Pull Requests Are Accepted At About The Same Rate, Regardless of Gender (techinasia.com)

An anonymous reader writes: Remember that story about how women "get pull requests accepted more (except when you know they're women)." The study actually showed that men also had their code accepted more often when their gender wasn't known, according to Tech In Asia -- and more importantly, the lower acceptance rates (for both men and women) applied mostly to code submitters from outside the GitHub community. "Among insiders, there's no evidence of discrimination against women. In fact, the reverse is true: women who are on the inside and whose genders are easy to discern get more of their code approved, and to a statistically significant degree."

Eight months after the story ran, the BBC finally re-wrote their original headline ("Women write better code, study suggests") and added the crucial detail that acceptance rates for women fell "if they were not regulars on the service and were identified by their gender."

Comments

Relationship

[Anonymous Coward]

Since both men and women have their code accepted at higher rates when you know their gender, I wonder if there is a relationship between knowing more about a person and accepting their code. Does knowing someone better mean you are more accepting of their work? If Beth is a working mother of 5 and you know this, does that knowledge make you more or less likely to approve of her code opposed to only knowing that someone made a merge request?

Re:Relationship

[dgatwood]

Actually, it's more like an uncanny valley. The first part was that for people outside the community (where you don't really know much about them), if you know nothing about the submitter, you're more likely to accept the code than if you know just a little bit about the submitter. But if you actually know them IRL, you're more likely to accept their code.

Re:Relationship

[Mr D from 63]

Since both men and women have their code accepted at higher rates when you know their gender, I wonder if there is a relationship between knowing more about a person and accepting their code. Does knowing someone better mean you are more accepting of their work? If Beth is a working mother of 5 and you know this, does that knowledge make you more or less likely to approve of her code opposed to only knowing that someone made a merge request?

I don't know, but surely there is something we can be outraged about. We must look harder.

Re:Relationship

[AmiMoJo]

And that is exactly what you would expect. People tend to be open minded when they don't know anything about a person. Once they know enough to categorize them, all the biases creep in. Then as they get to know them, they start to see them as individuals again.

Re:Relationship

[lgw]

It must be raining frogs: I agree with AmiMoJo. This is how the human mind works - we are pattern engines. We fit what we see to the patterns we've internalized, even when we see very little. Usually, that works quite well: even our vision is mostly synthesized from minimal data (other than the very center of our field of view, the rest is mostly fake detail), but it works so well we don't notice. But sometimes it doesn't quite work out, from biases that aren't accurate to optical illusions. Helps to be aware of it all.

Re:Relationship

[ShanghaiBill]

Why are these studies even being funded?

The original crappy study, that TFA debunked, was funded by the National Science Foundation under grant number 1252995. So your tax dollars paid for it.

I am not totally opposed to all research related to gender discrimination, but when researchers publish statistically invalid garbage, they should be banned from future grants.

Re:People of all genders should team up ...

[AmiMoJo]

Because it's flamebait. It's got everything - gender, hipsters and millennials, wise older workers, leftists, claiming other people's issues aren't real... It could easily have been copy/pasted from the /. troll playbook.

I hope it's a sign that people are finally getting fed up with the denials and moaning about it all being a non-issue, and are actually interested in commenting on the story. There is an interesting story here, but the GP completely ignores it.

Re:Misleading title

[AmiMoJo]

The actual complaint to the BBC is quite insightful:

Complaint
A reader complained that the headline of this article was misleading, that the study on which it was based was so flawed as not to merit reporting, and that the terms of the report were not duly impartial in relation to the question of the benefits or otherwise of workforce diversity in particular fields of employment.

Much of it is a standard anti-feminist argument, but the bit about the headline was found to have merit:

Outcome
Whether the study should have been reported was a matter of legitimate editorial discretion and, in the ECUâ(TM)s view, the article did not deal with matters which were controversial in the sense which would require a balance of views. However, there were no grounds for believing that the women among the cohort selected by the study were representative of women in general, and thus no basis for generalising about womenâ(TM)s relative ability. To that extent, the headline was inaccurate.
Partly upheld

Note that they are saying the research itself and the idea that there might be gender bias is not wrong or controversial, just that you can't infer from the study, which only looked at women on Github, that all women experience this bias.

The Slashdot summary is actually worse than the BBC article. It inaccurately summarises both the original article, the correction and the study.

Re:it is STILL discrimination

[amiga3D]

I'm against discrimination in my organization. Not for some thing about right and wrong but simply because I feel the best should rise to the top. Gender be damned. It doesn't make sense to decide anything on the gender or sexual preference or color or religion or much of anything else about the person who does the work except for the quality of their work. Most people where I work feel this way too. We have very few women (it's aircraft repair and modification) but those we have are mostly excellent at their jobs and respected by their male coworkers. They constantly try to hire more women but I've seen women show up for an interview, take one look at the work environment, and just walk out. Somehow though, it's alway spun that it's our fault women don't work here.

Re:People of all genders should team up ...

[AmiMoJo]

I'm not a hipster, a millennial or a leftist. But I do see gender issues in tech, for both men and women. It's not about dividing people and games, it's about making things better for everyone.

For example, there was a guy who had his first child at one place I worked. The boss was very upset when he said he wanted to take his full paternity leave allowance, and implied he would never have hired him if he had known. He then spent the next six months berating him for not being a "real man" and not making "his woman" bring up the child.

He ended up quitting before the child was born in the end (long story). If you don't think it's a problem then fine, you are entitled to that, but don't be a dick by moaning about other people trying to address it. It's not a conspiracy against you, unless you think the manager was right.

Careful, your slip is showing

[Solandri]

Anyone else notice the glaring sexism in the media coverage of this story about purported sexism in programming culture?

"get pull requests accepted more (except when you know they're women)."

When it seems like men are getting preferential treatment, the story is portrayed as discrimination against women.

"Women write better code, study suggests"

When it seems like women are getting preferential treatment, the story is portrayed as women being superior.

I propose journalists be forced to write these stories without knowing ahead of time which gender came out on top in a study. After the story has been written, the editor can go back and insert the proper gender-specific word or pronoun.

Paper's got Issues

[Cinnamon Beige]

You've done the exact same thing you've accused BBC of, claiming that overall pull requests are accepted at the same rate when this is not true. Women were statistically significantly less likely to have their code accepted than men as outsiders and statistically more likely to have their code accepted as insiders.

Citation please? From the way the entry here is (horribly) written, it looks like gender is less of a factor than being a known coder to the group--which means that if you're going to do the statistical comparison properly you're going to have to be careful to make sure you break it apart on both the gender variable (unknown vs known woman vs known man) and on 'known to community' variable, being careful to make sure the numbers you're comparing match across the board.

Doing a bit of work to find the paper itself reveals that it's not peer-reviewed which probably explains why a read-through of the methodology is leaving me extremely skeptical on if they controlled for this obvious confound, especially since there might be additional confounds--for example, I'm not really seeing any attempt to control for the possibility that the rate at which skilled female coders deliberately adopt genderless/gender-neutral personas may be higher than it is for skilled male coders, meaning that the higher acceptance rate (of 6.2% higher then men's) is actually an artifact of that--and hey, since somebody might as well point out the elephant in the room, let's bring up the possibility that some people might have their GitHub persona more accurately reflect their gender than their LinkedIn profile because there are some employment discrimination suits you don't want to be the one trying to win.

I'm also somewhat skeptical of the whole idea that changing larger parts of code is automatically less cautious, since this seems a rather bad assumption on the whole. I would expect the amount of testing done before submitting the code to be a better indicator--but you can't measure that so easily, even if a small tweak could easily enough result in a kludge in the code becoming suddenly broken, meaning that a more cautious coder might end up submitting a larger amount of changes because they're both submitting the tweak and what is necessary to get the program working again after the tweak. (I'd actually be unlikely to submit them separately, unless the tweak isn't essential to what I did to get the program work again and the kludge itself is a particularly ugly thing that needs fixing no matter what.)

These are the kinds of things that will get brought up in the peer review process for a paper, so really, I'm not surprised in the least that it isn't.

As a side note, I am, however, pleased to have learned from this paper that, in fact, I can have my LinkedIn profile be gender-neutral, which is my personal preference.