Monday, September 15, 2014

Gender Bias in Guest Observer Programs

A recent paper on arXiv (1409.3528) by I. Neill Reid at STScI is an eye-opener.  It presents a detailed study of gender bias in HST guest observer selection.  The results are very clear: female PIs are systematically less successful in winning HST observing time than their male colleagues.

The HST review process has been carefully designed and tweaked over the years to be fair.  Conflicts of interest and competitive bias are dealt with by having proposals judged in panels that do not have members involved in those proposals.  Institutional conflicts of interest are also guarded against.  Gender bias is much harder to deal with, particularly if it is in a form of unconscious bias.  The first step in addressing any such bias is to determine if it exists and to what extent.  That is the purpose of the study.

The study covers HST Cycles 11 through 20 from 2001 to 2012.  The number of proposals submitted in those cycles was 9400 and number accepted was ~2100.  Since proposers were not required to give their gender, there was worked needed in the study to determine the gender of the PI for each proposal.  This was done by first name identification and web searches where necessary.

One of the primary results of the study is shown in the figure below.  The success rate of male PIs is seen to be higher than that of female PIs in every cycle.  The overall success rate for male PIs is 23.5% compared with 18.1% for female PIs.

One question that obviously comes up is what the gender diversity of the Time Allocation Committee diversity was.  There was an effort made by STScI to increase the fraction of women on the panels which resulted in a factor of more than two increase from Cycle 11 to 20, from 18% to 50%.  Interestingly, this did not result in a noticeable change in female PI success rate relative to male PIs.

The paper points out that previous studies have found that "unconscious" or "implicit" bias can be a significant factor in the scientific community.   Scientists participating in peer review are instructed to be unbiased and, as whole, try hard to achieve that.  Still peer review is a subjective process with personal judgment required and unconscious factor can enter in.  This may well be the cause of the HST results. 

One result of the study is that reviewers are now made aware of the overall lower success rate of female PIs in past reviews at the pre-review briefing.  Also, unconscious bias is discussed with them.  It is not clear yet if this will cause in changes in the results.  A welcome trend is that there is an ever increasing number of proposals submitted by female PIs, from 19% in cycle 11 to 26% in cycle 21.