Suppressing Intelligence Research: Hurting Those We Intend to Help (2003)

The contradiction between the ultimate value of pure scientific inquiry and the detriment of results that could be construed to be racist is an issue I've puzzled with some in the past. The conclusion I've come to is that it's part of our social contract to never act on any such correlation - any result that says that X person may behave like Y must be totally disregarded when interacting with any individual (except for, say, medical use). This extends beyond racism and sexism from disregarding institutionalized prejudices into actively ignoring discovered information. It's a fine line, for sure.

Are you sure they're being flagged? I think HN has been experimenting with showing recent submissions more on the main page in order to remedy the problem of /newest being a ghetto no one votes in.

IMHO we're better off without certain kinds of intelligence research. I can't imagine any good that could come from distinguishing a certain group of people as more or less intelligent. On the other hand, this information would undoubtedly be used as a rationale for all kinds of discrimination.

I think we should distinguish between the individual and the group.

When dealing with a specific individual, it is important to avoid short-circuiting inferences based on some group membership. This does not require any suppression of the truth: it is a fact that intra-group variance is higher than inter-group variance, and a member of the group with the lower average may surpass a member of the group with the higher average.

For example, let's say you have two candidates for a tech job, one black and one Asian. If you were forced to pick one based on that fact alone, it may be true that picking the Asian would be the better strategy, but it is clear that choosing an employee based on race alone is completely asinine. Both fairness and good business practices require that you evaluate both candidates on their individual merits; at which point any prior probabilities based on race become moot, since you now have specific individual information to decide on. And it is entirely possible that the black candidate will turn out to be the better pick.

However, it is wrong to try and transfer this to the collective level and claim that the racial composition of your workforce should match that of the general population. That would be the expected outcome if and only if it were the case that aptitude for tech jobs is equally frequent amongst different races. However, scientific evidence does not seem to support that assumption; in fact, it seems to strongly support the opposite.

In that situation, it is not just that fairness at the individual level does not lead to "fairness" at the collective level: the two objectives are actually at odds. If aptitude is not equally distributed and you still want to force the racial composition of your employees to match that of the general population, the only way to do that is to pick some worse candidates over better ones just because of their race. In order to be "fair" to groups, you have to be unfair to actual people.

Which, by the way, is a very good reason why this research should not be suppressed. Making policy decisions based on falsehoods is bad enough, but treating people unfairly because of falsehoods is even worse.