[1] http://www.psychologytoday.com/blog/between-the-lines/201204...
1. Isn't a woman, Jessica Livingston (Paul Graham's wife) on the interviewer panel and a part of the application process? It's not just Paul Graham himself.
I'm quick to defend Paul in the same way others are quick to blame him. It seems we as a diverse society are so conditioned to enforcing equal extremism that any time we don't see an industry, a workforce, or a group equally divided between male/female, black/white, gay/straight we immediately sound off the alarm and go on a witch hunt. All of this without considering that certain groups of people are better at something than others. For instance, African Americans make up only 12-14% of the population but over 60% of the NFL. Jewish people make up less than 6% of the population yet they make up almost 100% of entertainment industry executives (see Joel Stein article in the New York Times if you don't believe me). We hold up the majority to a level of standards that the minority cannot even reach. There's this stigma that if you have nice things, you cheated to get them, didn't earn them, and must divide them and share them with everyone else or else you are sexist/racist.
The solution isn't bias in the other direction, but to look for ways to remove the bias. This is why in science we have things like double-blind studies, for example. In music, doing auditions behind a screen seems to have been effective.
Putting systematic measures in place against bias also tends to help with self-selection, since it assures applicants that they have a fair shot. I believe that's what the original poster was asking for. I don't know what the best solution is for something like Y Combinator, but it seems worth giving it some thought. Of course, it's not going to be so easy as performing music behind a screen.
The percentages you cite show this is a problem in many industries. I doubt that 50% is achievable, but I also don't think it's helpful to either say "these people are sexist" or "yeah, but everyone does it." Those are both examples of moralistic thinking. The solution is to move beyond that sort of thing and treat this as a problem to be solved.
We all have some biases, and taking reasonable efforts to mitigate them has worked very well in other fields, the typical example being the screen for orchestra auditions.
One interesting place to start is Harvard's "Project Implicit". They have a massive publication list[1] and you can even test your own implicit reactions[2].
There are plenty of other scientists testing things like whether people judge women as less competent. A quick google search pulled up a PNAS paper where they did an experiment on women in science, for example.[3]
This is just the tip of the iceberg, of course. There is a whole host of related work, testing other sorts of biases and using other methodologies. I'd suggest a search on your favorite academic search engine for "implicit bias".
[1] https://www.projectimplicit.net/papers.html [2] https://implicit.harvard.edu/implicit/ [3] http://www.pnas.org/content/early/2012/09/14/1211286109#aff-...
