[waterhouse]

Absolutely. This is one of the reasons affirmative action is poisonous—it creates rational reasons for discrimination. Which will, of course, be exaggerated by actual racists; but it's much harder to argue "Discriminating against group X is bad" when it's public knowledge that group X has passed a lower bar.

[menzoic]

Affirmative action doesn't require a lower bar. Why do people keep saying that? It's also been banned in California since the 90s.

[ralusek]

It literally mathematically requires a lower bar. If your criteria for hiring a software engineer is that an applicant demonstrate proficiency in a given set of tasks, for example, you will result in a pool of applicants sorted by proficiency. Say we have 100 slots open to hire people, and we would normally just take the top 100 applicants according to this proficiency criteria.

Now, introducing any criteria which results in a different set of applicants than the top 100 necessarily lowers the bar of proficiency. It doesn't have to be affirmative action, it could be any arbitrary change in criteria. "We want just as many people named Michael as people named Jim." Well, if that wasn't the case in the original set of applicants, you're no longer getting the top 100. Affirmative action by race is no different. It's not that there is a debate here, as I said, it is that you are necessarily lowering the bar of proficiency by introducing another set of arbitrary criteria.

Maybe there were 6 Michaels and 1 Jim in the top 100. In order to balance them out, there would likely have to be some Michaels removed, some Jims added, and some people from other common names removed. Every person that was removed for a Jim from outside of the original 100 was more qualified and excluded due to the arbitrary criteria. The bar was lowered.

[satyrnein]

This assumes that the existing system is already a perfectly ordered meritocracy. Affirmative action supporters are coming at it from a different hypothesis: that 20 of the applicants pass the objective proficiency bar, and now people will apply subjective criteria like "culture fit", who has "leadership potential", etc, and this is where bias creeps in. Somehow the qualified candidates from certain groups don't make the cut as often you'd statistically expect (if you also do not believe in group differences).

If this is the starting position, then it is mathematically possible for affirmative action to deliver more equitable outcomes without lowering the objective bar.

So yes, affirmative action produces suboptimal results if you believe the world is already perfectly fair, and the "losers" weren't as qualified, due to differences between groups in preferences or abilities. Alternatively, affirmative action provides a slight correction to an unfair world, if you believe that all groups are equally capable, and differences in outcomes indicate how much bias is left to overturn.

[username90]

> that 20 of the applicants pass the objective proficiency bar, and now people will apply subjective criteria like "culture fit", who has "leadership potential", etc, and this is where bias creeps in.

If that was the case they wouldn't drop SAT. Fact is they want to accept people with worse objective scores, this whole discussion and article is about that fact. Instead they will use "culture fit" and "leadership potential" to discriminate against Asians and bring in more desirable minorities.

[satyrnein]

I think the debate is over whether the SAT is as objective as we all want it to be. I'm sympathetic to the argument that rich kids with tutors have an advantage, but I personally don't think the solution is to drop it entirely.

[waterhouse]

> This assumes that the existing system is already a perfectly ordered meritocracy.

No, just that it's as close as you can get with the imperfect tools at your disposal. Even if your tools are really bad—like, if the variation in scores on your test suite is 10% ability, 90% luck—well, if that's the best tool you have for sorting by ability, then you should use it, and to the extent that you ignore its recommendations in favor of racial or other preferences, that will lower the average ability of the candidates you accept. (Unless your tool is so bad that selecting by race outperforms it—which is a very unfortunate situation, and one that should be avoided as much as possible.)

[ralusek]

> apply subjective criteria like "culture fit", who has "leadership potential", etc, and this is where bias creeps in

This would no longer be the example I described. This is why people who support something approximating a meritocracy are typically in favor of any efforts to remove bias, and move in a direction of blind hiring. It is disingenuous to say that this is the objective of those in favor of affirmative action, however, as color/gender blindness is not their goal at all. The example here of getting rid of the SAT is a perfect example of that.

> you believe that all groups are equally capable, and differences in outcomes indicate how much bias is left to overturn

And this is the fundamental difference. Advocates of affirmative action/CRT believe that different population outcomes can be used as a de facto post hoc rationalization that the system which produced the outcomes must be necessarily biased in favor or against the groups. This is fallacious thinking. The conclusion doesn't even follow your own premise, and your premise is simply an assertion of what you believe to be true.

"[I] believe that all groups are equally capable, therefore differences in outcomes indicate that systems are biased." This is a fallacious statement. Capability is a minor, minor portion of the equation. Interest, culture, behavior, geography, income, wealth, history... Where do these fit into your model?

Let me tell you something about hiring. I've been responsible for hiring engineers on many occasions, and still am. If I were instructed to achieve, for example, 50/50 parity between male and female engineers: I would have to hire 100% of the female engineer applicants. If I were instructed to make sure that 13% of the engineers were black (to be in line with population levels): I would have to hire 100% of the black engineer applicants.

Your de facto reasoning that the reason that engineers are overwhelmingly white/east Asian/Indian/Eastern European is that the hiring system is favored as such. The pool of applicants, however, skews even further towards this representation. Almost all companies are already trying to capture a greater proportion of other demographics, and are simply unable to do so. But in regards to sacrificing proficiency, if you understand the proportionality of the applicant pool, your argument of not sacrificing proficiency completely falls apart. It's as I said, if I were to get 50/50 female representation, I would literally have to get rid of proficiency criteria altogether and literally hire every woman on the spot. It would absolutely be a massive hit to proficiency. That isn't saying that women are less proficient at engineering.

Lastly, I'm curious if you care about this for anything else. For example, Indians are extremely over represented in medicine as compared to their population. Filipinos are extremely over represented in nursing as compared to their population. Because you believe all group are equally capable, you surely believe that a cabal of Filipino nurses and their in group preferences are responsible for maintaining the hegemony of Filipino nurse supremacy, correct?

[satyrnein]

You made the strong claim that affirmative action mathematically requires lowering the proficiency bar. I responded that it depends on certain facts about the world, on which there is much disagreement (whether hiring/admissions is completely objective, whether groups have the same abilities and preferences, etc). I didn't even specify what I personally believed, but I see you dug up your talking points.

I'm responsible for hiring engineers as well, and I don't actually think discrimination plays much of a role at this point in the pipeline; I just rarely see candidates from underrepresented groups cross my desk. From here, it looks like a supply issue upstream (whether from preferences, abilities, or bias). So in this little corner of the world, I agree that affirmative action would require compromising on proficiency. Other corners may be different.

[waterhouse]

> Affirmative action doesn't require a lower bar.

The one strategy I've heard of that doesn't do this, and isn't something any rational organization who saw no intrinsic benefit to "diversity" would already be doing, is "spending extra recruitment resources to yield good candidates of the underrepresented groups". For example, you could send 5 recruiters to all-female colleges or majority-black colleges that aren't highly ranked in CS, in the hopes of turning up as many good programmer candidates as you'd get from sending 1 recruiter to a highly-ranked CS college. That indeed does not require a lower bar—although it spends resources in a way I'd consider wasteful.

But I don't think that's what affirmative action normally means, and I'm generally leery of allowing proponents to redefine the term more broadly (that type of thing enables motte-and-bailey argumentation).

Did you have another strategy in mind?

[908B64B197]

> Affirmative action doesn't require a lower bar

How else can it be done?

[unishark]

Dropping the bar altogether of course.

[jjeaff]

Stop using decisions factors that seem to benefit one race over another.