It is widely acknowledged by policymakers that efforts to restrict or suppress the sharing of false, misleading and harmful information must be carefully reconciled with the value of freedom of expression, as enshrined in national as well as international laws.² For example, a 2022 report of the secretary-general of the United Nations underlines that “responses to the spread of disinformation” (understood as information that is “inaccurate, intended to deceive and shared in order to do serious harm”) “should comply with international human rights law and promote, protect and respect the right of individuals to freedom of expression, including the freedom to seek, receive and impart information” (par. 11). Along similar lines, Europe’s recently implemented Digital Services Act says that service providers should take measures to mitigate the “systemic risk” of “disinformation,” “(giving) particular consideration to the impact on freedom of expression” (par. 86).

This standard acknowledgement of the need to temper anti-disinformation measures with respect for the value of freedom of expression is to be welcomed. Nevertheless, there is something quite misleading about the way the problem of “misinformation,” which I take to mean (intentionally or unintentionally) inaccurate/misleading and harmful information, is typically presented. One gets the distinct impression that there is a certain quantity of “misinformation” that can be straightforwardly identified by authorities, and then targeted by a variety of mitigation measures. On this picture, the regulator, whether a private Big Tech company, an organ of the State, or an international organ of governance such as the European Commission, can confidently identify “misinformation,” to wit, content that is inaccurate, misleading and harmful to this or that public value (e.g., public health, trust in public institutions, or rule of law), and then craft a strategy for restricting such content while giving due weight to the value of freedom of expression.

This picture, as I shall argue in this paper, vastly oversimplifies the nature of the threat posed by misinformation, and virtually ignores the significant risks of interventions designed to censor or suppress it, in particular the risk of wittingly or unwittingly sabotaging the truth-seeking function of public inquiry and debate. Undoubtedly, there are certain forms of communication that are rightly condemned by the moral sense of the community at large and widely viewed as deserving of legal censure or suppression, such as the transmission of child pornography and incitement to violence. The harms such communication poses to innocent citizens are quite evident to most reasonable citizens of good will, and very few would suggest that these sorts of communications contribute anything of value to public inquiry or reasonable self-expression. When it comes to claims advanced in the context of political and scientific debate among peaceful and law-abiding citizens, on the other hand, the case for censorship of any sort is considerably weaker.

The call to restrain or censor misinformation in the context of unfolding political and scientific debates, whether on the part of private corporations or governments, naively assumes that there is some selection procedure that can reliably pick out a special class of citizens who are uniquely qualified, on account of their superior knowledge or wisdom, to issue opinions on matters that are the subject of ongoing political and scientific debate, in a way that is epistemically authoritative or binding for the citizenry at large. By the same token, it naively assumes that those empowered to suppress inaccurate and harmful information are much less likely than anyone else to propagate or endorse “misinformation” (inaccurate and harmful information) themselves. If I am right, and there is no reason to assume that there is any selection procedure capable of reliably picking out a special class of persons uniquely qualified to arbitrate ongoing political and scientific debates, then the whole enterprise of curbing misinformation, at least in the context of unfolding political and scientific debates, is a fool’s errand. Indeed, if my argument goes through, then these sorts of speech restrictions do not actually involve any trade-off between freedom of expression and the public good of an epistemically sound discursive process: on the contrary, they involve a manifest harm both to freedom of expression andto the integrity of the discursive process as a vehicle for the discovery of truth.

The argument will unfold in four stages: First, I specify the sorts of speech restrictions that fall within the scope of my argument. Second, I argue that the effort by a regulator to restrict or suppress what he takes to be inaccurate/misleading and harmful information is unreasonable and counterproductive in the context of a lively public debate among peaceful and law-abiding citizens. Third, I argue that putting one’s faith in censorship as a tool for addressing misinformation in the context of political and scientific debate rests on a mistaken conception of a successful discursive process, as resting on pillars of truth artificially immunised from public challenge, rather than an evolving and unpredictable discovery process. Finally, I draw some practical conclusions from our discussion for our understanding of the problem of misinformation, and whether it admits of any convincing solution.

I am not arguing, then, that censorship is intrinsically evil or that speech should never be prohibited or penalised. Rather, I wish to cast doubt over the legitimacy and reasonableness of a limited class of speech restrictions, namely restrictions that either modify, restrain or eliminate communications (a) alleged to be inaccurate or misleading, and harmful to society, (b) representing one among several opinions held peacefully by law-abiding citizens in an ongoing political or scientific debate, (c) involving alleged harms that are not recognised by anything approaching a consensus among reasonable, law-abiding citizens (d) emanating from real individuals and organisations, rather than from bots impersonating individuals and organisations.³

False advertising, child pornography, and bomb-making manuals do not fall within the remit of my argument, since none of these represent one among several opinions held peacefully by law-abiding citizens in an ongoing political or scientific debate, while their alleged harms are recognised by a broad consensus among reasonable and law-abiding citizens. Nor do the opinions propagated by impersonal bots impersonating real persons fall within the scope of my argument, since patently fraudulent forms of communication have no plausible entitlement to be protected from censorship. By contrast, peaceful and law-abiding citizens can and do disagree about things like the risks and benefits of Covid vaccinations and lockdowns, the fairness and appropriateness of carbon taxes, the science of climate change, or the merits of competing political parties. Consequently, restrictions of speech about these sorts of issues do fall within the scope of my argument.

A variety of methods may be used to restrict misinformation: speech may be modified or edited by a censoring authority to bring it in line with an approved narrative; certain categories or instances of speech may be intentionally rendered less visible to its intended audience; certain speakers may be disabled from making further contributions if found guilty of breaking “content guidelines,” or this or that penalty may be attached to disapproved content, such as a fine or temporary suspension of the speaker’s access to a forum. Examples of these methods in action include the removal of black-listed books from Amazon, “shadow bans” by social media companies to reduce the reach of an author's social media posts, artificial suppression of certain results in Google search algorithms, and temporary and permanent account suspensions on social media based on content shared by a social media user.

This is no small problem, because there is no plausible basis for setting apart an enlightened class of thinkers with superior knowledge or intellectual endowments when compared with the rest of society. To begin with, nobody, not even the most educated or brilliant person, possesses perfect, infallible knowledge, whether on moral or scientific questions. No human being enjoys a form of knowledge or wisdom that is uniquely infallible or immune to challenge. The idea that there is a superior class of persons whose knowledge and insights automatically trump the knowledge and insights of others is inconsistent with ordinary experience, which confirms that people reputed to be highly knowledgeable and wise can make grave and even catastrophic errors.

Now, a defender of censorship might argue that I have set the bar too high: we may not require censors with infallible opinions, just censors whose opinions more reliablytrack truth than the opinions of those subject to their rules. All we need to justify a censorship regime, on this view, is a relative epistemic hierarchy, that is, a class of persons whose opinions on matters of public interest are substantially superior, on the whole, to those of the persons whose speech they are regulating. A censor need not be totally infallible, just a lot more epistemically reliable than the censored.

But assuming there is no natural or genetic marker of such superiority, how might the persons designing a censorship regime pick out the group of people whose opinions are epistemically superior to those of others? Let's assume, for the sake of argument, that there was in fact someone who, while not infallible, possessed a form of knowledge that was far ahead of most citizens, including their scientific peers, and therefore was qualified to stand in judgment over the opinions of others, flagging inaccurate and misleading claims for authorities to duly suppress. How might we go about identifying such a person, so as to hand them the power to arbitrate the moral and scientific claims of those less knowledgeable and wise than they?

One option would be to find a person possessing indisputably superior knowledge, to nominate the expert censors. But in that case, we would have to ask on what basis this nominator was picked out as epistemically superior and therefore eligible to nominate expert censors. In practice, superior knowledge in the expert censor, or in the person nominating the expert censor, would have to be identified or confirmed through a procedure that is publicly verifiable. Otherwise, we would be living under a censorship regime based on an act of blind faith in the superior knowledge of the censor, which would be completely contrary to the spirit of science and rational inquiry.

So what might such a publicly verifiable selection procedure look like? Since we cannot rely on blind faith in the epistemic superiority of the censor or the nominator of the censor, a sound nomination procedure would have to rely on some sort of publicly intelligible epistemic “short-cut” or proxy for superior knowledge or wisdom, of the sort that can be recognized by politicians and ordinary citizens who are not necessarily themselves endowed with superior knowledge, and do not necessarily themselves have intimate knowledge of the fields of knowledge they wish to see regulated. The most obvious such proxy is the reputed knowledge of a person or institution. For example, someone might be nominated as a censor because they have a Ph.D from Harvard University, or an impressive publication record, or a Nobel prize, or heart-warming letters of recommendation from other well-regarded experts.

The problem is, none of these credentials, no matter how impressive, can reasonably guarantee that someone is so outstandingly superior as a scientist or thinker that they deserve to stand in judgment over the claims advanced by their colleagues and fellow citizens. For neither moral nor scientific knowledge and understanding neatly track professional prestige. Indeed, professional recognition and adulation, which is influenced by non-scientific factors like politics and Groupthink, can push in a very different direction to scientific progress and enlightenment. The fact that one individual wins celebrity status among their peers and another does not, does not tell us which of these individuals is wiser or more insightful in their judgments. The fact that one scientist's work finds favour with a Nobel committee or attracts the patronage of an important institution does not necessarily mean that other scientists with different credentials, or less glamorous credentials, are less reliable or have an inferior grasp of reality.

Under a censorship regime controlled by what is alleged to be “expert knowledge,” those reputed to be wiser or more knowledgeable than others would have the authority to declare, by fiat, that the opinions of certain citizens and scientists should be censored or wiped from the public sphere, just because those currently occupying the chair of the censor view such opinions as “inaccurate/misleading and (either intentionally or unintentionally) harmful.” The problem is, reputational indicators do not uniquely pick out the brightest minds or the most reliable sources of information, from a God's-eye perspective. For someone may have acquired their reputation through dubious or fraudulent means. And even if someone has earned their reputation for wisdom and knowledge through honest means, this does not qualify that person uniquely to sit in judgment over others’ opinions. For it is quite normal to see sensible citizens and well-respected and credentialed scientists and scholars on both sides of a scientific or moral debate. Consequently, a censorship regime built on reputation or prestige will produce a situation in which disagreements among people reputed of roughly equal epistemic caliber are to be settled by a raw act of power. That is violence, not scientific debate and inquiry.

The fact is, there is no procedure we can employ to reliably identify an expert class whose views may be automatically considered to be epistemically superior to those of their peers, or deserving immunity from criticism. If we accepted that such a class of persons could be identified and entrusted with the task of unilaterally purging political and scientific debate of “misinformation,” we would have to reject the dominant understanding of the scientific enterprise as the presentation of evidence-based hypotheses susceptible to public refutation and correction within the scientific community. For under a regime in which certain individuals or cohorts can unilaterally censor what they deem false/inaccurate and harmful information, the opinions of the censors – who, as we have seen, are not necessarily better qualified judges than those they are censoring - are effectively shielded from public challenge, correction, or refutation by their peers. And this is the very antithesis of science and rational inquiry.

Even if we grant that there may be some instances of benign and enlightened uses of political and scientific censorship, the history of censorship suggests that these are the exception, not the rule. Historically, those wielding censorship powers have used said powers to silence their critics and protect themselves and society from uncomfortable truths and unsettling questions. This is a recurring pattern, from the political persecution of ancient Greek philosophers to book-banning in early modern Christendom, to book-burning in Nazi Germany, to the incarceration of dissident journalists in places like Turkey and China, to the suppression of coronavirus lab leak theories on Facebook. It would be naïve to believe that human beings have evolved to such a point that this historical tendency will be abruptly terminated by a more saintly or morally mature generation of rulers, or to think that 21st century Western governments and media corporations are somehow immune from this tendency. The flow of information and ideas shapes the mind of the public, and those who hold power in society, whether financially or politically, have a natural interest in ensuring that information and ideas in the public sphere do not harm their public image and their vested interests.

Censorship is as old as politics. It will always be in the interests of some (in particular, those who wield a lot of power over others) to control the flow of information and arguments, whether to protect their careers or reinforce a narrative that helps keep them in a position of power or influence. Powerful actors, given the opportunity, often do not hesitate to silence voices that undermine their policies or their social position, independently of the truth value of the censored content. Authoritarian rulers silence their political critics, since loud criticism of their regime is perceived as a political threat. Large corporations have a powerful incentive to silence whistle-blowers and any other voices that undermine their products or reputation. The perennial sway of these incentives does not mean that censorship powers will always and everywhere be abused for private gain. But it does mean that censorship powers will always be vulnerable to co-option by vested interests, and powerful actors will not be slow to recognise the strategic advantages such powers represent. In short, besides the fact that the epistemic superiority of the censor does not stand up to scrutiny, the ability to censor others is a power that invites abuse and corruption.

The human quest for truth is a bumpy discovery process, with unexpected twists and turns, not a form of inquiry whose outcome can be predetermined or rigidly controlled by a preconceived notion of Truth, uniquely available to a special anointed class of “experts”. The truth emerges gradually, through an ongoing process of correction and refinement, a process in which evidence and arguments play at least as important a role as epistemic credentials and prestige. As one author put it, “progress in the sciences is ragged and uneven, and each step, like each crossword entry, is fallible and revisable” (Haack 2008, 29). This process of correction and refinement can only occur under conditions in which participants in the conversation are free to advance their opinions and raise whatever objections they see fit to the opinions of others. Any attempt to protect a certain set of opinions from criticism and challenge artificially short-circuits the discovery process, substituting the dogma of the censor for an evolving consensus validated by rational scrutiny and debate. There is simply no way to decide, for once and for all, who is closest to the truth, or who is the most brilliant mind in the room, in the absence of open and uncoerced rational inquiry and debate. As the English liberal and civil rights activist John Stuart Mill so aptly put it,

The systematic restriction of misinformation can impede the discovery of knowledge in two ways: first, by suppressing or artificially reducing the visibility of contributions to the conversation that are disapproved of by the regulator. For example, a regulator may order a book to be taken off the shelves, or might decide to shadow-ban certain social media posts, rendering their content invisible to most of their intended audience. In this way, the regulator puts his thumb on the scales to ensure that certain perspectives and arguments are given much less attention than others. This obviously disrupts the discovery process by pre-emptively suppressing certain ideas or reducing their visibility just because this or that authority says so, rather than allowing them to be tested out along with others.

A second way censorship can impede the discovery process is by having a chilling effect on the public conversation. When people notice that opinions that wander outside the approved narrative are suppressed, censored, or penalized, they are less inclined than they would otherwise be, to challenge the dominant narrative. For example, if you have a Twitter/X account with a large following that you have built up over many years, and you know that voicing opinions critical of the dominant narrative, whether on transgender identity, Covid vaccines, or the war in the Ukraine exposes you to the risk of having your account either “shadow-banned” or shut down entirely, you may be inclined to avoid voicing dissenting opinions on such matters. If enough people stay quiet for fear of being censored, then the overall effect is to drastically reduce the presence of some opinions and arguments in the public sphere. This gives an unfair advantage to censor-endorsed opinions – an advantage that is won by violence and power, rather than by the force of the better argument.

The irrational character of interventions aimed at suppressing one side in an unfolding political and scientific controversy is abundantly evident from the history of censorship itself. Occasionally, a censor may get things right – this is statistically inevitable. But when the censor pre-empts the outcome of a complex debate based on his own subjective determination, he is crowning an epistemic victor prematurely, and this is bound to result in mistaken or implausible views being protected from criticism and taking on an unjustified aura of invincibility. Consider, for example, how Facebook and other social media companies censored claims attributing the coronavirus to a lab leak, on the basis of statements by prominent scientists, but then were forced to withdraw the policy when leading scientists – including those who had initially denied the lab leak hypothesis, such as Anthony Fauci – admitted such a scenario could not be ruled out. Or consider how critics of Covid vaccines were censored early on on platforms like Facebook and Twitter, but were finally allowed to air their criticisms, once official regulatory agencies came to acknowledge safety concerns surrounding some of the Covid vaccines. In this situation, valid criticisms that anticipated the findings of regulators were silenced, and a particular narrative denying safety concerns, a narrative that turned out to be false, was protected from public challenge.

But where does all of this leave us in regard to the problem of inaccurate/misleading and harmful information circulating in the digital public sphere? That there is inaccurate/misleading and harmful information circulating in our public sphere, and that this is sometimes quite deliberate and malicious, seems evident enough. For example, pharmaceutical companies have paid out large amounts of money in legal settlements for misrepresenting the safety or benefits of their products, in ways that put patients at unnecessary risk.⁴ Similarly, a citizen who is falsely assured that a life-saving medication is dangerous may forego an opportunity to promote her own health or her child's health. It seems hard to argue with the proposition that we all have an interest in reducing the amount of false, inaccurate and misleading information in the public sphere.

But it does not follow from this that we can readily trace inaccurate/misleading and harmful information to a few identifiable sources, in a way that is publicly verifiable, or safely entrust such a task to a specific institutional body. To begin with, the idea that those with the power to censor, whether governments, private corporations, or “fact-checking” services, can be counted upon to only share true and accurate information, and never indulge in misleading propaganda that serves special interests, is implausible and naïve. There is nothing about the office of a censor, whether in a private or government organization, that makes the holder, or his masters, uniquely immune to political bias, financial incentives, scientific error and confusion, or the tug of special interests.

For example, the very same governments that condemn “misinformation” perpetuated by their enemies almost always dress up their own war campaigns in noble, altruistic and patriotic language, even if they are fundamentally wars of conquest. Similarly, the authority to censor has been used historically to suppress political dissenters and defend authoritarian and totalitarian ideologies, whether communism in Russia or national socialism in Germany. More recently, we have seen social media companies use their authority, sometimes with the encouragement of governments, to suppress claims that later proved true or at least plausible, such as the idea that Covid emerged from a lab, the idea that mRNA vaccines carry significant health risks, and the idea that the evidence for the efficacy of community masking was far from conclusive.

All of this suggests that the perpetrators of misinformation are spread out across society, and may be private citizens, governments, private corporations, fact-checking organisations, or social media companies. If we erect a fact-checking or disinformation authority authority, that authority may itself be co-opted by private interests, it may knowingly propagate misleading information, or it may innocently embrace significant factual errors. Thus, drawing a clean line between sound and unsound sources of information is, it turns out, a very tricky problem, which cannot be solved by simply delegating it to this or that epistemic authority figure or institution. For any person or institution we choose to treat as epistemically authoritative is liable to silence true as well as false opinions, because we have no good reason to assume that such a person or institution will either be epistemically reliable on publicly contested issues, or detached from political and financial special interests.

This means that the problem of misinformation is not one that can be solved by an independent adjudicator who sits above the discursive process, intervening in a god-like fashion to purge the public sphere of false and misleading claims. However, it does not follow from this that there are no mechanisms whatsoever that we can leverage for weeding out false and misleading claims. To begin with, there is the discursive process itself. A free and open discourse permits conflicting truth claims to present competing arguments and bodies of evidence. Over time, the weaknesses in certain claims and theories may be exposed by the pressure of counter-arguments. This process may take years, decades, or even centuries. But certain ideas that are considered self-evident at one point of time may come to be seen as mistaken at a later time, due to the weight of emerging evidence and objections.

Having said that, we should not exaggerate the power of the discursive process. Although free and open discourse is the only way to put claims properly to the test and to compare and contrast the merits of competing arguments, the success of an open conversation at unearthing truths and exposing untruths will obviously vary according to the ethical and scientific quality of the participants, the quality of the moderation procedures, and whether the rules of the conversation encourage robust and civil dialogue, or permit participants to hold each other accountable for their claims.

The mere fact that the conversation is non-coercive and open to differing perspectives does not guarantee that the participants will be intellectually serious or committed to the pursuit of truth; nor does it guarantee that discursive institutions will be robust in the face of frivolous and malicious interventions such as flooding a channel with bots, shouting down speakers, or engaging in vacuous insults. Finally, even in the best case scenario, in which the discursive process is well designed and undertaken by honest and ethically serious people, there are likely to be many complex and difficult issues upon which people of goodwill will disagree for the foreseeable future, or upon which the fact of the matter may remain uncertain for the foreseeable future.

These are important caveats concerning the power of free inquiry as a vehicle for uncovering truth. Notwithstanding these caveats, if my argument goes through, then we can at least rule out one strategy for protecting the integrity of the discursive process as intrinsically ineffective and counterproductive, namely centralized or top-down suppression of claims deemed by the censor to be untrue or misleading, whether through private censorship (e.g., by social media corporations) or government-sponsored censorship (e.g., prohibition of certain books, or legally requiring social media companies to censor “misinformation”).

The temptation to use heavy-handed censorship tools to purge political and scientific discourse of “misinformation,” i.e., false, misleading and harmful content is not likely to go away anytime soon. In some cases, people will pursue censorship strategies because they mistakenly but honestly think it is possible for a censor to identify and restrict misinformation in a publicly authoritative and non-partisan manner, and that this will help improve the quality of public discourse; in other cases, people will favour censorship strategies because they think this will protect their own public image or vested interests. Whether the push for censorship of political and scientific debate is well-intentioned or not, it must be firmly resisted, because ironically, as I hope to have shown in this article, the effort to suppress what is alleged to be “misinformation” actually subverts rational inquiry and undermines the conditions under which false, harmful and misleading information can be effectively exposed.