Following up on a comment by Diane Greco about her experience at MIT [no permalink; search for “MIT's New President”], Hanna Wallach at join-the-dots asks about the absence of women in theoretical machine learning.
Diane Greco: Even now, it seems to me that the dissertations produced in my department were theoretically interesting to the extent that they were written by men, and merely fussy and pedantic when they were written by women, myself included. It was as if we were afraid to be geniuses, preferring to hide behind large data sets just like I hid behind my hat.Hanna Wallach: Are women shying away from theoretical machine learning and “hiding behind large data sets” in areas such as computational linguistics, bioinformatics, computer vision? I also wonder if it is that women prefer to engage in computing with a purpose, to focus on an application for those abstract and theoretical techniques.... With so few women undertaking research in fields like machine learning—theoretical or applied—it’s rather dangerous to conclude anything in either direction, however I do think that when reading Diane’s comment, it’s worth considering that women may simply find enjoyment and perceive greater utility in more applied work.
Huh? Is this actually true? Gender imbalance in computer science is a profound embarassment, but is it really worse in more theoretical areas? Several of the strongest computational geometry and algorithms researchers are women, even at MIT. Sure, some of them are clearly application-driven, but others are unabashedly theoretical. And there are more women entering the field every year. Even the few women I know in computer vision aren't exactly “hiding behind large data sets”, as Hanna puts it.
What's going on? Is my testosterone-addled memory that inaccurate? Do women really prefer applications to theory?
From my own experiences in math, I see a few factors.
1. Undergraduate experiences. Based on my recollections of people who happen to come to mind, many of the men started grad school in a stronger position. I was fortunate to have attended an excellent undergraduate school, but most of the women in my program came in with weaker backgrounds; they were at a decided disadvantage at the beginning of graduate school. Do the best faculty want to take on the weakest grad students as advisees? What kind of project would you suggest for a struggling grad student? I had the same struggles with the material -- just a few years previously.
2. Risk-taking. When pure-math work is going badly, there is nothing to publish and nothing to talk about. The theorem could turn out to be wrong or someone else could prove it first. More applied work seems to be a bit more steady work and a bit safer (she says without knowing anything about applied work). This factor comes up in different ways for different people.
3. Available mentors. Several women in my graduate program chose their research area because one faculty member is especially well-known for being a good advisor to women. A limiting tradeoff.
Posted by: Rudbeckia Hirta | September 27, 2004 at 04:31 PM