Women, men and software — Alexander Serebrenik

Some people (men) wonder why women do not participate in StackOverflow.

Alexander was interested in this topic, so he tried to figure out what exactly was going on. Of course, in an automated way. Some people have names which are relatively easy to match to a gender, like Gunter. This is sometimes local, Andrea is a girl’s name in the Netherlands, but a male name in Italy. Information can be also obtained from profile pics. Source code is here.


The tool could not find a gender for 37,5% of people. We could ignore them, but this is a large share! Alexander explored 3 scenario’s:

  • Ignore them
  • Assume they are all women
  • Assume they are all men

A question from the audience: could some of the male profiles be female owned? There is some anecdotal evidence (from online poker) that women take on male profiles to seem more seriously, and males take on women profiles to get more mercy.

Independent of which of the three scenarios you consider, the results are as follows:

Interestingly, the results are different from (left column) mailing lists of Drupal and WordPress. The full paper is here.

This can be explained from behavioral economics! Women and men perform similar in a non-competitive environment, but lose effectiveness when competing, especially when competing against men.

Great question from the audience! Are women more effective in collaborative environment? Alexander does not know, but the next part of the talk will be about collaboration and diversity.

Is diversity good?

Some people say no: People like to be in groups with people that are similar [Byrne], and treat members of their own better[citation missing]. Some people also say it is good, because diverse problem solvers outperform high ability performances [Hong & Page] and multicultural groups are more creative [missing]

So, Alexander examined GitHub. He looked into team size, forks, project age and compared them to gender diversity and tenure diversity. His study suggests that increased gender and tenure diversity are associated with greater productivity (more commits)  Secondly, he investigated turnover. Do gender and tenure diversity impact that? Yes. More tenure diversity -> more turnover, but gender diversity -> less turnover.

So, that is a good story! Gender diversity increases productivity and decreases turnover. BUT!

More info in the paper

Codes of Conduct

Over the past few years, many projects have adopted Codes of Conduct. Alexander examined use of them, and found that contributor covenants are way more common than C0C’s. He also conducted interviews with writers of them and categorized them:

The big open question is whether codes of conducts work? This is of course really hard to measure. My guess is, that if you could even measure the number of incidents, they would go up after the CoC, because it makes people feel more comfortable to speak up.

Sadly I could not find a version of this paper, it will appear at this year’s SANER.