[pygar]

If you have 5 equally good candidates, 4 men and 1 woman that woman has a 1/5 chance of being selected. The others also have a 1/5 chance as individuals but the candidate selected has a 4/5 likelihood of being male.

If you want to hire more women what do you do exactly?

[gameman144]

In this scenario, you should expect to have a representation of around 1 woman per 4 men, if the rate of qualified candidates is 1 woman per 4 men. If you want to hire more than that ratio, you have to be doing some discarding of otherwise qualified candidates on the basis only of being men.

[pygar]

Sure. But even if you want to hire at about 1 women per 4 men you would need to actively do so. I.E. create a hiring policy in which she would be selected instead of one of the other equally qualified men.

[gameman144]

Not at all! Passively choosing from the equally-qualified-candidate distribution across the board will yield representation that's right in line with that distribution.

There may be some groups that are entirely men and some that are entirely women, but the aggregate result will approach the true population of qualified candidates.

[samatman]

Yes, it's true that if you want to hire some group of people above their percent rate in your applicant pool, you're going to have to put your thumb on the scale for that group of people.

But this has the opposite effect from what you were just claiming. It means you hire worse candidates merely on the basis of their immutable characteristics.

Or, equivalently, you discriminate against other candidates on the basis of their immutable characteristics. Which is illegal in the US.