Re:Ask a Female Engineer: Thoughts on the Google Memo
by hackinthebochs on Wednesday August 16, 2017 @11:13AM
(#15028085)
I don't see how you don't recognize this as effectively lowering the bar (if not intentionally).
>Additional rounds of interviews are more like trying to accurately diagnose a condition using multiple different tests because the initial test is known to have poor sensitivity and will produce false negatives.
The difference is that what they're testing for isn't a binary proposition (do you have the disease or not), but a spectrum (what is your skill level). Viewing this in terms of false-positives or false-negatives is insufficient. If we think of programming skill as a spectrum, we can ask what is the average top-% of candidates who pass the interview (we might guess its top 5% of all developers). If everyone has the same test then the average top-% is unchanged regardless of any efforts to get more minorities to take the test. But once you start giving more tries to minorities your average top-% necessarily reduces.
Whatever your test is designed to admit (say you're interested in hiring only the top-10% of developers), the average of those who pass will be higher precisely because of the chance factor. Being significantly better than the intended cutoff gives you a better chance at passing and so those who pass skews towards better than the intended cutoff.
I'm not saying whether this is a good or bad thing, but the average skill of those who pass must reduce. It is very straightforward to see this as effectively lowering the bar.
>Additional rounds of interviews are more like trying to accurately diagnose a condition using multiple different tests because the initial test is known to have poor sensitivity and will produce false negatives.
The difference is that what they're testing for isn't a binary proposition (do you have the disease or not), but a spectrum (what is your skill level). Viewing this in terms of false-positives or false-negatives is insufficient. If we think of programming skill as a spectrum, we can ask what is the average top-% of candidates who pass the interview (we might guess its top 5% of all developers). If everyone has the same test then the average top-% is unchanged regardless of any efforts to get more minorities to take the test. But once you start giving more tries to minorities your average top-% necessarily reduces.
Whatever your test is designed to admit (say you're interested in hiring only the top-10% of developers), the average of those who pass will be higher precisely because of the chance factor. Being significantly better than the intended cutoff gives you a better chance at passing and so those who pass skews towards better than the intended cutoff.
I'm not saying whether this is a good or bad thing, but the average skill of those who pass must reduce. It is very straightforward to see this as effectively lowering the bar.