I was dismayed by Bostrom’s racist email, more so by his apology, but most of all by many EAs' reactions to both. Based on defenses of Bostrom’s apology here and on Twitter, a substantial number of EAs seem to think it was fine that Bostrom failed to disavow the racist claims made in his original email. (“Are there any genetic contributors to differences between groups in cognitive abilities? It is not my area of expertise…”)
My goal is not to make an important conversation about racism into one about sexism, but to instead suggest that Bostrom's words and the community's response have broader implications.
People who believe there are innate links between race and IQ may also believe there are innate links between sex and IQ. I first learned about the variability hypothesis—i.e., the idea that females display less variability in cognitive ability than males—from a male EA. Unlike the claim that “some races are smarter than others because of genetics,” there is data to suggest the variability hypothesis may be true in some places and for certain kinds of intelligence, though this hypothesis has often been leveraged in sexist ways.
If many EAs buy into the variability hypothesis, then when I am in the EA spaces I run in—which are full of men who graduated from elite universities—some may assume I am less smart than they are not just because they are smart, but also because I am a woman. For instance, if they’re at the 97th percentile of intelligence, rather than assuming there’s a 3% chance any given person is as smart as they are, many will employ higher priors for men than for women.
Women become attuned to which men are employing what priors about their intelligence because it matters when deciding who to befriend, date, collaborate with, be mentored by, and so on. It feels different to talk to a man who is open to the possibility that you’re as smart as he is and one who has all but ruled this out. (Imagine asking your parent for something as a kid—you could generally tell the difference between a “no” where the door wasn’t totally shut, and one where it was.)
I have had my intelligence underestimated both within and outside of EA. But three things make intelligence-related sexism particularly insidious within EA. First, EA is heavily male. As a result, I am more likely to experience sexism in EA spaces than in environments where women are better represented. Second, the discourse around Bostrom’s email has convinced me that it is more socially acceptable within EA to harbor certain racist or sexist beliefs (as compared to, e.g., academia), so EAs who do may not work very hard to think—or act—differently. Finally, many EAs place a high premium on certain, narrowly defined forms of intelligence. In other spaces, while I might at least benefit from men thinking: “I’m smarter than you, but you probably bring other important stuff to the table,” in EA spaces, it often feels more like: “I’m smarter than you, and that’s primarily what matters.”
If something causes me to leave EA entirely, it may well be sexism. I simply don’t have to do as much extra work to prove myself in other spaces (including ones where people are every bit as smart), and having to do this work saps me of my energy, confidence, and time; in short, of my ability to be an effective altruist.
This was a powerful paragraph, thank you for sharing this.
I'm sorry to hear that you've experienced sexism both within and outside EA.
Just to clarify your view, you said that:
But an implication of the hypothesis is that men will make up a greater proportion of the "intelligent" people in those places for those kinds of intelligence.
Do you think it would be fine to use this information as a prior in those contexts?
Even if there are cases in which it would theoretically be reasonable to employ different priors for men vs. women, I doubt people will be able to reliably identify these cases, choose appropriate priors, and correctly apply the priors they've chosen. When you couple these challenges with the fact that there are significant downsides associated with trying to discriminate in a principled way (e.g., harming people, alienating people, creating self-fulfilling prophesies, making it harder for members of an already disadvantaged group to succeed, etc), it seems like a bad idea to base priors on the variability hypothesis in basically any context.
IMO, I agree with this. I think that a lot of the defenses of Bostrom's comment are bad, and I don't really like many of the defenses shown here.