Amid all the angst about the small percentage of women who work in computer science, a new research paper has a surprising finding: Women may actually be better coders than men. The problem is that women's code is viewed much more favorably when people don't know that a woman wrote it.
The researchers, both computer science professors, looked at so-called pull requests, or proposed changes to a software project's code, submitted on GitHub, a Web-based repository of open-source code. Some 78.6 percent of pull requests initiated by women were accepted, compared with 74.6 percent of those initiated by men.
From this data, the researchers concluded that women were suggesting more useful changes to the code than men were. But first, they looked for other reasons that women's code was more likely to be accepted. Was it addressing more urgent problems, maybe? Or were the women's changes smaller or easier to implement, involving fewer lines of code? Neither of these proved to be true, and in fact, the changes suggested by women, on average, involved significantly more lines of code than those of men.
The researchers then split the coders into groups. First they separated those whose gender could pretty easily be figured out from their identities on GitHub from those whose gender was more hidden. They also classified each person as either an "insider" on a particular project, and whose gender was probably known to others on the project, or an "outsider," about whom less was probably known.
Within the realm of insiders, the acceptance rates of women's pull requests was about the same regardless of whether their gender could be determined from their profile. But it was different when the requests came from outsiders. In those cases, women had acceptance rates of 71.8 percent when they used gender-neutral profiles. Their acceptance rates dropped to 62.5 percent when it was clear that they were female.
"Women have a higher acceptance rate of pull requests overall," write the researchers, "but when they're outsiders and their gender is identifiable, they have a lower rate than men."
The paper has not yet been peer-reviewed.
What GitHub has in common with orchestras.
This is not the first study to find bias against the work of women. In one, scholars were asked to read and rate research papers; unbeknown to them, the names had been changed to change the gender of the authors, and the scholars rated the papers "written" by men as better than the ones that appeared to be authored by women. Another found that women tend to receive better job performance ratings than men, yet are less likely to be considered for promotion.
Professor Ben Barres, a neurobiologist at Stanford, recalls hearing that his work was "so much better than his sister's." But Barres is transgender; his alleged sister, Barbara, is simply Ben's earlier identity.
But perhaps the GitHub study has even more in common with an experiment conducted far outside the world of computer science. In the 1970s and '80s, fewer than 5 percent of musicians in the top five symphonies were female. Today, the number stands closer to 25 percent. Of course, there have been many cultural changes over that time. But the implementation of gender-blind auditions has also played a huge role.
Many auditions for orchestras are now conducted behind a screen, which was not the case in the 1970s. So the people listening, and judging, have no way to know if the musician is a man or a woman. The screen increases a woman's chances of making it past the preliminary rounds by about 50 percent. It also greatly helps a woman's chances of winning in the final round.
Perhaps orchestras are not the only ones that could benefit from a more gender-blind approach. As the researchers of the GitHub study write, "The trends observed in this study are troubling. The frequent refrain that open source is a pure meritocracy must be re-examined."