Philip Greenspun Reviews “Lean In”

  My basic gripe with Bell-curve arguments against female   
  genius is that at the high end, in the exponential tail, 
  fine-tuning is needed

Well, actually no fine-tuning is needed. The Central Limit Theorem is most valid for a sum process in the middle of the distribution. It converges most slowly in the tails [1,2]; indeed, if your goal is to model the tails you really are dealing with a max-phenomenon rather than a mean-phenomenon.

In this case, the underlying sum process could be a bunch of small alleles of roughly equivalent size, each contributing to high IQ (viz. a QTL model for a multifactorial trait [3]). In that case, if you cared about the tails rather than the body, the discrete chunkiness of the underlying binomial distribution becomes more crucial.

Put another way: Bell Curve arguments (aka "statistical genetics") are at the lowest level based on discrete alleles, not perfect Gaussians. Now, those in the field would love to get better models of the genetics of highly intelligent people -- but to study that kind of thing you need to move to China and work at BGI. Remember, thoughtcriminals who propose genetic explanations for behavioral phenomena are "berated" in the US [See jackowayed's wonderful admission against interest up thread].

  International Mathematics Olympiad.. Notice also the 
  political fine-tuning of the variance argument: it doesn't 
  dare suggest that mean female intelligence is lower 
  (because we'd all reject that). God of the gaps, anyone?

While we're talking about fine-tuning, why do you cherry-pick a few stats without acknowledging that the history of science is male? Do the thousands of male names (Newton, Maxwell, Einstein, Gauss, Euler) that have inscribed their name into history count as a datapoint here? Do the Fields Medals? The Nobel Prizes in Physics? The faculty of math and science departments around the world? The gender of the inventors of the locomotive, the aeroplane, and the automobile? The names of those men who built steam engines and search engines?

I know why. All the conquests and murders in history are counted against men; but is a little odd that every male invention is counted in the demerits column too! I think the idea is that said ancient men ostensibly discriminated against women, shoving them out of the way before they could figure out the value of pi. Without said invidious discrimination women - biologically, neurologically, hormonally, genetically different women - would have been tearing it up on the math tip at the same rate. Just as they have been dunking from the free throw line ever since we started the WNBA.

Bottom line: you can't have it both ways. If achievement in science and engineering is to be a signal, if you are to cite any stats related to IMOs and whatnot, you need to take on board the enormous imbalance in the favor of men on historical measures of sci/eng aptitude and achievement. All due respect to Noether, Daubechies, and Curie -- but the prior probabilities of achievement are not equal.

  If you must hire very good people, and you have the wrong   
  model of how very good people are distributed among the 
  population, you will fail.

That's right. Your statement is: if you have the wrong model, you will fail.

The logically equivalent contrapositive is: if you succeed, you didn't have the wrong model.

So given that Google, Facebook, Twitter, Youtube succeeded with highly male software engineering staffs, logically they did not have the wrong model of how very good people are distributed among the population.

[1] http://math.stackexchange.com/questions/314659/central-limit...

[2] http://www.cs.toronto.edu/~yuvalf/CLT.pdf (see page 2)

[3] http://en.wikipedia.org/wiki/Quantitative_trait_locus#Multif...