Contra Grant On Exaggerated Differences
Briefly:
1) Scott’s dismissal of the Hyde meta-analysis is unfounded. He claims it lumps a bunch of common interesting claims with a bunch of nothing claims that find nothing. But here are some of the things Hyde found small or no difference in:
Math (several studies), science, aggression (one study out of several), non-physical aggression (several studies), and leadership (several studies). All of these are things for which any person on the street could identify the stereotype and which way it’s supposed to point, and probably at least some would still believe it.
As an additional nitpick, when Scott lists things Hyde “found gender differences in” he lists things even if they were only found in one of several studies in the meta-analysis. If all of the things he listed were found in every study that studied them, there would be no way of getting near the headline 78% figure.
2) Scott is making a very strong assertion here based on casual reading of charts. Closer reading of the same charts shows that Scott’s reading is specious.
For example, Sweden is generally agreed to be less sexist than Japan and indeed has more female CS students (29% to 20%). Portugal, which would be fairly typical in sexism for Europe, does exceptionally well in gender balance (49%). Italy, similarly not the most sexist country in Europe, is doing pretty okay (42%). And if you look closer you find some VERY strange things, like how Yemen’s rate is 59% but Oman’s, right next door with a similar culture, is 26%.
Furthermore: if you look at the study Scott cites about Big Five personality traits, there are some obvious problems with it (it’s got a table which is the epitome of fishing for p-values; seriously, sex ratio in smoking?). It also doesn’t seem to have been replicated at all as far as I can tell from Google, although of course I could just be unable to find a replication.
There’s also the more obvious problem that it’s actually totally meaningless in this context. What would that have to do with this? You can’t just wiggle your eyebrows suggestively in the direction of a link and claim you’re done, you would need to actually prove it, or at least provide evidence for it.
3) The section is fine as far as the narrow claim about Harvey Mudd, but I can make a fairly strong argument that the broader claim is false.
Grant himself puts in a very suggestive graph which Scott doesn’t bother to argue with. If you look at the social context you find that drop happens about when the social connotations of computer science change from “secretarial work” to “math”. You can see in this famous Cosmo article that women were being held up not just as equally capable computer programmers as men but as more capable in the 1960s. (There’s a different explanation here, which could also be the case but which I find more likely to be an offshoot of the underlying problem.)
4) There’s a lot of bluster and repetitions of long-debunked arguments like “but you feminists never care about getting women into coal mining!” in this section that I’m not going to get into specifically.
A lot of the strength of the argument is debunked by that same graph in the same article which Scott must have already seen. It peaks along with every other job and then decreases. If it was an innate aptitude thing you’d expect it to hit a ceiling and then stay roughly level over time, not decrease.
That’s a strong sign of shifting social norms, which Scott never considers. He just assumes that all social norms get more permissive over time. Of course, if you look at the evidence it’s pretty clear that in fact, Cosmo was writing articles about how many women there were in CS and how it was like cooking dinner in the 60s, and that seems pretty absurd now.
A lot of this is also Scott, in somewhat typical Scott fashion, misapplying studies. For example, using Shashanni 1997 to prove that women aren’t responding to negative stereotypes is like using the many repeated studies which prove that most Americans are unwilling to endorse racist beliefs to a survey taker to prove that racism is dead. In fact, if you read it closely, you find it actually shows almost exactly the opposite: the female students in that study were less likely to have used a computer, to have a computer at home, to say they were confident about using a computer, and to say they liked using a computer. The study also found that liking and confidence decreased for the female students (increased for male students) if they perceived at least one of their parents thought computers were more appropriate for boys, and increased for both male and female students if they got encouragement from their parents. The only thing that study found was equal was a belief in the broad equality of men and women with computers, which Scott cherry-picks to say that stereotypes don’t matter when in fact the whole point of the study is that they do!
Sidenote on the hormone stuff: Scott, and other LWers, and the trans community, have all observed that trans women are disproportionately likely to be programmers, particularly relative to cis women. This seems to me to be pretty strong evidence that it’s not hormones (which trans women share with cis women) and is socialization (which they don’t).
5) Trying to have a debate about the science, despite me debating the science, is not something I think is useful here. Science should be debated by scientists. I don’t want to debate some random Google employee about psychology; that just produces a lot of bad psychology. Making very strong claims like the ones in question about shaky science is never a good look and generally ends up in a giant quagmire of an argument such as this one.
