Even assuming that IQ is completely correlated with ability to do the job, your pool of applicants that is “already selected for IQ” is not selected with 100% accuracy. In other words, being in the pool is a proxy for IQ. You’re better off using two proxies than one (unless the effect of one proxy is nil when conditioned on the other proxy)--if you want to maximize the applicant quality, you should pick people from the pool, but prefer men when comparing two applicants who both are in the pool.
The basketball example here is bad because you actually can measure someone’s height. If you can measure it you don’t need proxies for it. You’re not going to measure “ability to do the job” without using proxies and even if IQ is completely correlated with it, you’re not going to measure IQ without using proxies either.
Height is a proxy for basketball ability and the résumé is a proxy for research ability. It is possible that the latter is a worse proxy than the former, but it seems unlikely that it is much worse.
(ETA: Especially given that the latter enables you to go to arXiv and look at the applicant’s actual research output so far, whereas knowing how tall someone is doesn’t enable you to watch all the basketball matches they have played in.)
I didn’t downvote, but your second paragraph has a problem in that in the basketball example height is only a proxy for the ability to play basketball.
The first paragraph is a bit iffy, too, because proxies have different effectiveness or usefulness. By the time you’re estimating someone’s ability to do the job on the basis of a resume, the male/female proxy becomes basically insignificant.
In any case, I think that the better language here is that of priors. It’s perfectly fine to have different priors for male job applicants than for female job applicants, but once evidence starts coming in, the priors become less and less important.
in the basketball example height is only a proxy for the ability to play basketball.
IQ is a proxy for the ability to do well in a job, too. I was ignoring that, so in the analogy I would have to ignore that for height and ability to play basketball.
proxies have different effectiveness or usefulness
Arguing “the other proxy is so much better that we don’t need the original one” is not the same as “the other proxy is better, and that’s all we need to know”, and is even farther from “we can measure it so we don’t need a proxy at all”.
Even assuming that IQ is completely correlated with ability to do the job, your pool of applicants that is “already selected for IQ” is not selected with 100% accuracy. In other words, being in the pool is a proxy for IQ. You’re better off using two proxies than one (unless the effect of one proxy is nil when conditioned on the other proxy)--if you want to maximize the applicant quality, you should pick people from the pool, but prefer men when comparing two applicants who both are in the pool.
The basketball example here is bad because you actually can measure someone’s height. If you can measure it you don’t need proxies for it. You’re not going to measure “ability to do the job” without using proxies and even if IQ is completely correlated with it, you’re not going to measure IQ without using proxies either.
Height is a proxy for basketball ability and the résumé is a proxy for research ability. It is possible that the latter is a worse proxy than the former, but it seems unlikely that it is much worse.
(ETA: Especially given that the latter enables you to go to arXiv and look at the applicant’s actual research output so far, whereas knowing how tall someone is doesn’t enable you to watch all the basketball matches they have played in.)
The problem is that height is a number, whereas it’s hard to translate résumé into a number without resorting to various gameable metrics.
So what?
It’s a lot easier to measure height then evaluate a résumé.
Is there some reason why this was modded down aside from political incorrectness?
I didn’t downvote, but your second paragraph has a problem in that in the basketball example height is only a proxy for the ability to play basketball.
The first paragraph is a bit iffy, too, because proxies have different effectiveness or usefulness. By the time you’re estimating someone’s ability to do the job on the basis of a resume, the male/female proxy becomes basically insignificant.
In any case, I think that the better language here is that of priors. It’s perfectly fine to have different priors for male job applicants than for female job applicants, but once evidence starts coming in, the priors become less and less important.
IQ is a proxy for the ability to do well in a job, too. I was ignoring that, so in the analogy I would have to ignore that for height and ability to play basketball.
Arguing “the other proxy is so much better that we don’t need the original one” is not the same as “the other proxy is better, and that’s all we need to know”, and is even farther from “we can measure it so we don’t need a proxy at all”.