A paper presented the outcome of a large scale study on gender bias, by comparing the acceptance rates of contributions from men versus women in an open source software community.  

The paper, Gender Differences and Bias in Open Source: Pull Request Acceptance of Women Versus Men, found that when contributors could be identified as men, their contributions were accepted more. The reverse was the case when the gender of the contributor could not be ascertained.

Overall, the results suggested that although women on GitHub may be more competent overall, bias against them exists nonetheless.

The paper presents an investigation of gender bias in open source by studying how software developers respond to pull requests, proposed changes to a software project’s code, documentation, or other resources.

The researchers investigated whether pull requests are accepted at different rates for self-identified women compared to self-identified men. The researchers chose to study GitHub because it is the largest, claiming to have over 12 million collaborators across 31 million software repositories.

The researchers hypothesized that pull requests made by women are less likely to be accepted than those made by men. Prior work on gender bias in hiring, where a job application with a woman’s name is evaluated less favorably than the same application with a man’s name, suggests that this hypothesis may be true.

Results showed that women’s pull requests tend to be accepted more often than men’s, yet women’s acceptance rates are higher only when they are not identifiable as women. In the context of existing theories of gender in the workplace, plausible explanations include the presence of gender bias in open source, survivorship and self-selection bias, and women being held to higher performance standards, the researchers concluded.

 

 

Referenced:

Terrell J, Kofink A, Middleton J, Rainear C, Murphy-Hill E, Parnin C, Stallings J. (2017) Gender differences and bias in open source: pull request acceptance of women versus men. PeerJ Computer Science 3:e111 https://doi.org/10.7717/peerj-cs.111