In the face of a gigantic participation gap, I am not persuaded by numerology. I don't find the data you have on this issue to be particularly compelling. I think it relies primarily on a vain misconception a lot of people have about the demands on software developers.

I think you find it persuasive because you work in a small specialty of CS that is particularly math-oriented (even if I don't buy that your facts apply to your particular specialty).

So in the face of weak but suggestive data, we should instead accept an alternate hypothesis with a single data point vaguely alluded to?

"Single data point vaguely alluded to". That's funny. Women compromise 20% of the software development workforce. For every woman developer, there are four men. Against that fact, you've got statistical studies on the variance of quantitative ability between men and women which purport to explain why women don't chair math departments and win Fields Medals, and which you'd also like to use to explain why there are four men writing SQL queries to every woman.

Just curious, what evidence (if any) would cause you to change your views?

Also, given that you keep bringing up CS as somehow unique, does this mean you attribute other STEM fields without women (e.g. physics, aerospace engineering) to a different cause?

Those fields are tiny compared to software development, so I guess I'm willing to believe (a) that there isn't enough social impetus to correct them and (b) whatever biased CS is probably very similar to what biased those STEM fields that have the same problem as CS.

Results from unbiased work-sample aptitude tests would influence my views. But they wouldn't influence the views of the "pro-gap" faction; I expect the results would be explained away as "well of course that cohort of women captured by your sample does well on the test; it's all the other women that have the problem."

Anyways, I'd pay attention to that, because: the biggest fallacy I see in arguments that suggest the gender gap is benign is the notion that it attributable to intrinsic differences in intellectual attributes between males and females; when people attempt to explain those differences, they appear not to connect to software development in any but the most naive, high-school guidance counsellor kinds of ways.

My sensitivity to that fallacy has, like, the inverse vector of what I think yours is: I started my career with a huge deficit in math (my worst subject in high school, which is where my formal education stopped), but was successful regardless, and not because I plugged form fields into databases: after doing somewhat interesting systems programming for an ISP as a network admin, my first real dev job was compiler-theoretic. So I'm automatically suspicious of the belief that software development is closely tied to math aptitude.

Your claim (b) confuses me. First you try to paint CS as somehow special and requiring no quant skills, unlike the rest of STEM. Thus, the well supported "quant skill -> less women" correlation doesn't apply. But then you try to suggest other STEM fields do have the same problem as tech.

As for why intellect gap arguments persuade you and not me, I think I understand now. I tend to believe that there are very few independent dimensions in human intellect, you seem to think there are many. I'd suggest investigating psychometrics in a bit more detail - I found this idea very surprising, but it holds up quite well. Take tokenadult's linkdumps on hiring procedures as a good starting point.

Also, let me emphasize that I don't claim my theory is perfect - I just claim it's the best supported. Quantitatively, all the correlations point in the right direction. I see no reason to exclude other theories, e.g., Sapna Cheryan's "women avoid nerds" theory ( http://faculty.washington.edu/scheryan/research.htm ) - the correlations at least line up there, admittedly anecdotally (doctors and lawyers are high status, computer and math geeks are low status).

The idea that CS, math and physics are somehow uniquely unkind to women seems weak to me. I see no plausible reason why "bog-standard human behavior" would be worse in STEM - if anything, you'd expect objective fields to care less about human behavior. (That's probably why computing is so tolerant of weirdos.) Note that you haven't even provided anecdotal evidence that computing is somehow worse than law in this regard.