Quantcast

Mathematics May Be Better Than Random Selection At Airports

November 18, 2010

Computer scientists argue that racial profiling to rule out potential terrorists is less effective than random searches, but simple math could be an even better solution.

“When you have any profiling at all, it quickly becomes less effective than random sampling,” University of Texas professor William Press told the AFP news agency. A paper on this subject appeared in Wednesday’s issue of the journal Significance.

Profiling does not work because “you end up screening the same innocent people over and over again, just because they happen to be in a profiled group,” Press said.

Previous studies showed that any rise in success due to racial profiling is actually due to increased levels of law enforcement.  More police focusing on one group will catch more criminals since fewer police and resources are focused on other groups.

“It is simply better to do uniform random sampling, which means everyone who shows up at the airport should have the same chance of being screened in the same way,” said Press.

However, he has come up with an idea that might offer a better solution.

“It is this thing called square root sampling,” he told AFP.

Using his method, screeners would approach a group thought to be, for example, 100 times more likely to be harmful, and then check the square root of that number, or 10 times.

“That actually would be better than uniform (random) sampling. The trouble is there is no good way to do that.”

Press teaches university-level statistics and uses the example for his students.

“One could imagine a system in which people’s risk factors are evaluated and as you show up in airport you know, in some computerized automatic way the computer flashes either red or green and does this square root business which would be some form of optimal profiling,” he told AFP.

“But I don’t know anyone who actually thinks you could make such a system work.”

On the Net:


Topics: Sampling, HTML, AFP


comments powered by Disqus