The Accidental Mathematician



This new study in PNAS on gender bias in science hiring is already making rounds everywhere, but in case haven’t seen it yet, here’s part of the abstract:

In a randomized double-blind study (n = 127), science faculty from research-intensive universities rated the application materials of a student—who was randomly assigned either a male or female name—for a laboratory manager position. Faculty participants rated the male applicant as significantly more competent and hireable than the (identical) female applicant. These participants also selected a higher starting salary and offered more career mentoring to the male applicant. The gender of the faculty participants did not affect responses, such that female and male faculty were equally likely to exhibit bias against the female student. Mediation analyses indicated that the female student was less likely to be hired because she was viewed as less competent.

The article is open-access, so you should be able to read it from home; if not, then the summaries here and here have the numbers.

The applicant’s resume was the same in all cases, except for the name (Jennifer or John, two names that were judged by the researchers to be equally “likeable”) and gendered pronouns. (It is available on the PNAS site as part of the “supporting information”.) It is fairly generic, calibrated to show a candidate who looks potentially promising, but not stellar. The candidate mentions research experience with two faculty mentors and a co-authored journal paper, and has plans to apply to graduate school in the future. A few details were added that, depending on the evaluator, might or might not matter: dropped out of a course, slacked off a bit early in college but then got serious and made up for it. (If you asked me, none of these would raise concerns.) Faculty responses were broken down by the respondents’ gender.

On a competence scale from 1 to 7, John received mean scores of 4.01 and 4.1 from male and female faculty respectively. Jennifer’s scores were 3.33 and 3.32, about 0.7 less than John’s. Similar differences were observed in the “hireability” and “mentoring” categories (the latter refers to the faculty member’s willingness to mentor Jennifer or John). The difference in mean salaries recommended for the candidate was also significant: 30,520 and 29,333 for John vs. 27,111 and 25,000 for Jennifer.

There are several important points here. One is that the subjects were actual scientists, doing what we normally do in the course of our work. The resume did not look fake or contrived – it would fit right in with the paperwork I receive all the time as a potential supervisor or member of selection committees. The authors do not study gender bias by proxy, trying to draw a straight line from girls and boys playing with dolls and trucks to faculty composition in top science departments. They test us on behaviours that have direct and immediate impact on women in science, and find us biased.

The “mediation” part is crucial. The scientists were not actively seeking to discriminate against women. They offered similar salaries to candidates that they perceived as equally competent, suggesting that, in their minds, they were evaluating the candidate purely on merit. The problem is that the female candidate was judged to be less competent, evidently for no reason other than gender, given that the resumes were exactly identical except for the name. The unconscious bias was “mediated” into different perceptions of the candidate’s competence.

I’m sure that most of the participants, believing themselves unbiased, would be shocked to see the results. In fact, I’d like to see a web test based on this experiment that deans, department heads, hiring committee members, journal editors, conference organizers and other decision makers would be required to take before assuming their responsibilities. I suspect it could be an eye-opener for many of us.

That the bias is unconscious and involuntary is confirmed by another finding: the female candidate was rated higher on a “likeability” scale. In other words, faculty respondents reported “liking” the female applicant better than the male one, even as they judged her to be less competent, were less willing to hire or mentor her, and recommended a lower salary. It confirms something I’ve believed for years now: it’s a blind alley for women to worry too much about being “liked”.

There’s much more to unpack here, from the responses all over the internet, to the better practices we could adopt in hiring (and elsewhere), to the myriad ways in which we interpret resumes and supplement them with other information, to possible explanations of why female scientists recommended lower salaries overall. It’s good that I’m on sabbatical, because that’s enough material for several posts. They should be forthcoming soon.