Sam Harris in his blog:
Imagine that you work for the TSA and are executing a hand search of a traveler’s bag. He is a young man in his twenties and seems nervous. You notice that he is carrying a hardcover copy ofThe Girl with the Dragon Tattoo. You pick up the book and ask him if he likes it. He now appears even more nervous than before. You notice something odd about the book—the dust jacket doesn’t seem to fit. You remove it and find a different book underneath. How do you feel about this traveler’s demeanor, and the likelihood of his being a terrorist, if the book is:
A. The Qur’an (in Arabic)
B. The Magic Mushroom Grower’s Guide
C. Overcoming Impotence: A Leading Urologist Tells You Everything You Need to Know
If you care more about A than B, C, or D, as I think you should, you are guilty of religious profiling (and calling it “behavioral profiling” doesn’t change this fact).
The funny thing about my “racism” is that I would probably be more concerned if the young man in this example were light-skinned, like me, than Middle Eastern. Why? Because he would have had to make a great effort to learn Arabic. Is there anything intrinsically sinister about learning Arabic? No. I wish I knew Arabic. But it is one more detail that fits the profile of someone who is deeply committed to the worldview of Islam and disposed to conceal that fact. Are all such people terrorists? Of course not. But every person who attempts to blow himself up on an airplane, now or in the foreseeable future, is likely to come from this group. Of course, if that changes, we should alter our view of security accordingly. If the Ku Klux Klan were to declare a broader war on civilization and begin a campaign of suicide bombings, we would have to keep an eye on that profile too (and being nonwhite or Jewish would help smooth your path through security).
More here. And a response by security expert Bruce Schneier:
The right way to look at security is in terms of cost-benefit trade-offs. If adding profiling to airport checkpoints allowed us to detect more threats at a lower cost, then we should implement it. If it didn’t, we’d be foolish to do so. Sometimes profiling works. Consider a sheep in a meadow, happily munching on grass. When he spies a wolf, he’s going to judge that individual wolf based on a bunch of assumptions related to the past behavior of its species. In short, that sheep is going to profile…and then run away. This makes perfect sense, and is why evolution produced sheep—and other animals—that react this way. But this sort of profiling doesn’t work with humans at airports, for several reasons.
First, in the sheep’s case the profile is accurate, in that all wolves are out to eat sheep. Maybe a particular wolf isn’t hungry at the moment, but enough wolves are hungry enough of the time to justify the occasional false alarm. However, it isn’t true that almost all Muslims are out to blow up airplanes. In fact, almost none of them are. Post 9/11, we’ve had 2 Muslim terrorists on U.S airplanes: the shoe bomber and the underwear bomber. If you assume 0.8% (that’s one estimate of the percentage of Muslim Americans) of the 630 million annual airplane fliers are Muslim and triple it to account for others who look Semitic, then the chances any profiled flier will be a Muslim terrorist is 1 in 80 million. Add the 19 9/11 terrorists—arguably a singular event—that number drops to 1 in 8 million. Either way, because the number of actual terrorists is so low, almost everyone selected by the profile will be innocent. This is called the “base rate fallacy,” and dooms any type of broad terrorist profiling, including the TSA’s behavioral profiling.