Megan Greischar led a discussion on some of the literature concerning implicit bias. Here is her summary:
I discussed different ways implicit bias is tested for in published literature. Williams & Ceci 2011 PNAS find no consistent pattern when comparing the percentage of PhDs held by females and the percentage of female hires for tenure track positions and infer that there is no systematic bias, and we discussed when these percentages could be misleading (e.g., when departments are growing at different rates). Thomas et al. 2015 PLoS ONE instead model demographic changes in faculty numbers (rather than percentages) and concluding that both the hiring and retention processes must be equitable to achieve parity.
Moss-Racusin et al. 2012 PNAS sent identical CVs for a lab manager positions and found that both male and female faculty ranked male applicants as more competent, hireable, and deserving of mentorship than female candidates. Faculty believed they were ranking real candidates who wished to obtain feedback on their applications. Using a different approach, van Dijk et al. 2014 Current Biology examined the probability of becoming a principal investigator, finding that being male significantly increased the odds of becoming a PI given the same publication record.
Williams & Ceci 2015 PNAS conclude that current faculty (male and female) show a 2:1 preference for hiring female candidates for tenure track positions. Their study differs from previous work in that the faculty knew they were judging hypothetical candidates (“Drs. X, Y and Z”). They were also unambiguously strong applications, which might be expected to reduce bias of research into racial bias (Ginther et al. 2011 Science). Williams & Ceci’s study design reduces bias from gendered language (and they do not examine the effect of gendered language in this study). Haynes & Sweedler 2015 Analytical Chemistry highlight these issues in their response to the Williams & Ceci study.
The subsequent discussion focused on how to detect and address bias. We explored a range of reasons why new hires might be perceived (or perceive themselves) to be less qualified than other candidates. The problem of perceived differences in quality may be especially severe for spousal hires and individuals hired as part of explicit efforts to increase diversity, even when those hires are clearly productive and influential scientists in their own right. We discussed how to deal with those perceived differences.