Hide table of contents

TLDR: assuming that people are open to policing harmful speech, what kinds of speech should they actually choose to suppress? It seems to me that common tropes about harmful speech are just partly correct. Speech-cops should continue to suppress prejudice, but they should be more cautious about policing some ideas which are only vaguely associated with prejudice, and they should be more willing to police speech that contradicts some key tenets of the EA community.

Note: I haven't personally done any serious kind of speech-policing.


There seems to be increasing interest and capability in the West for policing speech in an attempt to suppress harmful attitudes in society. Speech can be policed either through legal means (e.g. hate speech laws) or via private action (protests, firing, milkshaking, online censorship, selective publication of research, etc). For this post I will ignore the distinction and treat speech policing as one issue. Also, I won't address whether speech policing is actually good or bad.

Instead, I will focus on the specific choice of which kinds of speech are most harmful, i.e. best to police. Speech policing has been advocated and performed against a variety of different ideas, often in very different partisan directions. And there are few good arguments about which kinds of speech are more harmful or beneficial than others. Assuming that one were to police harmful speech, which attitudes most deserve to be suppressed? Most speech-policers don't seem to put the proper amount of care into considering this question.

It should go without saying that information should only be suppressed among certain kinds of unethical people, if we can be so precise. The worry with harmful speech is almost universally that it can inspire unethical people to take bad actions. An ethical person, by definition, will only use information to do ethical things. For instance, if you tell an Effective Altruist that immigrants will lower domestic wages, she will shrug her shoulders and say that the needs of the many outweigh the needs of the few, unless it really makes immigration a bad idea all-things-considered; on the other hand, if you tell a narrow-minded nativist that immigrants will lower domestic wages, that will motivate him to be more strongly xenophobic. When someone is unethical, we seek to control them; when someone is ethical, we seek to empower them. Information sharing is empowerment; information denial is control. For this reason, we must absolutely reject speech policing and lying to EA community members. Instead, I am just hypothesizing about speech policing that might be pursued on college campuses, on general social media, etc.

Note that speech policing is very controversial, with some arguing for a classically liberal policy of restraint and tolerance. I won’t directly address this controversy here. I don't really have a stance on it, and I'm not arguing in favor of speech policing. But no matter what your view is, you must realize that speech policing incurs objective costs. It takes time and effort, it creates controversy and grievance, it causes people to be fired from productive labor, it makes victims feel unhappy and oppressed, etc. These downsides might be outweighed by the benefits of restricting speech, but they are still real and cannot be ignored. Therefore, merely being certain that a particular kind of speech is harmful is not sufficient to justify suppressing it. The speech has to be more harmful than a threshold in order to outweigh the practical harms imposed by speech policing. I don’t know what that threshold is, but it is still an important principle to keep in mind when deciding which kinds of speech are better or worse candidates for policing.

Another thing to remember is the risk of implicit censorship by association. If people who believe X also disproportionately believe Y, then censoring X-believers can suppress Y from the public discourse.

Keeping these issues in mind, here are my preliminary thoughts on which specific kinds of speech are more or less harmful. They are informed by the conclusions of the Candidate Scoring System, which not only rates specific candidates but also serves as a general compendium of political judgments and priorities. If you disagree or are confused by political assumptions made in this post, you may find answers in a relevant section of the Candidate Scoring System, specifically the “policy positions” and “weights” for relevant issues. However, some of my views in this post are based on personal judgments that are not written into CSS.

Harmful speech

Calls to violence

This one is pretty straightforward. Agitating for killings or other violence against minorities, political opponents, cops, landlords, etc can only do harm. An exception would be calls for properly orchestrated state violence against threatening actors, such as routine policing and strikes against terrorists, which are much more likely to have good consequences.


Speciesism, defined as not caring about animals merely on the basis of their membership in a different species, is clearly wrong and harmful.

In reality, perhaps the majority of “speciesism” is not like this. Instead, people believe that animals deserve less moral status by dint of being unintelligent, or being intended to serve humanity, or some other reason. But these attitudes too are good candidates for speech policing.

Vicious racism

Racism is tough to define, but a core type (discussed by philosopher Jorge Garcia) is specific disregard for the interests of people of other races. This is straightforwardly wrong, and its potential harm is demonstrated by a variety of regimes and nonstate actors, both historical and contemporary, who have pursued genocide, ethnic cleansing, biased policing, harsh sentencing, immigration restrictions, zoning restrictions, and other bad actions. Vicious racism is not the sole cause of racially destructive policies, but it presumably plays a major role.

Vicious racists don’t seem to uniquely hold other insights or valuable ideas, so the risk of implicit censorship is minimal.

White nationalism

None of the policies favored by white nationalists seem to be beneficial, and they pose a risk for committing major harms like mass deportations. In addition, they don’t seem to possess any unique body of useful ideas that liberal society ought to take into consideration.

Trans-exclusive speech

Claims that transgender people are not their transitioned gender, or that they are all mentally ill, or refusals to use their chosen pronouns, can create stigma that reduces their social acceptance. It reduces the prospects for antidiscrimination laws. It reduces the prospects for private and public funding of transition therapies. It reduces the prospects for children with gender dysphoria to get any kind of support. It has further been alleged to increase their risk of being murdered, but as far as I can tell, there is fortunately no significant evidence to substantiate the idea that there is an epidemic of hate-motivated trans murders.

The biological or philosophical reality of gender and sex is really not relevant here: even if all the transgender theory were wrong, that wouldn’t change the factual reality that affirming transgender status is socially beneficial.

The risk of implicit censorship is mild here. There are some significant concerns about people (especially children) being encouraged or pressured too much into irreversible or difficult-to-reverse gender transitions, and the broad suppression of gender-critical people or TERFs already seems to have interfered with debates about this issue; however health and legal professionals can do a fairly good job of working through this on their own. It’s OK if we are a tad over-enthusiastic in our early social attitudes towards gender transitions, we will learn more in short order. There are concerns about women’s sports, as the introduction of a small number of markedly biologically superior performers can reduce competitiveness and perceived fairness for the majority of female athletes (potentially outweighing the benefits of inclusion for a small number of transgender athletes); however this is simply not something that seriously affects a major part of the population. In any case, these concerns are perfectly well voiced by people who are basically fine with transgender identity and acceptance. It isn’t useful to bring actual transphobes to add to the chorus of what is being voiced more respectfully and eloquently by mainstream liberal actors (although some of these mainstream liberal actors are themselves regarded, incorrectly in my opinion, as transphobic in their own right).

So overall, it seems pretty straightforward that truly trans-exclusive or transphobic speech is harmful. On the other hand, just 0.6% of the population identifies as transgender and a good deal of therapy funding and trans-affirming communities and institutions already exist, so such speech may not always be harmful enough to outweigh the objective costs of censorship. For instance, Maya Forstater was fired for stating that men cannot change into women, and her termination was upheld by a court; while that speech act seems straightforwardly harmful, it may be much more harmful that her employer – a think tank which campaigns against global poverty – weakened its mission by firing her. Of course the think tank may have been responding rationally to popular pressure to fire her, but then the responsibility lies with the outsiders who applied that pressure rather than affirming the importance of their mission and organizational independence.

Anti-vaxx campaigns

This should need no elaboration.

Unscientific climate change denial

Assuming that someone is not presenting scientifically valid evidence to change the conversation, it is obviously harmful for them to generally promote denial of climate change.

In fact, even if man-made climate change isn’t real (though it really definitely is), it might still be good for people to believe in it, because the more direct cognitive, health and agricultural harms of air pollution already pose an important and neglected problem for humanity.

These policies are clearly and strongly beneficial. They are supported by robust combinations of empirical evidence, theoretical mechanisms, and the most common-sense liberal ideas of social fairness and freedom. Therefore, opposing them outright is likely harmful, although it can be beneficial to point out specific downsides in certain contexts.

Supporting Donald Trump

Trump is a clearly and particularly poor president, and both his impeachment trial and his election are forthcoming. Therefore pro-Trump sentiment is generally harmful.

(However, this doesn’t mean that pointing out particular virtues of Trump or refuting poor arguments against him is harmful. That can be healthy discourse.)

Selfishness and nihilism

Selfishness is a collectively harmful point of view, warping social and political attitudes in a variety of ways too numerous to list. Meanwhile, it is hard to see any upsides to selfish ideology. It’s true that harnessing selfish incentives can have good results, for instance in capitalism. However, that doesn’t mean that explicitly promoting that ideology over other attitudes is a good thing. It seems that selfishness/egoism is one of the better candidates for ideas to be suppressed.

Ethical nihilism could be considered similar, lacking in upside, and probably functioning similarly to selfishness for all practical purposes.

Disparaging EA/EAs

There are some legitimate criticisms of particular practices and ideas which are common in EA. While these are generally voiced adequately within the EA community by other EAs, it can still be beneficial to protect their freedom of speech in a public context, when they are being respectful and open to debate.

It is a different story when people criticize EAs for reasons that are less about substantive and fair disagreement and more about identity and affiliation. Complaints that EAs have the wrong political opinions, wrong cultural background, wrong race or gender, or the wrong mindset on rationality can be very pernicious and unresponsive to honest reason and ethical appeals. Insults, mocking, etc is of course very problematic, and accusations of racism/sexism and so on are particularly damaging in the current milleu. The harm of this sort of speech is that it can lead people to devote less or none of their time and money to important issues including global poverty relief, animal welfare, existential risk reduction, institutional reform and EA movement building. Even if someone makes it their mission to criticize just a part of EA, if they publicly do it in a cynical or inflammatory manner, the result is a general weakening of the overall prestige and power of the EA movement, to the detriment of all EA projects.

Because the EA movement is uniquely high-impact, it follows that poor criticism of the EA movement is uniquely harmful, at least compared to criticisms of other beneficial ideas, social movements and institutions.

Honorable mentions

Racial differences

Racial science could contribute to some of the same harms of racism. It seems particularly likely to promote restrictions against immigration, and such arguments are occasionally made by figures such as Jason Richwine. From the perspective of global welfare, this is not a good possibility. Countries will predictably err on the side of too much nationalism and too little immigration, and we should seek to steer them towards more cosmopolitan behavior. However, I think it would be incorrect to say that the main xenophobic political trends in America, India, Hungary, and other countries are substantially motivated by racial science. To the extent that laypeople on the right are motivated by a scientific kind of racism, it is more often a matter of ignorant assumption, not the most recent papers and blogs.

Carl (2018) describes downsides of the taboo on racial science. One is that it can increase people’s propensity to equate moral status with scientific attributes like intelligence (and if some kind of incontrovertible proof of racial differences were found, this would be very dangerous). Another is that the view of humanity as a ‘blank slate’ can inspire totalitarian efforts to remake humanity, or other policies which are more benign yet still wasteful. A third downside is that the taboo could interfere with useful medical treatments for people of different races.

I don’t see a good overall rationale for censoring careful studies of things like ‘human biodiversity,’ if the researchers reject xenophobic and nationalist ideas and write things responsibly. On the other hand, if someone more casually remarks that members of another race are inferior at something, that is harmful speech.

Stupefying narratives

One of the harms of anti-semitism is that it “makes people dysfunctionally stupid, unable to understand how the world actually works,” since they blame everything on a conspiracy. Simplistic ideologies which blame everything on unsubstantiated conspiracies by Jews, the mainstream media, capitalists, Ukrainian parliamentarians, etc could be considered stupefying narratives that just damage people’s ability to make sense of major issues.

There is a high risk of implicit censorship here. For instance, it seems that the West’s very strong attitude of anti-anti-semitism has buried legitimate criticisms of Israel and the pro-Israel lobby. Generally speaking, it’s hard to suppress irrational attachment to stupefying narratives without stamping them out entirely. We don’t want to make people ignore the influence of capitalists, the mainstream media, Ukrainian parliamentarians, George Soros, etc; we just want them to be less irrationally fixated on them. It is difficult to see how speech policing could be used effectively here, even if we ignore the obvious problem that censoring stupefying narratives will only entrench these people further in their conspiratorial mindset.

(Of course, vicious anti-semitism falls under the category of vicious racism, and is accordingly very harmful.)

Not harmful speech

Scientific disregard for animals

Some people argue that animals have little or no sentient feelings of pain. These views may not be accepted by the majority of neuroscientists, but they have enough potential merit that it cannot be appropriate to police them. The questions on the nature of consciousness are too murky.

Weighting animals less than humans on the basis of their sentience (or intelligence as a proxy for sentience) is almost certainly the right attitude to have, so of course it should not be policed.


Some people have condemned the idea of transracialism as offensive – for instance, see the controversy surrounding Rebecca Tuvel’s philosophy paper on the subject. However it’s not clear if this idea is actually harmful. There isn’t a serious problem of “race dysphoria” that needs to be solved via transracial practices, but it could be mildly beneficial for people to enjoy the ability to switch races/cultures. It might weaken racial and cultural identity, which could be bad by weakening a source of pride and valuable political mobilization, or could be quite good by weakening a source of sectarian tension and grievance. Generally speaking there just isn’t a clear case that transracialist ideas are harmful.

Human enhancement

Forcing people to go through eugenics like sterilization or abortions would be a very bad policy, though it is not as bad as some other kinds of bad policies promoted by racism. Fear of eugenics has caused some people to police speech which advocates it. However, there is little risk that benign measures (like scientifically informed dating services to prevent genetic diseases, or genetic engineering for parents to improve the happiness or intelligence of their children) would lead to such policies. Western society has very strong laws and norms against nonconsensual medical procedures (excepting cases where people are mentally ill), plus a deep antipathy towards any policies that evoke Nazism or scientific racism, so these new technologies are very unlikely to lead to oppressive policies. Meanwhile, there is a substantial chance that genetic engineering and other techniques could improve the health, ethics, happiness, productivity or other characteristics of new children. Some people worry that this could cause communities to wither – for instance, there will be no deaf community if nobody is deaf anymore. However, such organic shifts in social identities and culture are not harms.

So, while calls for coercive eugenics are harmful, speech which favors actions that are superficially similar but ultimately consensual and nonviolent is not. In fact, sweeping ideological attacks against human enhancement could be considered harmful speech in their own right.

Criticizing the American pro-Israel lobby

The pro-Israel lobby has a very strong and pernicious influence on U.S. foreign policy. Some people fear that condemnations of the lobby could lead to antisemitic attitudes, but widespread antisemitism is not a severe risk in American liberal democracy, and the costs to the Palestinian people (and even to the Israeli people, in the long run) from flawed American policy towards Israel appear greater.

A concluding point

It appears that common assumptions about what kinds of speech are harmful are just partly right. They correctly recognize many kinds of prejudice against different types of humans as being harmful, but can go too far in using only vague associations with prejudice as a justification for speech policing. Meanwhile, people must get better at recognizing the harms of advocating selfishness, speciesism, and false opinions on key policy issues.





More posts like this

Sorted by Click to highlight new comments since: Today at 2:41 AM
I don’t know what that threshold is, but it is still an important principle to keep in mind when deciding which kinds of speech are better or worse candidates for policing.

I think without more work on figuring out where this threshold is, no amount of proposals as to what is or is not over the threshold will seem beyond possibly being cases of "I just don't like this type of speech". Considering whether people believe a particular type of speech is beyond the harmfulness threshold to warrant censorship seems worthwhile as evidence about what may be harmful, but I don't think your presentation of what types of speech you consider harmful is that, thus much of this post to me reads like "what kinds of speech kbog thinks are harmful" and that's not very interesting or useful to me.

I'll also say that as a US citizen I'm culturally biased to immediately oppose any suggestion that we might censor speech, and only cautiously accept it if there is a really strong and compelling argument in favor of censorship for some particular type of speech. For example, of your entire list only "calls to violence" passes what I would consider the test of speech so harmful it should be restricted. I realize the US free speech norm is not globally shared, but it does make it hard for me to take your arguments seriously when you include, for example, political speech ("supporting Donald Trump"), since this is specifically one of the types of speech considered most in need of protection within US culture since it also the most tempting for political opponents to suppress on spurious grounds.

I'm not saying I can't be convinced some kinds of speech are harmful and that we should do more to restrict them, but I don't find your case compelling, especially around why you think some certain kinds of speech are harmful.

I think without more work on figuring out where this threshold is,

Well it will vary from context to context. When it comes to self-censoring, the threshold is almost nil - we almost instinctively refrain from saying things that we think are harmful. On the extreme end there are cases like harassment and legal prosecution which probably should have an extremely high threshold, like active incitement to violence. So it would be very complicated to address this.

thus much of this post to me reads like "what kinds of speech kbog thinks are harmful" and that's not very interesting or useful to me.

If you don't engage in any speech policing, then yes, I understand. This post only has relevance for those who would engage in some kind of speech policing. The question of whether speech policing is generally good or bad is not something I am taking a stance on here.

for example, political speech ("supporting Donald Trump"), since this is specifically one of the types of speech considered most in need of protection within US culture since it also the most tempting for political opponents to suppress on spurious grounds.

Legally, sure. In other contexts it is often the norm in America to deliberately avoid or forbid presenting information in a manner which can support certain political candidates and policies. Consider social media, partisan-leaning workplace cultures, biased media outlets, and so on.

I don't find your case compelling, especially around why you think some certain kinds of speech are harmful.

Well some of it is backed up by elements of CSS, others are more informed by other personal judgments. These are not rigorous arguments, but I figured that this would still be a step forward since no one else seems to have done this sort of systematic comparison and judgment (certainly not one that incorporates recent EA knowledge), and even my intuitive judgments should be useful due to my above-average familiarity with US politics and political culture.

I agree with most of your actual judgments here, although I'd be reluctant to draw lines as generally as you seem to have done.

On Donald Trump, I think he has bad values (engaging in many of the forms of harmful speech you describe here), but there have been instrumental arguments in support of him overall. I do suspect they're mistaken, but I'm not sure we should suppress support for Trump, at least when good faith and impartial arguments are presented. I think the same could apply to legal immigration or other practical policies here. If someone is making an argument for or against a position in good faith and in an impartial manner (e.g. without outright discounting the welfare of would-be immigrants), then we might not want to suppress such speech.

Maybe the kind + true + necessary standard could be applied to find exceptions to classes of speech we normally find harmful.

Ethical nihilism could be considered similar, lacking in upside, and probably functioning similarly to selfishness for all practical purposes.

FWIW, ethical nihilism is compatible with any normative view, including impartial ones. Brian Tomasik has written against moral realism (see especially the last section):


More from kbog
Curated and popular this week
Relevant opportunities