9 ms·
He referred to a "lowered bar" in the form of a reduced false-negative rate for "diversity candidates"; at no point in the document did he imply his female engi
by slavak 9y ago
He referred to a "lowered bar" in the form of a reduced false-negative rate for "diversity candidates"; at no point in the document did he imply his female engineer colleagues are less qualified. He also had a reference to apparently support this assertion, but since the hyperlink leads to an internal Google group we can't really know its contents unless a Googler informs us.
From the original memo:
> However, to achieve a more equal gender and race representation, Google has created several discriminatory practices:
> ...
> Hiring practices which can effectively lower the bar for “diversity” candidates by decreasing the false negative rate
- pimmen 9y agoLet's assume this guy knows statistics; by lowering the rate of false-negatives, you raise the rate of false-positives. This means that he's heavily implying Google has employed undeperforming engineers with these practices.
- rtx 9y agoIsnt that whats affirmative action is about. People need to own it. Yes skills may vary depending on hiring practices. But so what?
- doubleshame 9y agoThis. I wish there official response from Google would have just opened up to their stance and stood ground. Instead of flimsily trying to appease everyone. It would have been more respectable.
- benchaney 9y agoThis is only true if the total number of trials and the test outcomes stay the same, which is definitely not the case here.
- asdfologist 9y agoHonest question: without seeing the reference he used to back his assertion, how did you conclude that his belief about this is wrong? Or did you actually manage to read it?
- imron 9y ago> This means that he's heavily implying Google has employed undeperforming engineers with these practices. No he hasn't. Rather, he's implying that 'non-diverse' hires have a higher bar to to pass, and that by enforcing diversity quotas you tend to end up hiring exceptionally qualified people from the 'non-diverse' group, which can have the opposite of your desired effect, and perpetuate stereotypes about the performances of people from different groups. For example, let's say you have 100 people interviewing for 10 positions. Of that, due to current representation of women in tech let's say the candidates comprise 20 women and 80 men. Let's also assume that 25% of those interviewing meet the technical standards a.k.a 'the bar' required by Google (that makes for 5 women and 20 men). Google however is only hiring 10 positions, and in order to meet diversity requirements will hire 5 women and 5 men (50/50 split). Each one of the people hired is qualified to work at Google and meets the bar (none of them are underperforming). Without any diversity requirements a candidate would need to be in the top 10% of other candidates to get a position however now, for the women, they only need to be in the top 25% of other woman candidates to get the position, whereas the men need to be in the top 6.25% of male candidates. The bar has been effectively lowered for the women candidates and raised for the male candidates, despite all of them being worthy and capable of working at Google. If you are always hiring equal numbers of people from both groups, but from one group you are always hiring the top 6.25% of candidates, and the other group you are hiring from the top 25% of candidates, then over time you'll start to find that on average, one group has a larger percentage of high performers than the other.
- ummonk 9y agoNot true. You can lower the rate of false negatives without raising the rate of false positives using multiple measurements instead of just one, which as I understand it is what Google does for candidates from underrepresented groups.