Traditional Profiling Ineffective, Study Says
POLITICAL CORRECTNESS and legal concerns aside, a new study finds that traditional profiling to assess the need for secondary screening (at security checkpoints) is no more effective at rooting out criminals than random screening. The study highlights a mathematical approach that might be better at discovering potentially dangerous individuals, but it did not differentiate between behavioral factors and racial or ethnic ones.
Computer scientist William H. Press of the University of Texas at Austin conducted the analysis, which he wrote about in Proceedings of the National Academy of Sciences USA. Press’s study finds that strong profiling is no more effective than purely random screening. An example of strong profiling would be determining that a person from country A is 10 times more likely to be a terrorist than the general population and subsequently screening country A’s citizens 10 times more than everyone else.
However, Press has determined mathematically that there is a “golden mean” in profiling, which can be deduced by square root biased sampling. What that means is that someone should be screened the square root of the amount that they are statistically likely to be a terrorist or criminal. For example, if people who buy one-way airline tickets (a behavioral factor) are nine times more likely to be terrorists than others, then they should receive secondary screening three times more than the general population, with three being the square root of nine.
Press figured it out by developing formulas for detecting a “malfeasor” using the strong profiling approach and comparing those results with the effectiveness of the square root bias formula approach. Press had been working with the square root bias formula for a biological application dealing with genomes, and while walking through an airport one day, he realized that the formula could apply to secondary security screening.
Press explains: “If you simply screen the people with the very highest profiles, then it’s less effective than random screening, because you end up wasting your effort and screening those high [probability], innocent people over and over again…. And then also, you provide sanctuary for the terrorist who might just happen to not have a high-profile…. So, you need a compromise, where you screen high-profile people a little bit more, but not so much that you take effort away from screening everybody else where the terrorist might actually be lurking.”
With regard to terrorism, however, this approach would likely lead to only a small amount of profiled screening. “I kind of doubt that you can get accurate profiles that would flag someone as being more than a factor of, say, 10 times more than the average member of the public [of being a terrorist],” says Press. “And then by the time you take the square root of that, you’re at approximately three. And at that point, you’re using your profile to flag people only a bit more than just doing everybody randomly.”
Press questions whether that slight amount of additional screening is worth the other consequences that come with profiling, particularly if it is based on characteristics, such as country of origin or race, rather than on behavior. He also doubts that the government could develop statistical profiles accurate enough to justify implementation of square root bias sampling.
Barry Steinhardt, director of the technology and liberty program for the ACLU, an organization that actively campaigns against racial profiling, agrees that the government likely would be unable to employ square root biased sampling in security screenings and investigations. “These are agencies that operate in fairly simple ways. They can’t build systems that are that sophisticated,” says Steinhardt.
Still, some who favor profiling have seized on Press’s findings as an argument supporting its use in security screening.
“I think our current system which treats everybody equally…is basically a triumph of political correctness,” says public interest lawyer and George Washington University law professor John Banzhaf. Banzhaf says that targeted screening techniques have been shown to be effective in disease screening and in other situations, so why not apply them to security screening?
Steinhardt counters that while there is strong evidence that some diseases are more prevalent in one race than in others, there is “no such evidence for terrorism.”
The profiles do not have to be precise, Banzhaf contends. If the government can get a ballpark estimate, then that should be good enough to find a square root and make profiling more efficient, he says.
It is legal to consider race in matters of “national security,” according to the Department of Justice’s Guidance Regarding the Use of Race by Federal Law Enforcement Agencies, as long as the use respects constitutional rights. It is unclear if the Obama administration will make any revisions to this guidance.
Banzhaf argues that racial profiling in general is constitutional. He cites a recent Supreme Court case that found that using race as a factor in college and university admissions is constitutional due to the “compelling interest” in obtaining benefits of diversity. Banzhaf argues that “since the interest in preventing things like airplane hijackings is stronger than the compelling interest in having people of different races in a classroom,” the same standard should apply in racial profiling cases. Steinhardt disagrees with this interpretation, saying that in more relevant examples, such as cases of “driving while black,” courts have ruled against racial profiling.
Banzhaf also argues that profiling cuts down on the total number of people screened, allowing lines to go more quickly. “We have a situation where the United States has limited resources to search people boarding airplanes. If we can make the secondary screening twice as efficient, we save an awful lot of time, an awful lot more money, and we’re substantially more likely to stop terrorists.”
Supporters truly believe profiling enhances the chances of catching the terrorist. Opponents note that it may instead increase the chances of an atypical terrorist getting through the checkpoint unnoticed while security focuses its efforts on “the usual suspects.” Press has found mathematically that the opponents may be correct, except with use of the square root bias application.
As mentioned above, Press doubts whether the limited additional screening is worth the political and social ramifications of racial profiling. However, he says, “the policy-making should be informed by what the science or math really says.”