> If women have the same potential both to succeed at computer science and to enjoy the work as men do, the underrepresentation of women in the field is a sign of problems that, if fixed, would result in more people leading more fulfilling lives. I think that is worth spending money on, and Google seems to agree.
If some _individuals_ have a potential to succeed at computer science and are currently not doing so, then I agree, helping them to do so would lead to people living more fulfilled lives. But such individuals should be helped equally regardless of such factors as gender, race, nationality, sexuality, etc.
If it happens to be the case that most such individuals are women, then any well-chosen but fair (i.e. genderblind) criteria will end up helping mostly women anyway.
If those assumptions about the underrepresentation of women in the field hold, then it's much easier to identify women for whom computer science would be a good fit but haven't chosen it than it is to identify similar men. Paying the costs for any qualified college student to attend JSConf would be less effective than paying the costs for qualified women.
Can you suggest gender-blind criteria that would be successful? Discriminating less while achieving the same goal is an unalloyed good.
> Can you suggest gender-blind criteria that would be successful?
Well, the original formulation was "potential to succeed at computer science". So we need to measure potential (P) and success (S). Then the people we are interested in satisfy (P and not S).
S could be defined as being a professor of computer science (or related discipline), or as earning more than a certain amount in a computing-related field.
P could be determined by getting someone to submit a computing-based project they've worked on; the submission might consist of a tarball containing documentation and source code, which would obviously have to be assessed according to subjective criteria (to avoid untoward discrimination at this stage, the submissions might be anonymised so reviewers wouldn't know the name of the submitter.
Or P could be determined by some objectively-assessible criteria, for example asking people to write a program whose performance would be measured objectively (e.g. the Google AI Challenge -- http://csclub.uwaterloo.ca/contest/index.php )
But however you determine S and P, the set of people you're interested in falls naturally from that.
If some _individuals_ have a potential to succeed at computer science and are currently not doing so, then I agree, helping them to do so would lead to people living more fulfilled lives. But such individuals should be helped equally regardless of such factors as gender, race, nationality, sexuality, etc.
If it happens to be the case that most such individuals are women, then any well-chosen but fair (i.e. genderblind) criteria will end up helping mostly women anyway.