The researchers, who published their findings earlier this week looked at the behavior of software developers on GitHub, one of the largest open-source software communities in the world.

Based in San Francisco, GitHub is a giant repository of code used by over 12 million people. Software developers on GitHub can collaborate on projects, scrutinise each other’s work, and suggest improvements or solutions to problems. When a developer writes code for someone else’s project, it’s called a “pull request”. The owner of the code can then decide whether or not to accept to proffered code.

The researchers looked at approximately 3m pull requests submitted on GitHub, and found that code written by women was approved at a higher rate (78.6%) than code written by men (74.6%).

There are two implications from this study:

1. Women write better code because their code is accepted more if gender is unknown.

2. There is community-wide bias against accepting the code if gender is known.

Number 2 is unsettling because it speaks to a community bias.

Number 2 is also being questioned -- see below.

Top Reddit comment:

This conclusion was actually reached through very questionable methods (see the original preprint).

• The original hypothesis was that women's contributions would be rejected more often. It was not supported by the data.
• Then the researchers started looking for some way to justify their initial hypothesis in the face of this data.
• After multiple failed attempts they discovered that in one specific situation (if the gender is obvious and the woman is not a long-term contributor to the project) women's contributions appear to be rejected more often. Note that they used a heuristic to check whether the gender is obvious or not.
• They immediately assumed without any further research that it is due to gender bias. They didn't even consider what other factors could be at play, as they did when their initial results appeared to disprove their hypothesis.
• They didn't try to explain in any way why contributions are always rejected more often if the gender is obvious — whether male or female. Yet their graph clearly shows it.

Given such flaws, I don't think it's a good idea to promote this preprint as it is in any case.

Source: https://reddit.com/r/science/comments/45bq1q/data_analysis_of_github_contributions_reveals/

This may be taken as respected research or not. What is most illuminating, and vital, is that this shows the need to scrutinize in-depth any 'Research', 'Studies Show', ad infinitum. And to also delve into the bona fides of any outside people/sources who applaud or refute the findings. Not to be a cynic, but a realist. Thank you for pointing this out, Adam.

You're welcome, Marlene.

I think it's good that the research sparked the discussion.

"Gender Bias in Open Source: Pull Request Acceptance of Women Versus Men"

Josh Terrell1, Andrew Kofink2, Justin Middleton2, Clarissa Rainear2, Emerson Murphy-Hill2∗, Chris Parnin2 

1Department of Computer Science, Cal Poly, San Luis Obispo, USA; 2Department of Computer Science, North Carolina State University, USA 

https://peerj.com/preprints/1733v1.pdf

Thank you for posting the source.

From the ArsTechnica article:

Source: http://arstechnica.co.uk/information-tec...