Sunday, September 30, 2007

Criminal bias

Jonathan Ekman writes in a alumni magazine:
Only a social scientist “Shooter’s Choice,” July–Aug/07 could seriously argue that any possible bias against black suspects derives from “ambient social stereotypes” and not from rational discrimination based on statistical probability. The racial disparity in criminal activity in this country is widely known, yet for many in the groves of academe it must be ignored, explained away, or simply discounted.

Social psychologist Joshua Correll replies: Mr. Ekman’s question is one we often hear. It’s true that African Americans in the United States are disproportionately likely to be arrested for violent crime, and my colleagues and I recognize that this disparity may contribute to a pattern of bias that leads participants in our video-game studies to shoot black targets more quickly and more frequently than whites. It is not our intent to “explain away” the possibility that racial bias in decisions to shoot stems from more-or-less rational decision-making. Our intent is to understand the mechanisms that inspire this behavioral bias. In truth, no one would argue that participants in our studies shoot blacks more quickly or more frequently than whites simply because blacks are more likely to be arrested. We need to recognize that there is a crucial intervening process. People shoot African Americans more quickly because they subjectively, psychologically associate them with crime and danger. It is this psychological association, or stereotype—not arrest rates—that drives behavior.

Of course, the association may derive from racial disparities in criminal activity just as it may derive from movies, news reports, or gangsta rap music. The point is that the psychological representation is a critical, proximal variable. In our efforts to understand the behavioral phenomenon, we believe it is valuable to understand the processes that generate it. In this case, those processes seem to involve stereotypes.
Correll used a video game to study whether cops are quicker to shoot a black suspect than a white suspect. It is a good idea, and he gets some interesting results. But then he claims to distinguish between active prejudice, racial stereotypes, and rational assessment of statistical facts. He discounts active prejudice because blacks show the same biases as the whites. But then he jumps to the conclusion that it could not be rational decision-making so the explanation must be racial stereotypes.

Ekman is right. Only an academic social scientist would give such nutty arguments. Correll refuses to even admit that there is racial disparity in criminal activity, and merely alludes to a disparity is arrest rates. Cops are going to shoot the most threatening suspects, based on all available info. It is only common sense that age, sex, race, clothing, and any other aspect of appearance will be rational factors.

The proper way to test for bias is to compare the simulation with some actual threat measure. For example, if urban 25-year-old men are 100 times more likely to shoot at a cop than little old ladies, then cops doing a video game simulation should be 100 times more likely to shoot the young men. Any more or less would be bias.

No comments: