A new study that used AI to analyze 2M+ contributions by ~365K developers on GitHub finds users with white-sounding names may have more success on the platform (New Ai 2m Githublapowskyprotocol)

New Ai 2m Githublapowskyprotocol: Is the world of programming a level playing field? White-sounding names may get people further in their careers as developers than others, according to a new paper from UC Berkeley. The study found that developers with white-sounding names were more likely to be hired for project work and had an easier time getting feedback on their code. The findings were confirmed in two field experiments, one of which sent emails with varying degrees of whiteness to job applicants.

Data scientist Jonathan Quick, working with computer science professor Haonan Zhang, studied GitHub data on nearly 2 million GitHub users. The researchers looked at the interactions of 2 million developers between August 2015 and March 2016 to see how they were hired by companies. More than 2 million contributions were made by roughly 365,000 developers with a total of 32m unique names, for a mean score of about 1.4 per developer. The researchers also looked at two different self-reported surveys on GitHub users.

In the first experiment, they identified a sample of 7,500 users who had made more than 2,000 repos and had at least 1,000 stars on an existing repo. A subset of this group was randomly selected for close inspection: 300 developers with more than 2,000 followers and 1m total stars. From this group of 300 contributors, 100 were picked at random for a closer look.

“We show that developers with white-sounding names are more likely to be hired for an in-person interview,” says the report. “We also find that stars and forks from white-sounding users on a repository are significantly more likely to be accepted than stars and forks from users with East Asian names.”

In the second experiment, Quick used a dataset of job applicants’ GitHub profiles, writing samples and resumes submitted to unnamed companies over a six-month period. About 320 applicants were chosen at random and sent emails with varying degrees of whiteness. The resumes received a median score of 3.57 out of 5 in the experiment, with no one getting higher than that.

“We find that white-sounding names are associated with significantly higher callback rates,” says the report. “Recruiters rate white-sounding users as being easier to work with, providing more constructive feedback and providing better references. In contrast, we find no evidence that recruiters discriminate based on gender, with one exception: We show that recruiters appear to discriminate against Asian women.”

For the third experiment, Quick looked at pull requests (PRs) – where developers suggest improvements to another person’s code. Here he found the same results as in the second experiment.

The researcher also looked at what happens when a developer undervalues his or her own work and gets fewer stars from others on GitHub.