> If your candidates are 80% men and 20% women, and your hires are 80% men and 20% women, your process isn't discriminatory. DIscrimination is still happening, but it's somewhere before you in the pipeline. Some would argue that you have an ethical obligation to "reverse discriminate" to mitigate it, but that's much more ambiguous than plain discrimination being bad.
I'm not (right now) interested in ethical obligations to discriminate or reverse discriminate or anything. I'm interested in hiring the best candidate for the job.
If your candidates are 800 men and 200 women, and you need to hire the 100 best candidates from that pool, you have very little to believe that the 100 best candidates are roughly 80 men and 20 women. You have even less reason to believe that if you have a suspicion that there's discrimination happening in the pipeline.
So if your goal is hiring the best candidate for the job, you shouldn't look at the ratios of demographics coming out of a known-biased pipeline at all. If you end up hiring 80 men and 20 women, you should have your own reason why four times as many men than women were in the set of best qualified people.
If your goal is some vague social-justice politically-correct nonsense like ending reverse discrimination, then you care about putting the blame on the pipeline and slavishly following the numbers it gives you.
>> If your candidates are 800 men and 200 women, and you need to hire the 100 best candidates from that pool, you have very little to believe that the 100 best candidates are roughly 80 men and 20 women.
Why not?
Statistically, at least, this should be true, provided that the candidates have been treated equally before they came to you. If they have disadvantaged in that treatment, it will skew the outcome even more against the discriminated minority - e.g. if women tend to get shittier education in certain field because their teachers don't believe them as capable as men, then of those 200 female candidates, the proportion of ones that are less qualified is likely to be lower.
If you somehow end up with the numbers that are reversed, it's not at all unreasonable to ask why - there ought to be a cause, and for a difference that big, it should be prominent. It may well be a reasonable cause - maybe you just got a bunch of really bright women in that batch, and that will be reflected in their career records etc. But it may also be sexism or other form of discrimination not related to performance.
We do investigations of discrimination against minorities based on statistics like these all the time. For example, one of the underpinnings of the ongoing push for police reform in US is the disproportionate number of black people when it comes to police interactions, arrests (esp. for often substance-less crimes like "resisting arrest" or "disorderly conduct"), and convictions. Generally speaking, this is not considered proof of discrimination by itself, but when there's corroborating evidence (e.g. you find out that some PD had a racist internal culture, based on the SMS the officers send to each other etc), it can certainly strengthen the case to the point where discrimination becomes the default assumption, and another explanation would require substantial proof.
In this lawsuit, the guy does cite several specific examples that, at face value, do indicate discriminatory approach, so dismissing such a large disparity and such a quick rate of change as "it just happens, wasn't on purpose" out of hand is not feasible - it requires evidence to rebut.
I'm not (right now) interested in ethical obligations to discriminate or reverse discriminate or anything. I'm interested in hiring the best candidate for the job.
If your candidates are 800 men and 200 women, and you need to hire the 100 best candidates from that pool, you have very little to believe that the 100 best candidates are roughly 80 men and 20 women. You have even less reason to believe that if you have a suspicion that there's discrimination happening in the pipeline.
So if your goal is hiring the best candidate for the job, you shouldn't look at the ratios of demographics coming out of a known-biased pipeline at all. If you end up hiring 80 men and 20 women, you should have your own reason why four times as many men than women were in the set of best qualified people.
If your goal is some vague social-justice politically-correct nonsense like ending reverse discrimination, then you care about putting the blame on the pipeline and slavishly following the numbers it gives you.