An interesting article, but I was baffled by this line:
'The bigger mistake in Damore’s memo is one I see frequently: Assuming that job skills and performance can be deduced from differences among demographic groups.'
From my understanding, Damore does not argue this. Instead, he argues that job skill and performances in groups might be deduced from differences in demographic groups. I don't recall Damore every claiming that differences in a group make an individual less skilled.
I think this confusion arises from a misunderstanding of Damore's argument flow, which goes something like this:
1. Women as a group tend to be interested in things other than programming, so there are fewer female programmers overall relative to their population.
2. Because there are fewer female programmers, there are fewer female programmers applying to google.
3. The distribution of skills in the population of female programmers that apply to google is identical to the distribution of skills to the male programmers that apply to Google. (This is an important assumption that we don't have a lot of data for)
4. Google wishes to hire women in proportion to their representation in the general population (looots of anecdotal support for this point).
Conclusion: Google hires a larger percentage of their female applicant than their male applicants. Because of point 3, they will end up hiring more low skilled females than males[1].
You'll notice that a lot hinges on point 3. There is some data to suggest that women self select much more when applying to a job than men do, thus it's possible that the distribution of female engineers applying to google is either more high skilled, or the distribution is right shifted, than the distribution of skills among male engineers. If this shift was big enough, google would not require special considerations in order to hire a larger percentage of female, than male applicants. Since Google's discriminatory hiring practices are one of the worst kept secrets in the tech industry, it's unlikely that the skill distribution among female applicants is that skewed, though there is very likely some skew.
Point 1 is interesting because it suggests that the fundamental problem is not Google's fault so it can't be fixed by Google's hiring or retention practices. Even if point 1 is actually "Women as a group are discouraged from being programmers therefore there are fewer female programmers", it still puts the problem squarely outside Google's hiring and retention practices.
[1]The actual error being made here is much more complex. Since hiring is an error prone process, a certain percentage of unsuitable candidates will be hired. Since there are so many more men hired by Google than women, even lacking any positive discrimination in favor of women, there will be far more low skilled men hired than low skilled women. What would actually happen is that the error rate for hiring low skilled candidates would be much higher for women than for men. Absent any other intervention, this could lead to a lower retention rate for female hires as managers realize that their skills are not up to par. Thus Google also introduces a plethora of retention programs designed to raise the skill level of the female engineers.
Note that these low skilled female hires are not low skilled because they are women, but they are low skilled because Google was less stringent while hiring them. This argument is not sufficient to say if a particular female engineered is low skilled, only that if you were to evaluate and group the skills of all the female and male developers, the female skill distribution would have a larger left tail than the male left tail. Thus for a given female engineer, they are more likely to be low skilled, despite the fact that you will almost certainly meet more low skilled male engineers.