Friday, January 8, 2010

In Profile

Recently, profiling has been on the mind of John McGuinness over at Man Bites Blog, and he's written about it in a couple of recent posts. In one, he makes a very good point:

Obviously, simplistic racial profiling, in so far as it reduces people to their race or origin, is an affront to personal dignity.
But, perhaps, it misses part of the point in the broader argument against racial profiling, especially of the simplistic sort.

In the end, there are really three reasons to profile:
  1. Savings in labor - you don't have to interact with every single person.
  2. Lower miss rate than a purely random sample - given the fact that you're not interacting with each and every possible perpetrator (i.e., the entire population), you want to have the greatest chance possible of catching the actual perpetrators (i.e., the actual people who are up to no good).
  3. Lower false positive rate - every moment spent working over a non-perpetrator is, basically, wasted time. The fewer false positives the system generates, the better.
If your profiling regime doesn't make some progress towards achieving goals 2 and 3, you're not really doing anything at all helpful. Profiling has acquired a bad name here in the United States, because basically, it was popular for an off-label use - deliberate harassment of an unpopular out-group. Stopping, say, young Latinos in nice cars on the highway and searching them for drugs might have saved a state trooper the time and trouble of stopping everyone who might drive by. But it gives a pass to any non-Latino who may be a drug courier, and it results in stopping a large number of Latinos who aren't drug couriers.

It's safe to say, given the tenor of some public comments, that deliberate harassment of an unpopular out-group is part of what's driving some popular support for profiling, vis-a-vis terrorism. Coupled with an increased illusion of safety is the fact that the people who are going to have to put up with the problems and indignities brought on by false positives are unsympathetic, out-group members. But simplistic racial profiling, while it may slake a fearful public's need to feel that someone is applying the boot to their tormentors, isn't likely to substantially bring down the rate of false positives, or push the terrorist catch rate to anything near the 100% that we're commonly told is the only acceptable outcome.

No comments: