Monday, January 30, 2006

Racism or Discrimination?

Winterspeak at Asymmetrical Information posted on a topic I've been meaning to talk about for a while. Some Harvard researchers have come up with a test for subconscious bias. In brief, you are given two categories: in the racial test, say, black and good, and white and bad. You are given a series of words and faces, and asked to categorize each. Most people take measurably longer to sort words into white/bad and black/good categories than into white/good and black/bad; these results hold up even if the subjects are told that the test is testing for racist inclinations. The researchers interpret this as evidence that most people have mild but subconscious pro-white racism. Which is a sensible conclusion, as far as it goes.

But (and I confess a bit of trepidation at publishing this, since it could easily be taken in very much the wrong way), these subconscious leanings aren't necessarily irrational. It's a sad fact, for instance, that relatively more blacks are raised in impoverished neighborhoods, raised by single parents, and given low-quality educations, and are thus relatively more likely to commit crimes. So if you know nothing else about a person, it's reasonable to assume that a black person is likely to have a below-average education and above-average criminal tendencies, since the average black person does. The problem comes in when you do have additional information, and still cling to the generalization: I don't have any stats in front of me, but I'm pretty sure that upper-middle-class black students with good educations are pretty much the same under most/all metrics as upper-middle-class white students with good educations. I'd be a bit surprised if they weren't.

So the problem, as Winterspeak says, isn't when the first approximation is "the black person is, statistically, likely less well educated than the white person." It's when you don't update that based on additional, more specific information; when you learn that the black man has a degree from Harvard and the white guy flunked out of community college, the specific information completely trumps the general and race no longer should be used to evaluate educatedness.

However, I disagree with Winterspeak that this isn't racism. It seems to me—although here we get into more definitional territory than substantive disagreement—that if you're biased toward one race and away from another, you have at least some racist beliefs. You're certainly discriminating based on race; and whether the beliefs are true or not, that doesn't mean they aren't technically racist, or at the root of a lot of observable, unjustifiable racial discrimination.

Because people have a strong tendency to stick with their first evaluations, and not update with new data. A recent NYTimes story discusses this with regard to politics specifically, but in almost any field people don't like to admit that they were wrong, and so stick with their initial judgments long after new info should have caused them to reevaluate. Thus even after he knows enough to judge between the white and black job candidates on specific individual traits, a manager may still stick with his initial from-the-hip guess that the black candidate is less qualified simply because his skin is a darker shade. Because of this limitation, I think it potentially makes good sense for an organization to give some preference to groups that traditionally suffer from discrimination, to correct for examiners' latent biases. Better still would be not allowing racial information, or at least not allowing it until first impressions have already set, but if that isn't possible a conscious bias in one direction may correct for an unconscious bias in the other (to clarify, I'm not supporting legally mandated affirmative action; I'm leery of the government interfering either in the market or in racial evaluations. I simply contend that it may be rational in some cases for a private actor to implement some sort of affirmative action program of its own).

But this analysis also suggests that common methods for combatting lingering racist/sexist/etc sentiments are doomed to failure. For instance, awareness campaigns that try to convince people that racism is bad, or blacks are the equals of whites, or women are just as good on the job as men, or what have you. The Harvard tests measure subconscious and uncontrolled bias; scores didn't improve even when participants were trying to correct for any pro-white sentiments they might have. With the exception of some truly bigoted and often reprehensible figures (David Duke, I'm looking at you), I suspect this is how almost all bigotry survives, in an unconscious and perhaps partially justified initial bias in one direction for another. Encouraging people to exercise conscious bias in the opposite direction may be worthwhile, but simply arguing that racist beliefs are bad will get nowhere, because most of the targets don't knowingly hold them. And those who do knowingly hold them are unlikely to be much persuaded; activists' efforts may be better spent elsewhere.

Sunday, January 29, 2006

Real Update Coming Soon

Until then, two new things to chew on.

First, David Friedman, a leading libertarian/anarcho-capitalist thinker, has started a new blog. It's been really interesting so far; I'm adding it to the blogroll. Check it out.

Second, if you want to waste several hours of your life, check out You can browse through their catalogue of button slogans; I've killed way too much time reading through and chuckling at the more amusing ones.