NEEDLES IN A HAYSTACK: Terrorists are so rare, a new study concludes, that a heavy reliance on profiling is no more effective than random sampling of the population in locating them. Image: ©ISTOCKPHOTO.COM/LYAC
Profiling is a hot-button issue—civil liberties groups maintain that making assumptions based on race, nationality or ethnicity is unacceptably discriminatory, whereas some prominent conservatives argue that the method is an effective means to combat crime and terrorism, and therefore worth the social cost.
So who's right?
According to new research, it is no more effective to profile strongly—that is, subject individuals to increased scrutiny in proportion to their presumed likelihood of malfeasance—than it is to randomly flag individuals in the general population when it comes to rooting out terrorism. The reason, says study author William Press, a computer scientist and computational biologist at the University of Texas at Austin: terrorists are vastly outnumbered by innocents, and it's a waste of time and money to screen and rescreen the same benign people.
"The existing literature tends to assume away the problem that drives the result here," which is the inefficiency of oversampling, says John Knowles, an economics professor at the University of Southampton in England who has published many papers on racial profiling. "This is obviously an important problem in real life, and hopefully Press's result will draw academic attention to it."
The extent of profiling in the U.S. is difficult to quantify. According to Amnesty International USA, 24 states have laws on the books prohibiting racial profiling in one form or another, but only four states outlaw profiling based on religion. Guidelines for federal law-enforcement agents prohibit racial profiling generally but reserve some leeway in combating terrorism, illegal border crossings and threats to national security. And although the U.S. Department of Homeland Security's Transportation Security Administration (TSA), which conducts security screenings at airports, says it does not profile by either ethnicity or religion, it has been sued recently for allegedly doing just that.
If profiling were to be used for security purposes, Press says, it would actually be more effective in a weakened form. Mathematically speaking, it would be optimal to screen individuals in proportion to the square root of their presumed probability of malfeasance. In other words, someone deemed nine times as likely as the average traveler to be a terrorist would be subjected to three times as many screenings. (Such an approach, legal and ethical questions aside, would require accurate quantitative assessments of those probabilities—a tall order in its own right.) This so-called square-root-biased sampling would distribute resources more evenly across the general populace and thereby turn up more wrongdoers by casting a broader net, according to Press's model.
In pure political terms, might the study, published online today in the Proceedings of the National Academy of Sciences USA, be construed as a scientific argument for racial, ethnic and religious profiling, albeit in a diminished form? Press doesn't think so; square-root-biased sampling, if it were somehow implemented, would limit profiling so much as to "reopen the moral and ethical questions of whether the profiling is worth the social cost at all," he says. "Personally, I would say that it is not." The mathematics of the inefficacy of strong profiling, Press adds, "could actually lead one to a decision not to profile at all."