I wrote an initial draft of this post much closer to the Manifest controversy earlier this summer. Although I got sidetracked and it took a while to finish, I still think this is a conversation worth having; perhaps it would even be better to have it now since calmer heads have had time to prevail.
---
I can’t in good faith deny being an effective altruist. I’ve worked at EA organizations, I believe many of the core tenants of the movement, and thinking about optimizing my impact by EA lights has guided every major career decision I’ve made since early 2021. And yet I am ashamed to identify myself as such in polite society. Someone at a party recently guessed that I was an EA after I said I was interested in animal welfare litigation or maybe AI governance; I laughed awkwardly, said yeah maybe you could see it that way, and changed the subject. I find it quite strange to be in a position of having to downplay my affiliation with a movement that aims to unselfishly do as much as possible to help others, regardless of where or when they may live. Are altruism and far-reaching compassion not virtues?
This shame comes in large part from a deeply troubling trend I’ve noticed over the last few years in EA. This trend is towards acceptance or toleration of race science (“human biodiversity” as some have tried to rebrand it), or otherwise racist incidents. Some notable instances in this trend include:
- The community’s refusal to distance itself from, or at the very least strongly condemn the actions of Nick Bostrom after an old email came to light where he used the n-word and said “I like that sentence and think that it is true” in regards to the statement that “blacks are more stupid than whites,” followed by an evasive, defensive apology.
- FLI’s apparent sending of a letter intent to a far-right Swedish foundation that has promoted holocaust denial.
- And now, most recently, many EAs’ defense of Manifest hosting Richard Hanania, who pseudonymously wrote about his opposition to interracial marriage, cited neo-Nazis, and expressed views indicating that he didn’t think Black people could govern themselves.
I’m not here to quibble about each individual instance listed above (and most were extensively litigated on the forum at the time). Maybe you think one or even all of the examples I gave has an innocent explanation. But if you find yourself thinking this way, you’re still left having to answer the deeply uncomfortable question of why EA has to keep explaining these incidents. I have been even more disturbed by the EA forum’s response. Many have either leapt to outright defend those who seemed to espouse racist views or urged us to view their speech in the most possible favorable light without consideration of the negative effects of their language.
Other communities that I have been a part of (online or otherwise) have not had repeated race-science related scandals. It is not a coincidence that we are having this conversation for the fourth or fifth time in the last few years. I spend a lot of this post defending my viewpoint, but I honestly think this is not a particularly hard or complicated problem; part of me is indignant that we even need to have this conversation. I view these conversations with deep frustration. What, exactly, do we have to gain by tolerating the musings of racist edgelords? We pride ourselves on having identified the most pressing problems in the world, problems that are neglected to the deep peril of those living and to be born; human and non-human alike. Racial differences in IQ is not one of those problems. It has nothing to do with solving those problems. Talking about racial differences in IQ is at best a costly distraction and at worst a pernicious farce that risks undermining so much of what we all hope to achieve.
Why might one tolerate this?
One explanation for tolerating race science is that you think it is true and ought to be promoted. I don’t have very much to say to people in this camp.
The stronger argument (and the one I suspect more people are willing to defend in the open) is that truth-seeking requires engaging with uncomfortable and impolite questions. Limits on free discussion can hinder free inquiry in arbitrary ways and broadly chill speech that is necessary to challenge the current social consensus. There is a lot to say for this argument and we have gained a lot from our willingness to engage with ideas that run counter to common sense. No one took AI risk seriously 6 years ago and the vast majority of people still routinely fail to include animals in their moral circle.
However, this principle should not be taken as a maxim to be honored regardless of the cost. There is a bar for when dedication to no-holds-barred “intellectual inquiry” (more on that in a moment) should take a backseat to more pressing goals. Thinking that the bar is very high (as I do) is different from thinking the bar does not exist at all. I have a hard time seeing how curtailing discussion in EA of racial differences will restrict our ability to seek truths that expand our moral circle or develop better models of what AI alignment could look like.
Another compelling argument worth addressing is that the answer to bad speech is good speech; the truth will win out in a truth-seeking environment. The problem here is that some issues and lines of inquiry have higher-level distorting effects on the community. I was particularly struck by this passage from a recent review of Manifest, as I think it illustrates this dynamic in a microcosm:
[T]he organizers had selected for multiple quite controversial speakers and presenters, who in turn attracted a significant number of attendees who were primarily interested in these controversial topics, most prominent of which was eugenics. This human biodiversity (HBD) or “scientific racism” curious crowd engaged in a tiring game of carefully trying the waters with new people they interacted with, trying to gauge both how receptive their conversation partner is to racially incendiary topics and to which degree they are “one of us”.
Most people are turned off by race science. A rational person who knows they do not have time to thoroughly investigate every community or ideology they come across will rely on heuristics to make quick judgments about what they should take a deeper look into. “Do some people in this community seem to endorse or at least tolerate race science?” is one of the easier heuristics one can rely on to weed out unpalatable communities and ideologies. Had I known at the outset how many people in EA were seemingly tolerant or supportive of race science, I would have quickly applied that heuristic, dismissed the movement as a bunch of fringe weirdos, and moved on with my life. I am sure many people who would have been great contributors to the EA project made just that calculation, although we’ll obviously never hear from them.
When a community is permeated by exclusionary ideas, it becomes particularly hostile to the targets of those ideas. People of color, and particularly Black people, are underrepresented in EA. While questions of representation are complicated, I have a very hard time believing that there isn’t some non-trivial causal relationship between some EAs openly flirting with race science and the absence of racial minorities in EA. I’m not sure exactly how many people who otherwise would have been EAs have been turned off by some EAs endorsing race science or have left EA in frustration. That is by nature a difficult thing to track, but I think the qualitative case for assuming that it’s a non-trivial number is quite strong. I don’t intend to portray people of color as a monolith and recognize that there are likely some in the community who will disagree with my characterization (see e.g. this post defending Bostrom). As proponents of human biodiversity are ironically fond of noting, I am speaking in averages and talking about what I assume is likely to be true on balance.
On a hard-nosed epistemic level, we lose out from a lack of diversity. Intellectual insularity is dangerous. In my experience, very smart people systematically overestimate their ability to steelman every possible objection to their position. To truly truth test, you need to expose your arguments to those with radically different priors, viewpoints, and experiences.
One might respond to this by pointing out that I am advocating for making EA more insular by excluding people who want to discuss HBD and related concepts; perhaps I’m choosing the people who get offended over the offenders. My response is: yes, I am. To the extent that these kinds of decisions unavoidably have a zero sum element, I want to include people in EA who detest race science rather than those whose participation is contingent on discussing race science because they view it as a litmus test or hallmark of epistemic integrity, or who view the lack of toleration of race science as a canary in a coalmine for epistemic degradation.
Zooming out, this goes beyond pragmatic considerations and gets at the heart of the question of what kind of community we want to build. I want fellow travelers who are truth-seeking, yes, but also compassionate, wise, skeptical of unfettered attempts to maximize, and dedicated to making EA a welcoming space. There is another kind of community we could be, one that is committed to entertaining every controversial topic in EA settings, without respect to its relevance to EA and the damage it might cause. Tolerating flirtations with race science means choosing the latter.
Why does it have to be this thing in particular?
A culture of truth-seeking is one of EA’s largest comparative advantages; some may even say its greatest comparative advantage relative to other social movements. I could imagine advocates of a very strong truth-seeking culture saying something like: it’s true that HBD is unseemly and uncomfortable, but we have to be uncompromising in our interrogation of difficult questions, even those that run afoul of social taboos and common sense. After all, didn’t longtermism itself come from a willingness to engage with fringe and counterintuitive positions?
Broadly, my response to this is that one of these things is not like the others. Race science is noxious, both in the scale of historical atrocity that it has enabled and the harm that tolerating it does to our community. And crucially, it is also just so unclear to me why we need to talk about it at all. It doesn’t have anything to do with our core cause areas and certainly does not help us make any progress on the most pressing problems within those cause areas. We don’t spend much time on the forum discussing baseball either, and baseball certainly does not have the baggage of race science.
Reinstituting the taboo on discussing race science doesn’t mean that we need to chill free inquiry into other difficult and taboo topics. Here are nine things off the top of my head on which we could do difficult truth-seeking instead. All of these topics are more relevant to EA cause areas or the lives of EAs and taking any of these ideas seriously sends a pretty costly signal about a willingness to be open-minded:
- Insect welfare
- Polyamory and more broadly whether traditional conceptions of monogamous relationships ought to be challenged
- Digital minds
- Suffering risks
- Ending predation
- Whether it might actually be better for China to win the AI race (and more broadly whether the Chinese social and political system has advantages over Western liberal democracy)
- Pro-natalism
- The meat-eater paradox
- How EA should engage with major AI labs going forward
I have yet to hear a single person defend why race science needs to be the taboo issue of choice. I view the choice of race science as a litmus test for truth seeking with deep suspicion — it is hard not to view the people doing it as knowingly or unknowingly racially insensitive.
Some of the topics I listed above could make people uncomfortable. For example, someone who has been a victim of political repression in China may very understably have a strong reaction to a discussion of the possible benefits of the CCP system. An easy line to draw is whether the topic is completely irrelevant to the project of EA and has the possibility of demeaning or excluding people based on their immutable characteristics. Race science clearly falls into this bucket, as would conversations about whether trans people actually exist. Rejecting this fairly narrow and workable exception to otherwise strong truth-seeking norms strikes me as quite dogmatic. We should be wary of absolute principles; at the end of the day, dogmatism is antithetical to truth-seeking, ironically so when the dogmatism is about truth-seeking itself.
On the subject of good faith
The view that says that good speech will eventually win out over bad speech in a truth seeking community also assumes that there is some sort of fair debating game going on in which all participants are conversing in good faith. One of the aspects of EA I have found refreshing is its willingness to give people the benefit of the doubt and to fully hear them out to understand the best version of their ideas. However, I have often seen this bleed into a view (especially among rationalists) that as a best practice we should more or less take what people say at face value. This is, to put it simply, not how the world works. In the real world, people lie about their beliefs and intentions; they obfuscate and strategically misdirect. Online neo-Nazis are particularly adept at taking advantage of this.
I want to be clear that I am not accusing any particular person of being a secret Nazi; that is an extremely serious accusation that requires a very high burden of proof. It is well-documented that online racists use strategic ambiguity, saying things that push the envelope and Overton window only to back down and say it was a joke or that what they said was a joke/misunderstood/taken out of context. I don’t think that we should take Richard Hanania at his word that he is entirely reformed given his history of quoting neo-Nazis under a pseudonym. I think straightforwardly taking him at his word would be naive. We should still strive to read others in good faith and not allow debates about important ideas to collapse into ad hominem attacks; I am very specifically arguing that we should make an exception to this presumption of good faith for people advocating for racism or race-science-adjacent ideas.
Another intuitive response is that bad ideas and speech go underground if you try to censor or suppress them. Therefore, it is best to confront bad ideas directly, in the open. In general I think this is quite compelling, but I don’t think it’s right in the very specific circumstance I’m talking about. We’re already talking about these ideas out in the open and the costs in terms of damage done to the community are quite high. As I’ve said, it’s not clear to me why the EA movement has to be the forum for challenging race science anyway. EA is already fairly professionalized and probably needs to become even more so to achieve its lofty ambitions. Movements and communities that are more professional and focused tend not to associate with race science or the people who promote it.
Where do we go from here?
I think it is important to be concrete about what you’re advocating for, especially on mushy topics like community attitudes and discursive norms. I would like to see something in the direction of the following reforms.
The EA Forum should ban any discussion of race science, “human biodiversity”, or racial differences in IQ. I would ask readers to remember that I am only talking about the EA Forum here. I’m not calling to use the power of the state to censor these ideas. If you really, absolutely need to talk about race science, there are other places you could go. Just not here and not in our community.
Major EA organizations and leaders should publicly disavow race science and human biodiversity. EA is quite top heavy. Major organizations like CEA and OP taking a strong and clear stance on an issue has a large effect on shaping community norms. Normally such organizations should be cognizant of the power they hold and restrained about using it, but I think this is an exception.
EA funders should avoid giving money to people or organizations with a history of associating with race science. This would easily include funding anything associated with Richard Hanania, as well as the other highly controversial speakers at Manifest. I am also of the view that absent a much more genuine and detailed reckoning with his racist past, EAs should have a much higher bar for supporting Nick Bostrom. My view overall is that it’s better to be safe than sorry when it comes to this sort of stuff. Given how bad it is for EA to be associated with race science, in terms of community composition and broader reputational damage, I think it’s worth missing out on a small number of potentially valuable projects or unfairly cutting a few people off if the net effect is a much healthier community.
EA should avoid any public association with people who have a history of making statements sympathetic to race science. This has similar issues to the last suggestion. Following this prescription almost certainly means dissociating ourselves from people who really can offer us something that would help us solve our most pressing problems. This tradeoff is starkest in the case of Bostrom, who has made undeniably massive contributions to longtermism, but I would expect to see it at a smaller scale, e.g. Richard Hanania could plausibly have something to contribute about prediction markets at Manifest.
EAs should be empowered to speak out against race science and its proponents. I mean speak out in the sense of saying “I find these views extremely disconcerting and don’t think they have any place here” rather than trying to debate the merits of whether there are racial differences in IQ. I worry that the last couple years of polarization conversations on the forum caused people holding more “average” views relative to the community to keep quiet. I am writing this post with the sincere hope that I am in fact speaking for a sizable number of people who have also found recent conversations on the forum upsetting and out of step with the values they hold and the movement they thought they joined. If you find yourself in this group, know that you are not alone.
---
If we really believe that we are working on the most pressing problems in the world, we need to be serious and hold ourselves to a higher standard. No one else who is seriously working on problems at the highest levels of importance openly tolerates any association with race science. We shouldn’t either.
Imagine yourself in the future where we have failed (assume for the sake of argument that you still exist in this future). Take a hard look in the mirror and ask yourself: do you really believe that we failed because we weren’t tolerant enough to people who wanted to debate race science? Or did we fail because we stayed too insular, too online, too unwilling to professionalize and accept the burdens of integrity imposed by the monumental task we set ourselves to?
You say that you want to save the world? Then act like it.
Forum users can, should, and do downvote posts that are bad, distracting, etc. (The trolls should soon get the message and leave.) I'm very opposed to top-down hierarchical interventions of the sort you describe. I don't particularly think that EA spaces should host "unequivocal condemnations" of things that (as you rightly note) have nothing to do with EA, so I'd also encourage people to downvote those. It's groupthinky and cringe, and risks being massively off-putting to the kinds of independent thinkers who value epistemic integrity and have little tolerance for groupthink or witch-hunts, however meritorious the message (or wicked the witches).
You should look into how universities work! Academic freedom means that individual professors are free to condemn whatever views they find obnoxious. They're also free to invite speakers that their colleagues find obnoxious, and sometimes they do (even, e.g., at Princeton). Their colleagues -- many of whom work on important problems! -- must then tolerate this. Note that many of the best universities follow the Chicago Principles & Kalven Report guidance on institutional neutrality, according to which the university leadership should express no official opinion on matters that aren't directly relevant to the running of the university. The university is a community of scholars, of diverse opinions, not a political party with a shared orthodoxy.
I would much prefer for the EA community to model itself on universities than on political parties.
Again, that doesn't mean that anything goes. We already have the solution to bad contributions. It's called 'downvoting'.
Effective altruism is meant to be a social movement, not a university debate. And unlike in a university setting, there are zero requirements for someone to be accurate or to have relevant expertise before posting here.
It is common here for people with little expertise in a topic to do an arbitrary amount of online research and throw out their resulting opinions. This results in something like the post where someone cited "mankind quarterly" for their human genetics posts, without mentioning that it was a publication with a history of white supremacy, fraud and incompetence. That issue was caught, eventually, but I guarantee you the forum is riddled with similar problems that are not caught.
For a regular topic, these loose standards may be acceptable, as it makes it easier to throw out ideas and collaborate, and the air of loose discussion makes things fun. Someone may chime in with corrections, someone may not, ultimately it is not a big deal.
But when it comes to race science, the consequences of this sort of loose, non-quality controlled discussion is worse. As the OP mentioned, you drive away minorities, and make the forum an unpleasant place to be.
But it also might convince more people to be racist. At least one white supremacist has traced their radicalisation pipeline to go through lesswrong and Slatestarcodex. That was just one person out of forty, so perhaps it was a fluke, or perhaps it wasn't. Perhaps there are a few that didn't go all the way to posting on white supremacist forums, but became just a little bit more dismissive of black people on job applications. I don't know how high the cost is, but it exists.
The way I see it, the forum should either hold back any race science related post and ensure that every claim made within it is thoroughly fact checked by relevant independent experts, or it should just ban the things. I prefer the latter, so we don't waste anybody's time.
In fairness, expertise is not required in all university settings. Student groups invite non-experts political figures to speak, famous politicians give speeches at graduation ceremonies etc. I am generally against universities banning student groups from having racist/offensive speakers, although I might allow exceptions in extreme cases.
Though I am nonetheless inclined to agree that the distinction between universities, which have as a central purpose free, objective, rational debate, and EA as a movement, which has a central purpose of carrying out a particular (already mildly controversial) ethical program, and which also, frankly, is in more danger of "be safe for witches, becomes 90% witch" than universities are, is important and means that EA should be less internally tolerant of speech expressing bad ideas.
You seem to be imagining the choice as being between "host bad discussions" or "do something about it via centralized hierarchical control". But I'm trying to emphasize a third option: "do something about it via decentralized mechanisms." (Posts with negative karma are basically invisible, after all.)
The downside of centralized responses is that it creates a precedent for people to use social/political pressure to try to impose their opinions on the whole community. Decentralization protects against that. (I don't so strongly object to the mods just deciding, on their own, to ban certain topics. What especially troubles me is social/political pressure aimed towards this end.)
As I see it, the crucial question to ask is which mechanism is more reliable: top-down control in response to social/political pressure from vocal advocates, or decentralized community judgment via "secret ballot" karma voting. As I see it, the primary difference between the two is that the former is more "politicized" and subject to social desirability bias. (A secondary effect of politicization is to encourage factions to fight over control of this new power.) So I think the decentralized approach is much better.
One important difference between the Forum and most other fora is the strong vote -- a minority faction can use their strong votes to keep things in circulation unless the majority acts the same way. The Forum is also small enough for brigading to be a major concern.
I think encouraging people who take the view to just strong downvote race science material off the Forum poses its own set of epistemic problems for the Forum. And encouraging them to "merely" downvote may not be effective if the other side is employing the strongvotes.
Maybe this proposed solution would be more viable if there were special voting rules for current topics at mod discretion -- e.g., 500 karma required to vote, no strongvotes allowed? I'm not sure, though. If all the race science folks vote and everyone else mostly stays away, the result is a false sense of community views.
Still, I support a ban -- race science is off topic, there are other places people can go if they want to talk about it, and these discussions cause significant harmful effects. If discussions of why the New England Patriots and New York Yankees were a scourge on humanity were causing these problems for the Forum, I'd support a ban even though I believe those teams are. :)
Karma is not a straightforward signal of the value of contributions
This statement and the idea of karma as the decentralized solution to the problems OP describes feels overconfident to me. To reference this comment, I also would push back on karma not being subject to social desirability bias (ex: someone sees a post already has relatively high karma, so they’re more inclined to upvote it knowing that others on the Forum or in the EA community have, even if they, let's say, haven't read the whole post).
I would argue that karma isn’t a straightforward or infallible signal of “bad” or “good” contributions. As those working on the Forum have discussed in the past, karma can overrate certain topics. It can signal interest from a large fraction of the community, or “lowest-common-denominator” posts, rather than the value or quality of a contribution. As a current Forum staff member put it, “the karma system is designed to show people posts which the Forum community judges as valuable for Forum readers.”
I would note, though, that karma also does not straightforwardly represent the opinions of the Forum community as a whole regarding what’s valuable. The recent data from the 2023 EA Forum user survey shows that a raw estimate of 46.5% of those surveyed and a weighted estimate of 70.9% of those surveyed upvoted or downvoted a post or comment. Of 13.7k distinct users in a year, 4.4k of those are distinct commenters, and only 171 are distinct post authors. Engagement across users is “quite unequal,” and a small number of users create an outsized amount of comments, posts, and karma. Weighted upvotes and downvotes also mean that certain users can have more influence on karma than others.
I appreciate the karma system and its values (of which there are several!), and maybe your argument is that more people should vote and contribute to the karma system. I just wanted to point out how karma seems to currently function and the ways it might not directly correlate with value, which brings me to my next point…
Karma seems unlikely to address the concerns the OP describes
Without making a claim for or against the OP’s proposed solutions, I’m unsurprised by their proposal for a centralized approach. One argument against relying on a mechanism like karma, particularly for discussions of race on the Forum, is that it hasn't been a solution for upholding the values or conditions I think the OP is referencing and advocating for (like not losing the potential involvement of people who are alienated by race science, engaging in broader intellectual diversity, and balancing the implications of truth-seeking with other values).
To give an example: I heard from six separate people involved in the EA community that they felt alienated by the discussions around Manifest on the Forum and chose to not engage or participate (and for a few people, that this was close to a last straw for them wanting to remain involved in EA at all). The costs and personal toll for them to engage felt too high, so they didn't add their votes or voices to the discussion. I've heard of this dynamic happening for different race-related discussions on the Forum in the past few years, and I suspect it leads to some perspectives being more represented on the Forum than others (even if they might be more balanced in the EA community or movement as a whole). In these situations, the high karma of some topically related comments or posts in fact seemed to further some of the problems OP describes.
I respect and agree with wanting to maintain a community that values epistemic integrity. Maybe you think that costs incurred by race science discussions on the Forum are not costly enough for the Forum to ban discussion of the topic, which is an argument to be made. I would be curious for what other ideas or proposals you would have for addressing some of the dynamics OP describes, or thoughts on the tradeoffs between allowing/encouraging discussions of race science in EA-funded spaces and the effects that can have on the community or the movement.
Academic freedom is not and has never been meant to protect professors on topics that have no relevance to their discipline: "Teachers are entitled to freedom in the classroom in discussing their subject, but they should be careful not to introduce into their teaching controversial matter which has no relation to their subject. Limitations of academic freedom because of religious or other aims of the institution should be clearly stated in writing at the time of the appointment."
If, say, a philosophy professor wants to express opinions on infanticide, that is covered under academic freedom. If they want to encourage students to drink bleach, saying it is good for their health, that is not covered.
We can and should have a strong standard of academic freedom for relevant, on-topic contributions. But race science is off topic and irrelevant to EA. It's closer to spam. Should the forum have no spam filter and rely on community members to downvote posts as the method of spam control?
You elsewhere link to this post as a "clear example of a post that would be banned under the rules". That post includes the following argument:
The post concludes, "EA's existing taboos are preventing it from answering questions like these, and as new taboos are accepted, the effectiveness of the movement will continue to wain."
You may well judge this to be wrong, as a substantive matter. But I don't understand how anyone could seriously claim that this is "off topic and irrelevant to EA." (The effectiveness of the movement is obviously a matter of relevant concern for EA.) People's tendency to dishonestly smuggle substantive judgments under putatively procedural grounds is precisely why I'm so suspicious of such calls for censorship.
As an Ashkenazi Jew myself, saying "we'd like to make everyone like Ashkenazi Jews" feels just like a mirror image of Nazism that very clearly should not appear on the forum
I'm not making any claims either way about that. I'm just pointing out (contra Matthew) that it is clearly not "irrelevant spam". Your objections are substantive, not procedural. Folks who want to censor views they find offensive should be honest about what they're doing, not pretend that they're just filtering out viagra ads.
I think it is irrelevant, and in every context where I've seen it presented as 'on topic' in EA, the connection between it and any positive impact was simplistic to the point of being imaginary, while at the same time promoting dangerous views - just like in the post you quoted.
Topical relevance is independent of the position one takes on a topic, so the rule you're suggesting also implies that condemnations of race science are spam and should be deleted. (I think I'd be fine with a consistently applied rule of that form. But it's clearly not the OP's position.)