue that the speech harms some or all of the audience. Those directly targeted by hate speech or speech that questions their right to belong (or even to exist) are usually held to be the primary victims. The harms referenced are of three different kinds: feelings of isolation and depression (O'Keefe Reference O'Keefe2016); harms to the functioning and academic performance of their targets, for example, through the mechanism of stereotype threat (Kukla Reference Kukla2018); and harms to the dignity and equal standing of members of targeted groups (as Waldron Reference Waldron2012 points out in a somewhat different context). The harms are primarily held to arise from the speech itself, but they may arise instead (or also) from contextual facts. Jeremy Fantl (Reference Fantl2018) points out that there is a big difference between the harms that may arise from merely being exposed to certain ideas and those from knowing that they are being expressed by an invited speaker. He suggests that this difference arises from the fact that an invited speaker is accorded a degree of respect, and members of targeted groups therefore feel betrayed by the community that extended the invitation to the speaker.

A common response to all these arguments in favor of no-platforming invokes the value of free speech—speech restricted only by the limits set by something like the Millian harm principle. In particular, it invokes it value for the hearers. Through the clash of ideas, proponents of open speech point out, truth emerges, and sometimes from unexpected places. Bad speech may sometimes contain in it, perhaps buried, an overlooked truth. Moreover, exposure to arguments for false claims allows people to develop critical thinking skills that are essential for a democratic polity. In fact, the full development of our critical thinking skills requires exposure not only to arguments for false claims but also to claims that arouse strong emotions in us. We need to learn to reason not only when we are calm but also when we feel attacked. Defenders of open speech may also invoke John Stuart Mill's claim that we only properly understand our own views when we have heard and rebutted the best arguments against them. For Mill, if we are not exposed to the strongest cases against our views, we will hold correct opinions “as dead dogma, not a living truth” (Mill Reference Mill and Collini2003: 38).

Thus, advocates of open speech cite (mainly) epistemic considerations in its defense. They justify exposure to controversial views, even repugnant views, on the basis that directly or indirectly, the exposure tends to lead their audience to have better justified beliefs. Both the direct and the indirect justification for open speech are well illustrated by a recent polemical book by Greg Lukianoff, president and CEO of the Foundation for Individual Rights in Education, known as FIRE. The direct justification consists in the (putative) fact that in encountering beliefs we are disposed to reject, we put our beliefs to the test, either because the claims we encounter may be partially true, or because they may give us an opportunity to understand our own beliefs better. Thus, we must be willing ‘to acknowledge that even the opinion or person we find the most abhorrent might reveal some portion of the truth of which we are unaware. It is also important to recognize that even if a speaker does happen to be entirely wrong, we might learn more about our own beliefs or about the complex relationships among beliefs by allowing that person to speak’ (Lukianoff Reference Lukianoff2014: n.p.).

The indirect justification consists in the (putative) fact that exposure to bad arguments inculcates habits in the audience that will stand them in good epistemic stead in the future. That is, such exposure may not lead directly to better beliefs but will help to develop capacities that lead to better beliefs. A respect for dissent inculcates ‘healthy intellectual habits, such as giving the other side a fair hearing, reserving judgment, tolerating opinions that offend or anger us, and recognizing that even people whose points of view we find repugnant might be (at least partially) right’ (Lukianoff Reference Lukianoff2014: n.p.).

These epistemic considerations have a particular power in the context of higher education, given that its aims are very significantly (though not, of course, entirely) epistemic: the inculcation of true belief and of the capacities to acquire it. Lukianoff and Jonathan Haidt (Reference Lukianoff and Haidt2015) end their widely discussed broadside against no-platforming with the words of Thomas Jefferson upon the founding of the University of Virginia: ‘This institution will be based on the illimitable freedom of the human mind. For here we are not afraid to follow truth wherever it may lead, nor to tolerate any error so long as reason is left free to combat it’.

We might therefore understand the debate over no-platforming as pitting two different, arguably even incommensurable, sets of values against one another. On the one hand, defenders of open speech uphold the importance of a set of epistemic values, which they associate with the Enlightenment, while their opponents point to the likelihood of certain kinds of harms.

While this seems to me a reasonably accurate description of the current debate, there have been recent stirrings of epistemically based defenses of no-platforming. One response takes issue with the way in which opponents of no-platforming characterize the aims of the university: Robert Mark Simpson and Amia Srinivasan (Reference Simpson, Srinivasan and Lackey2018) maintain that the distinctive kind of institution a university is entails exclusion of certain kinds of speech incompatible with disciplinary norms. Eric Heinze (Reference Heinze2018) and David Estlund (Reference Estlund and Lackey2018) outline a possible defense of no-platforming that justifies some interference with the free expression of ideas on the grounds that it can counter the unfair advantages certain instances of speech have, stemming from the power or economic clout of speakers. (Heinze himself rejects this view on the grounds that it is a mistake to think that restrictions on democratic norms can ever help to secure democratic goals. While Estlund does not have Heinze in mind specifically, much of his essay can be read as a reply to this argument. Estlund argues it is a fallacy to think that because certain norms conduce to a particular goal in ideal conditions, pursuing that goal in nonideal conditions requires respect for those same norms.)

Another kind of epistemic defense of no-platforming takes issue with what might be seen as open-speech advocates’ naïve understanding of reasoning. Defenders of no-platforming might invoke a range of evidence from psychology that suggests that speech causes beliefs (or, at any rate, doxastic states) in ways that bypass reasoning capacities, and therefore independently of how well justified the speech may be. Speech may play a role in the inculcation of implicit biases. It may also lead to the development of false explicit beliefs, perhaps through the sleeper effect, whereby claims that are initially regarded as false come to be accepted, due to the initial claim being more memorable than the rebuttal (Pratkanis et al. Reference Pratkanis, Greenwald, Leippe and Baumgardner1988; Lewandowsky et al. Reference Lewandowsky, Ecker, Seifert, Schwarz and Cook2012); through fluency effects, whereby familiarity with a claim makes it seem more plausible (Begg, Anas, and Farinacci Reference Begg, Anas and Farinacci1992; Weaver et al. Reference Weaver, Garcia, Schwarz and Miller2007); or through so-called Spinozist accounts of belief formation, according to which rejection of claims we are exposed to is effortful and often—perhaps even always—unsuccessful (Huebner Reference Huebner2009; Mandelbaum Reference Mandelbaum2014). These kinds of evidence have been cited in defense of avoiding, or even restricting, speech due to its effects on beliefs and behavior (Moles Reference Moles2007; Levy Reference Levy2012; Mandelbaum and Quilty-Dunn Reference Mandelbaum and Quilty-Dunn2015). While some psychological evidence may in fact have an import that tends in the opposite direction (Ceci and Williams Reference Ceci and Williams2018), this evidence might be used to mount a plausible defense of no-platforming.

If proponents of no-platforming appeal to these epistemic considerations, however, they commit themselves to the acceptability of epistemic paternalism: the management of an agent's epistemic environment without (or regardless of) her consent, in order to promote her epistemic welfare. While epistemic paternalism may be justifiable (Ahlstrom-Vij Reference Ahlstrom-Vij2013), it is controversial. In appealing to it, proponents of no-platforming are brought uncomfortably close to agreeing that they infantilize the audience. To treat others paternalistically is, after all, to treat them as lacking (in certain respects) the capacity rationally to make their minds up for themselves. Perhaps such paternalism is justifiable because the Enlightenment conception of reason is unrealistic. Nevertheless, intellectual autonomy is not a good we can jettison without qualms. If we are able to defend no-platforming without flirting with the infantilization of the audience or disrespect for its intelligence, we ought to do so, and if we can avoid appealing to controversial psychological research, so much the better.

That is my aim here. I argue that in many instances, no-platforming may be justifiable for precisely the reason that refraining from lying to people is justifiable: respect for people's intellectual capacities and autonomous reasoning requires us to avoid deceiving them. Bad speech presents its audience with genuine evidence, evidence that they ought rationally to take into account in forming beliefs and that retains evidential value for them even if and even after it is rebutted or balanced.

Disagreement as Higher-Order Evidence

First-order evidence is evidence that directly bears on the truth of a proposition. Higher-order evidence (at least as this phrase is commonly used in epistemology) refers to evidence bearing on the reliability of the relationship between the first-order evidence and the proposition it is supposed to support or undermine, or the reliability of the agent who is assessing that relationship (Christensen Reference Christensen2010; Lasonen-Aarnio Reference Lasonen-Aarnio2014). Higher-order evidence is a familiar feature of the epistemic landscape. We regularly accuse one another of partisan biases, or worry that our partiality to loved ones may blind us to their faults. When we consider whether we should reduce our confidence in our beliefs (“perhaps Uncle Charlie really did take that money?”), we take higher-order evidence into account.

The higher-order evidence on which epistemologists have lavished most attention is the evidence stemming from disagreement. The discovery that you disagree with someone often provides higher-order evidence, in light of which you are rationally constrained to adjust your beliefs. Consider this (now hackneyed) kind of case, Restaurant Check:

Given that certain conditions are satisfied (in particular, that the other person is an epistemic peer), discovering that I disagree with someone else in a case like this provides higher-order evidence that should lead to a reduction in my confidence in my belief. I may have good evidence that the other agent is an epistemic peer. Perhaps the bill is always left to us on occasions like these, and we always perform the arithmetic in our heads, independently of one another. We almost always agree on the result (the math is not difficult), and on those occasions when we have disagreed, she has been right about half the time. There is nothing about this occasion that should lead me to think that either of us has a special edge tonight (neither of us has had any alcohol, neither is particularly tired or stressed, and so on). In that case, it seems I should reduce my confidence in the figure I have come up with. Her disagreement provides me with higher-order evidence: The most likely explanation of our disagreement is that (at least) one of us made a mistake in our arithmetic, and since I have no more reason to think that she has made a mistake than to think that I have, it seems that I should reduce my confidence (she should do likewise, for the same reason).

While most philosophers accept that in cases like Restaurant Check, we should reduce our confidence in the proposition over which we disagree, other cases show that this is not always true. Consider this example:

Perhaps Jennifer has extensive evidence that Harry is her epistemic peer. Nevertheless, his disagreement does not give her any reason to reduce her confidence that 2 + 2 = 4. That is because this sum is too easy: it is not credible that the best explanation of the dispute is that one or other side has made a reasoning error. It is, however, a mistake to conclude that cases like this show us that disagreement is not epistemically significant (as Lackey recognizes). While we should not reduce our confidence in the target proposition, the disagreement does provide evidence in the light of which we should (greatly) reduce our confidence in the epistemic peerhood of the other agent. The best explanation of the disagreement is that Harry is either insincere or delusional. In a case like this, Jennifer should hold fast with regard to her arithmetical beliefs because the symmetry between her and Harry is broken: given how easy the sum is, it is not credible that he made a mistake of a sort the she could easily have made. She can therefore rely on her ‘personal information’ (Lackey Reference Lackey, Haddock, Millar and Pritchard2010: 318 that she is not insincere or delusional to dismiss Harry from peerhood.

We might sum up the lessons from the literature on the epistemic significance of disagreement as follows: when two peers reach different conclusions on some moderately difficult (for these agents, in this context) question, this disagreement provides each of them with higher-order evidence that one or other has made a mistake in reasoning, unless there is some reason to break the symmetry between the disputants. When peers reach different conclusions on some very easy question like 2 + 2, or whether the world is more than five years old, the disagreement provides evidence that the one who answers it incorrectly is not (on this occasion or with regard to this proposition) sincere or competent. Conversely, when peers disagree about a very hard question, the disagreement is expectable, given the difficulty of the question, and provides little higher-order evidence at all (suppose my friend and I both try to do the following sum in our heads, under time constraints: 899 x 79 ÷ 4.31, and we come up with different responses. Given that both of us are only ordinarily numerate, our disagreement should not lower our confidence in our answers, in large part because we should have little confidence in the answer in the first place. If—almost miraculously—we agreed on a prima facie credible answer, on the other hand, our confidence should increase significantly). Importantly—for the discussion to follow—it is not only the disputants for whom disagreement and agreement is higher-order evidence. Third parties witnessing the dispute, and who might otherwise have acquired knowledge via testimony from the participants, also have higher-order evidence in the light of which they should adjust their beliefs.

The literature on the epistemic significance of disagreement often presupposes a highly idealized and very demanding conception of epistemic peerhood. My epistemic peer, as standardly understood in this literature, has the same evidence that I have and the same cognitive abilities. It is likely that these highly idealized conditions are never satisfied in the actual world (Matheson Reference Matheson2015). But as Jennifer Lackey (Reference Lackey, Haddock, Millar and Pritchard2010) points out, setting the bar for peerhood so high comes at the cost of ensuring that the debate loses contact with the real-world cases that motivate it in the first place. In fact, actual disagreement puts pressure on us to conciliate; that is, to moderate our own views by moving them closer to those expressed by the dissenting speaker.

One reason why real-world disagreement puts pressure on us to conciliate is that disagreement in the public sphere is often between representative individuals: representative of sufficiently large number of sufficiently expert people. When two economists disagree in public about whether interest rates should rise (say), each typically represents a school of thought; to that extent, investigation of their competence and access to evidence is somewhat beside the point. Perhaps the first really is better trained or more competent than the second, such that the second is not her peer. But these facts do relatively little to reduce the significance of their public disagreement, inasmuch as each is representative of a broader trend within economic thought. Perhaps the second is an ill-chosen representative, but he is representative (in the relevant fashion) nevertheless. Typically, his claims are backed up by the work of a large number of highly competent individuals, who formulate their theories in the light of a wide range of evidence (as, of course, are hers). For ongoing debates like these, evidential equality can be assumed: had the other side access to evidence their opponents lack, they would have pointed it out by now. Since disputes between representative individuals in the public sphere typically reflect ongoing disputes between large numbers of highly competent and well-informed individuals, we can set aside worries about cognitive and evidential equality.

Disputants are representative individuals, in the way required for their disagreement to be epistemically significant, if each represents a sufficiently large group of sufficiently well-qualified individuals. The numbers count and expertise counts, each independently. My confidence in the figure I have reached when trying to calculate my share of the restaurant bill should decrease further when I discover that two of my epistemic peers disagree with me, rather than just one: given that they agree with one another and have reached their answer independently, the best explanation of my dissent is that I have made a mistake (conversely, if my peers agree with me, my confidence should increase). But numbers count only when the opinions are sufficiently informed and arrived at independently. An epidemiologist should not be shaken in her opinions about vaccines by the knowledge that thousands of people on Facebook disagree with her.

The epistemic significance of disagreement (and agreement) also varies as a function of the expertise of the speaker and the hearer (holding numbers constant). Eni may rationally dismiss Mark's dissent from her prediction concerning tonight's game, because she is a football expert and he is not; Mark, in contrast, ought to reduce his confidence very significantly, perhaps deferring entirely. Even large numbers of ill-informed dissenters should carry little weight with a genuine expert. In cases like this, the symmetry between the disputants is broken because one side has significantly more expertise than does the other. Disagreement and agreement both provide substantive higher-order evidence with regard to the subject of the dispute when the best explanation is some kind of mistake and not a lack of peerhood between the disputants.

Bad Speech as Higher-Order Evidence

Speech of the sort that gives rise to controversy over platforming is dissenting speech. Often, it is dissent from the mainstream view (as in the cases of climate-science skeptics and anti-vaxxers); sometimes is dissent from more locally hegemonic or influential views (as in the case of so-called trans-exclusionary radical feminists). Such dissent offers us higher-order evidence. The weight that should rationally be given to it by a nonexpert audience is a function of the speaker's apparent expertise, and the extent to which she is reasonably taken to be representative of other experts, with those properties independently contributing to the weight nonexperts should give to the dissent.

Those given a platform are reasonably assumed to be both expert and representative. Most controversial speakers can credibly claim relevant expertise. Consider some recent controversies. In the United Kingdom, a petition to deny Germaine Greer a platform at Cardiff University because of her (allegedly) transphobic views attracted more than three thousand signatures (Morris Reference Morris2015). The veteran gay rights activist Peter Tatchell and the feminist Julie Bindel have each been subject to attempts at no-platforming for similar reasons (Faye Reference Faye2016). While there is controversy over what it takes to possess genuine expertise on issues like these, it is obvious that all three have some genuine claim to relevant expertise as long-time activists in the sphere of gender politics. In the United States, targets of no-platforming campaigns include Ayaan Hirsi Ali, for alleged islamophobia (Leef Reference Leef2014) and Charles Murray, for alleged racism (Beinart Reference Beinart2017). Again, these are individuals with claims to expertise: Hirsi Ali speaks on Islam with the credibility that comes with being a former believer, raised in a Muslim country, and Murray is a prominent academic with multiple relevant publications. Closer to home, the prominent bioethicist Peter Singer has been refused a platform on several occasions, due to his views on disability (Blackford Reference Blackford2015). Again, Singer has a very plausible case for qualifying as possessing relevant expertise.

Some controversial speakers, admittedly, do not have strong claims to expertise themselves. They may nevertheless reasonably be taken to be representative, inasmuch as they can credibly claim to be able to present the expertise of others (some science journalists, like the controversial Nicholas Wade, fit this description). They may be taken to deserve a hearing on that basis. The numbers count more heavily for assessing their credibility than for those who possess expertise themselves. But the numbers count for both. Accordingly, controversial speakers typically go to some lengths to represent themselves as supported by significant numbers of sufficiently expert individuals. They may, for instance, claim to be the mouthpiece for the silent majority, or for a large group of experts who have been silenced by a campaign of intimidation. Well-known examples include the so-called Oregon Petition, often cited as evidence that more than thirty thousand scientists dissent from the consensus on climate change (in fact, few of the signatories are scientists of any kind, and less than 0.5 percent have PhDs in a relevant science; Grandia Reference Grandia2017), and the Discovery Institute's ‘A Scientific Dissent from Darwinism’, presented as evidence that hundreds of experts reject the theory of evolution (Scott Reference Scott2009: 254).

As these examples show, claims to representativeness may be gross exaggerations; equally claims to expertise may be false. For the nonexpert audience, it is difficult to assess the extent to which these claims are justifiable; we (we, since everyone is a nonexpert in most domains) are forced to rely on various proxies for expertise to assess credentials. Such proxies include a record of publication in peer-reviewed journals, the possession of a PhD from a reputable institution, the receipt of prizes, and so on (see Levy Reference Levy2018 for discussion). Publication in the mainstream media and invitations to speak at universities and similarly respectable institutions figure among the proxies for expertise and representativeness we utilize. That is, these institutions play the explicit or de facto role of certifying expertise, and therefore both the claim to credibility and the claim to representativeness (especially the former) are bolstered by the very fact of invitation.

An offer of a platform is a signal that those who issue the invitation consider the person worthy of a respectful hearing. It is a signal that the inviters consider the speaker sufficiently expert, or sufficiently representative of expertise to have an opinion on that topic that should be taken into consideration. It is therefore also a signal that the speaker's dissent cannot be explained away as incompetence or insincerity. Because it certifies that we cannot dismiss their dissent on these grounds, the signal increases its epistemic significance, and thereby increases the pressure to conciliate. The strength of these epistemic signals given by the invitation is a function of the epistemic credentials of the platform. A major newspaper gives a much stronger signal than a local outlet or a little known blog. An academic department or a university gives a much stronger signal than does a fringe student organization (indeed, some platforms may actually detract from the credibility of the speaker). The stronger the signal, the greater the boost to the speaker's credibility and thereby to the epistemic significance of their dissent.

In fact, the credibility-conferring power of invitations is common knowledge, as witnessed by the way in which those who write opinion pieces for well-known outlets advertise that fact prominently. The very writers most prominent in opposing no-platforming are good examples. The faculty information posted about Jonathan Haidt on the New York University's Leonard N. Stern School of Business website mentions fact that ‘[h]is writings appear frequently in The New York Times and The Wall Street Journal, and he has given four TED talks’ (New York University 2019). FIRE notes that its president and CEO, Greg Lukianoff, ‘has been published in The New York Times, The Wall Street Journal, The Washington Post, Los Angeles Times, The Boston Globe, and numerous other publications. He frequently appears on TV shows and radio programs, including the CBS Evening News, The Today Show, and NPR's Morning Edition’ (FIRE n.d). Haidt, Lukianoff, and their organizations cite these facts because they recognize that their frequent invitations to appear in these outlets confer considerable credibility on them.

It is plausible that provision of a platform provides some degree of certification of representativeness, as well as of credibility. We may reasonably assume that the apparent expert given a platform does not speak only for herself, even before we have heard her claim representativeness. Credibility and representativeness may be traded: Nobel Prize winners, for example, might reasonably be given a platform despite lack of representativeness (Linus Pauling's advocacy for megadoses of Vitamin C, is a good example). In cases like these, invitation may not provide evidence of representativeness. But audiences seem warranted in assuming that reputable media organizations and other institutions filter out unrepresentative individuals unless they are genuinely exceptional in their expertise.

There is in fact indirect evidence that ordinary people infer representativeness from the provision of a platform. The consensus on climate change is very strong; roughly 97 percent of scientists with relevant expertise accept it (Cook et al. Reference Cook, Nuccitelli, Green, Richardson, Winkler, Painting and Way2013; Carlton et al. Reference Carlton, Perry-Hill, Huber and Prokopy2015). But ordinary people vastly underestimate the extent of scientific consensus (Leiserowitz et al. Reference Leiserowitz, Maibach, Roser-Renouf, Feinberg, Rosenthal and Marlon2014; Cook et al. Reference Cook, Oreskes, Doran, Anderegg, Verheggen, Maibach and Stuart Carlton2016). The degree of underestimation is particularly pronounced in the United States, where only 42 percent of the population believe that there is a scientific consensus on climate change (Hahn, Harris, and Corner Reference Hahn, Harris and Corner2016). This misperception may arise from the norm of ‘balanced’ reporting, whereby mainstream media organizations offer a platform to (so-called) skeptics to avoid the appearance of bias; whatever the cause, skepticism receives a degree of coverage far out of proportion to the degree to which it is genuinely representative of scientific opinion (Boykoff and Boykoff Reference Boykoff and Boykoff2004; Theel, Greenberg, and Robbins Reference Theel, Greenberg and Robbins2013).

Dissenting speakers provide first-order evidence for the claims they make. It is this first-order evidence on which the debate over no-platforming has focused. It is solid or specious? Should it be rebutted? But they (or perhaps their appearance) also provide higher-order evidence. As I discussed above, the strength of this higher-order evidence depends on what the best explanation for the dissent is. For experts in the topic, this may be obvious. For them, given their expertise, the question may be easy, or easy enough to rule out certain answers, or they may be able to dismiss a disputant from peerhood because they neglect certain evidence, for instance. But the audience lacks the expertise to break the symmetry between the disputants. Instead, the nonexpert must rely on various markers of expertise and representativeness, among which the provision of a platform by a reputable institution looms large. The invitation confers credibility on the invitee, certifying their expertise or representativeness, and it thereby constrains us to give their dissent significant epistemic weight.

Proponents of open speech call on us to listen to dissenting speech and make up our own minds. They acknowledge that the speech may be misleading but argue that misleading speech is best responded to by ensuring balance. Bad speech should be rebutted. While they have a persuasive point when it comes to first-order evidence, it bears emphasizing that higher-order evidence is much harder to rebut. We can argue with Charles Murray about the existence or causes of moral decline, or the extent to which IQ measures anything significant, but we cannot credibly deny that an invitation to speak at Harvard confers credibility. No amount of argument addressed to first-order claims can rebut this kind of higher-order evidence; calls to engage with the speaker, to address their arguments and present better ones, are powerless to engage with it (indeed, they may even boost the credibility of the speech, by presenting it as worthy of discussion). The only way to begin to rebut the higher-order evidence is to break the symmetry, by downgrading the speaker from peerhood or by attacking the epistemic credentials of those who issued the invitation: to attempt to show that they were not acting in good faith, for example. Rebutting higher-order evidence often requires ad hominem argument: we need to undermine the credibility of the speaker, not rebut her arguments. These are not, of course, the kinds of argument that defenders of open speech advocate when they advocate open dialogue.

It should also be emphasized that rebuttals of first-order evidence can be expected to reach a much narrower audience than those in receipt of higher-order evidence. Those who actually attend a lecture, read a transcript or summary, or listen to a podcast are in the receipt of the first-order evidence it provides. But the second-order evidence is broadcast to everyone who learns that the speaker defends a particular controversial claim and that they have been provided with a credibility-conferring platform. Indeed, the speaker need not even defend the controversial claim: it is enough that they are known to defend it. Charles Murray apparently devotes little time to The Bell Curve or its claims today, but invitations for him to speak constitute higher-order evidence in favor of its claims—perhaps even of what its claims are widely taken to be, whether they are accurately represented in the public mind or not—so long as he is taken to stand by them.

Defenders of open speech sometimes claim that those who object to speakers can seek to balance them, by inviting speakers with a contrary view. Balancing a speaker addresses the higher-order evidence that the initial invitation provided, not by undercutting it but by providing further higher-order evidence, this time for a contrary view. That leaves the initial higher-order evidence in play at full force. Suppose that a speaker is given a credibility-boosting platform and uses it to make claims that are false. Balancing that speaker by ensuring that their claims are rebutted by a rival speaker may ensure that all his first-order claims are addressed and shown to be false. But the higher-order evidence provided by the invitation remains in force.

Higher-Order Evidence and Epistemic Paternalism

As shown above, if proponents of no-platforming appeal to epistemic harms—the formation of false beliefs—by citing arational mechanisms of belief formation, they commit themselves to advocating some kind of epistemic paternalism. They must advocate the control of exposure to speech, with the aim of ensuring that more accurate beliefs are formed. Given, however, that both paternalism and the psychological mechanisms that might be invoked in its defense are controversial, it is important to point out that it is rational to be guided by higher-order evidence in coming to or adjusting our beliefs. It is no more paternalistic to attempt to reduce the amount of misleading higher-order evidence to which people are exposed than it is to attempt to reduce the number of lies to which people are exposed.

Higher-order evidence is genuine evidence. Most philosophers hold that cases like Restaurant Check, the disputants should lower their confidence in the figure they have reached, because the most likely explanation of their disagreement is that at least one of them has made a mistake, and neither has any good reason to think that it is more likely to be the other person than themselves. Philosophy being philosophy, there are of course some high-profile thinkers who maintain that at least in some circumstances, we ought to hold fast to our well-justified first-order beliefs despite recognizing that there is higher-order evidence of the kind that others take to motivate conciliationism. It is worth noting, however, that even those who maintain that we may permissibly believe that p at the same time as believing ‘my total evidence does not support p’ would admit that such a conjunction of beliefs is uncomfortable, and that other things equal, it would be better to avoid generating misleading evidence. (For a review of the prospects of ‘level-splitting’ views and the ways in which such a conjunction is uncomfortable, see Horowitz Reference Horowitz2013.) For third parties, moreover, the conciliatory verdict with regard to these kinds of case seems compelling. I cannot acquire knowledge via testimony from you if I know that your claim is contested by someone who is no less likely to be right than you are. Thus, I cannot acquire knowledge via testimony from an expert if I know that her claim is contested by a large body of well-qualified people. Their dissent is evidence for me, and I rationally take it into account in acquiring or changing my beliefs.

Speakers given a platform are thereby certified. The provision of a platform is evidence that their claims are worth taking seriously; their expertise (or borrowed expertise, when they do not claim it for themselves but purport to report on it) is thus testified to. Perhaps more weakly, the provision of a platform is also evidence that their claims are representative of a sufficiently large group of sufficiently expert speakers. Of course, these markers of credibility can be and often are disputed, but they cannot be defeated entirely. While an argument can demonstrate that the evidence you cite for p does not in fact support p, as long as the platform provided to you to say it is reputable, it just does confer credibility on you, no matter what anyone says. Evidence against the proposition that the speaker should not be attributed credibility can certainly be produced, but the evidence in favor of the proposition is not thereby eliminated. (Analogously, if it seems to me that I saw Frankie stabbing Colonel Mustard, providing me with evidence that she was elsewhere gives me a reason to think my perception was misleading; it does not eliminate its evidential force for me.)

It is paternalistic (justifiably or not) to withhold genuine evidence from an audience, on the grounds that it is misleading. But it is not paternalistic to prevent misleading evidence from being generated in the first place. Compare two ways in which a politician's minders might go about trying to prevent voters from getting the false impression that she is sympathetic to racists. They might pressure or bribe journalists from reporting that she attended a dinner hosted (unbeknownst to her) by racists, or they might turn down the invitation to attend. In the first case, they suppress misleading evidence on the grounds that it gives rise to a false impression. In the second, they ensure that the evidence is not generated in the first place. If there are reasonable grounds for thinking that avoiding the impression was in the epistemic interests of the audience, then they act paternalistically in the first case: they manage the agents’ epistemic environment without (or regardless of) their consent, in order to promote their epistemic welfare. But they do not act paternalistically in the second case. Nothing is withheld from the audience.

Similarly, it is paternalistic to suppress evidence that there are experts or even self-proclaimed experts who dissent from a proposition. It is paternalistic even if we have good reason to think that these experts (genuine or not) are mistaken. But it is not paternalistic to avoid generating evidence (or additional evidence) that dissenters are credible or representative. Refusing dissenting speakers a platform (or protesting the provision of a platform, with the intent that on this or future occasions they should not be given one) is working to prevent the generation of higher-order evidence, not paternalistically suppressing evidence. Whatever our views on the justifiability of epistemic paternalism, it should be uncontroversial that there are many instances in which preventing the generation of misleading evidence is justifiable.