Gender bias affects contributions to the open source community, according to a paper published Monday in the open access journal PeerJ Computer Science.
Female programmers’ suggestions for code changes in open source projects — called “pull requests” — were accepted more often than those of their male counterparts — 78.7 percent for women versus 74.6 percent for men — when gender was unspecified. However, that changed when the gender of a pull request’s author could be identified.
Authors who could be identified by name or a profile picture as women had lower pull request acceptance rates (58 percent) than those who could be identified as men (61 percent).
On the other hand, women with gender-neutral profiles had higher acceptance rates (70 percent) than any other group, including men with gender-neutral profiles (65 percent).
“Our explanation for why all women were doing better than women who could be identified as women is gender bias,” said Emerson Murphy-Hill, an associate professor of computer science at North Carolina State University.
Murphy-Hill authored the study with Josh Terrell, a former undergraduate at Cal Poly; Andrew Kofink, a former undergraduate at NC State; Justin Middleton, a Ph.D. student at NC State; Clarissa Rainear, an undergraduate at NC State; Chris Parnin, an assistant professor of computer science at NC State; and Jon Stallings, as assistant professor of statistics at NC State.
Higher Bar for Women
The study’s findings, which were published in preliminary form last year, are based on analysis of 3 million pull requests from 330,000 GitHub users, including 21,000 women. GitHub is an online programming community that fosters collaboration on open source software projects.
Despite their being vastly outnumbered on GitHub, women overall performed better than men on pull requests.
“Our intuition is that the women who are on GitHub — if they’ve made it to that point in their careers — they may be more competent to the average man in that position,” Murphy-Hill told LinuxInsider.
“Prior research has shown that women in the workplace are held to higher performance standards,” he pointed out.
By the time women reach the point in their careers where they’re contributing to open source projects, most of the less competent women have exited the career path, Murphy-Hill maintained.
Assessed with Different Lens
The findings by the NC State researchers aren’t applicable to women programmers alone, noted Barbara Annis, a founding partner in the Gender Intelligence Group.
“When women are identified as women, their ideas tend to get dismissed more, not only in the tech industry but in other industries as well,” she told LinuxInsider.
“We see women in organizations being assessed with a different lens,” she added.
For example, in an experiment the Gender Intelligence Group performed at a financial services company, one group of women had gender identifiers on their resume while another did not. Of the gender-identified women, only 18 percent received job interviews. However, 71 percent of the non-gender identified got interviews.
“They still didn’t get the job, because in the interview you could see the candidate was a woman,” Annis said.
“It can be worse in the programming field,” she added, “because the culture and mindset there is still very male.”
What might be the next step for the researchers? A look at the people judging pull requests could be enlightening, Murphy-Hill noted.
“Prior work suggests that women tend to be especially hard on other women,” he said.
“When some people read this paper, they think men are the villains here,” Murphy-Hill continued. “I don’t think that’s the way to read the paper at all. I think everybody has implicit biases, whether they’re men or women, and biases can manifest no matter who you are.”
Another possibility is that the original research may have focused on the wrong issue, suggested Jennifer Bryan, an associate professor at the University of British Columbia who teaches data analysis and statistics.
“I thought the story should have been much more about how few contributions were coming from women,” she told LinuxInsider, “than the rather modest differences in acceptance rates.”