None of that is really relevant, though, because we aren't talking about a generic OSS conference. We are talking about a PHP conference. What is the percent of women who are PHP programmers?
Actually, that number would not be all the useful, either. Such surveys are going to find out what percent of various groups use PHP. That would be somewhat useful for estimating what the conference attendees should look like. Conference speakers, on the other hand, should be people are are doing new things with or to PHP. That's going to be a much smaller group than the people who use PHP, and so is likely to have much more skewed demographics compared to the larger PHP user population.
[1] https://www.wired.com/2017/06/diversity-open-source-even-wor...
It was fully booked for men during the first event, but only one woman joined. The second event, fully booked for women and only two men.
Third event they called it Wellness Evening, with no mentioning of any hair pack or grill, and it ended up fully booked for both men and women. Forth event similar.
The probability of those outcomes happening through naturally probability is extremely unlikely. It is almost like people will interpret subtle gender identity in events, and then have it impact their decision to join.
As other repliers have already stated: a) you chose a flawed survey to get your "10%" figure, b) the percentage of "women in open source" is far different, for a variety of reasons, than the percentage of women who are likely to submit a proposal to speak at a PHP conference in Germany, c) you dismiss any possibility that unusual ratios could possibly be due to reasons other than bias (is lung cancer biased because it overwhelmingly affects older people?)
You've offered zero fundamental evidence of bias (i.e. they filtered their mailing lists, they adjusted female speaker ratings, etc), other than the math doesn't work out the way you'd like, which isn't surprising given the invalid inputs and assumptions you started with.
https://www.nature.com/news/why-women-talk-less-at-conferenc...
While skewed outcomes can make you suspect bias, they simply are not sufficient evidence to conclude that there is bias.
For example, around 50% of the population is male. Yet 0% of all births are to males. Obvious discrimination against male fathers! Silly me, no, of course not bias. There is some other factor at work.
Or take the 100m dash as Olympic discipline, or in fact most of the track and field running events. Extreme racial skew. Bias?
And no, I am not saying that these same factors are at work here, just that you cannot conclude bias from unequal outcomes. In fact, as far as I know the science, it would be exactly matched outcomes that would actually be highly suspect.
See also: Simpson's Paradox ( https://en.wikipedia.org/wiki/Simpson%27s_paradox ). Again, not saying that this is an instance of Simpson's paradox, but definitely saying that just because something looks like bias when superficially examining outcomes, that does not at all mean that bias is the actual cause.
And of course there are a lot of potential reasons why there would be a skew, many of which have been identified as fairly reliable gender differences, including willingness to take risks,
> 10% of OSS software programmers are women
The numbers I found were lower, more in the 1-5% range, skewing toward the lower end:
https://www.techrepublic.com/blog/software-engineer/it-gende...
https://link.springer.com/chapter/10.1007/978-3-319-39225-7_...
https://www.itprotoday.com/linux/gender-roles-search-open-so...
To sum up, your argument is flawed for at least 3 reasons:
1. Your numbers for the baseline are wrong
2. Your assumption that any skew in outcomes proves bias is wrong
3. You ignore actual gender differences that can explain different outcomes
I doubt that’s actually the case. But 1/250 does seem very low.
