The flaw in “racial profiling” for terrorists

Watching old footage of Ann Coulter turned up an argument heard in this country many times since 9/11. Coulter begins by citing patterns in the terrorist demographic. She lists bombing attacks in which Americans have died, concluding, “The perpetrators have all had the same eye color, hair color, skin color and half of them have been named Muhammad…This is not racial profiling; it’s a description of the suspect.”

Of course, she is being characteristically cheeky. She advocates “racial profiling” by name on the basis of this demonstrated pattern. In short, a terrorist is more likely to come from x-racial-group than from y or z groups; therefore, we are warranted in searching for terrorists among that group particularly. This might include singling out men fitting this description for random baggage or ID checks in subways, or funneling them through a separate check-in line at airports.

For the sake of argument, let us assume that Coulter’s data on terrorists is correct (though it isn’t); it is the form of the argument that is most suspect. It gives too much weight to the (statistical) relationship of racial groups to one another—that is, to the relative percentages of terrorists within each group. If we’re thinking properly, what should matter instead is the relationship of terrorists to their own racial groups.

The important point is not what percentage of terrorists are ‘middle eastern males with funny sounding names,’ versus males of some other group—but rather, what percentage of ‘middle eastern males with funny sounding names’ are actually terrorists.

Analogy #1: For all I know, men named George are .000003% more likely to be serial killers than men with other names. This would this hardly mean that shaking down a bunch of Georges would be a wise deployment of police resources. Shit, even if 100% of serial killers were named George, those Georges who actually commit serial murder are such a tiny minority among men named George that the strategy would still be suspect.

Analogy #2: Imagine we have a haystack which has some probability of containing a needle. (That is, there is some probability that one of the straws is a needle.) Let this probability match that of a given, random Arabic man’s being a terrorist; drawing a random straw is as likely to yield a needle as “drawing” a random Arabic man is likely to yield a terrorist. Let us assume this method of finding needles is ineffective, counterproductive, even immoral; also, that we have some far better method of finding needles in haystacks—using magnets, X-ray, floating the straw on water so the needle sinks, etc. We still want to root out needles, but have long abandoned the strategy of drawing random straws.

Now, imagine we discover that all along there has been a second haystack nearby which has an even lower probability of yielding a needle than our haystack. Perhaps we discover several more, each with some probability lower than the original, but still more than zero. It has become clear that a needle is more likely to come from the first haystack than from any of the others. Still, it would be irrational in the highest to conclude that we should, on this basis, resume our random straw draws. The simple fact that a less promising haystack exists does not magically make checking this stack a good idea, if it wasn’t a good idea before.

Similarly, the simple fact that terrorists are more likely to come from middle eastern men than from some other group doesn’t mean that the likelihood of randomly finding them among middle eastern men is very good at all.


The obvious question is just how good that likelihood is. I haven’t exactly crunched the numbers; you can do the math if you like. But there are millions of men in the world who fit Coulter’s “profile” and very nearly zero of these commit terrorist acts against Americans. Even fewer do so in those stereotypical ways that profiling would address. Even fewer operate in the U.S., where ours laws can actually penetrate. Clearly, we are dealing with numbers akin to those Georges who commit serial killing. It is quite likely that if we incarcerated every other Muslim male in the world, it would register nothing in practical terms to diminish the odds of the next terror attack. Yes, we can theoretically halve a .000003% chance of something. Getting married later in life will halve one’s chances of committing suicide someday. Buying a second ticket will double one’s chances of winning the lottery. There is shit you could do right now to halve your chances of being brainwashed by a cult or eaten by a mountain lion. Who gives a shit? Differences of this infinitesimal grade should no more drive policy than they drive anybody’s consideration of anything else in the real world.

This is not to mention that radical Islamists come in all “colors” and (duh) will easily work around any profile we make. Plus, racial profiling is counterproductive; it alienates those communities which are most critical for intelligence on the potential attackers that move and live among and gain cover from them. “Profiled” individuals tend to avoid law enforcement as much as possible.

Finally, as with finding needles in haystacks, we have a much better alternative strategy for fighting terrorism. Granted, jihadists will cite a number of gripes against the US if you ask them. Some of these concern cultural factors like our women’s liberation and sexy music and movies. But according to the evidence, these aren’t the “root” reasons they turn to terror. As I’ve argued elsewhere (link below), the violence is a response to US foreign policy in and toward Muslim countries and populations. Thankfully, these concerns are quite reasonable, technically solvable, and are morally “overdetermined”—that is, they should be met for a host of reasons even aside from fighting terror.

[See the last boldface section of the post here.]


4 responses to “The flaw in “racial profiling” for terrorists

  1. I agree that profiling is a bad idea; but your argument is flawed. In your analogy about killers named george, there is no connection between a person’s name and whether or not they kill. Any pattern would be coincidental. Conversely, some Islamic extremist groups advocate for the killing of Americans. People from certain countries, of certain religions are being asked to kill Americans, so the connection between these groups and their desire to kill people is not coincidence as it would be with the name George.
    Profiling does not work though because it has been found that terrorist groups are actively seeking out people to commit terrorist acts who do not “fit the mold” and hold US visas (among other things). Thanks for your thoughts.

  2. Joe,

    Thanks for reading. I am sympathetic to your point here. That is, I think the “George” example is probably a bad analogy. It is misleading. However, I think it is a valid one, that is, it does not make for a formally invalid argument. My mistake is poor writing, not poor logic.

    But first, I simply do not know if a particular name is connected to serial killing in a non-random way. This would have to be argued, not merely assumed. Maybe “George” (or some name) is popular in certain ethnic/age groups among which serial killers are overrepresented. The idea is not that crazy; again, people like Coulter argue that “Muhammed” is preponderant in those ethnic groups from which anti-US terrorists are typically drawn. It becomes less crazy as we are dealing with a set with very few members; just a few more or less of the name is all it would take to register in the stats. Or some other causal story might hold.

    However, I will assume for the sake of argument that your point is correct—that “any pattern [between a name and propensity to serial kill] would be coincidental.”

    As you see it, my analogy is flawed because it assumes the connection is not coincidental. That is, it assumes something which is not actually true. But since when can we not use counterfactual scenarios in analogies? A parent might tell a child, “Don’t steal from your sister, because—What if I stole something from you?” An analogy is drawn between the child’s stealing from his sister, and the parent’s stealing from him. It is not invalid just because it is false that the parent ever stole from the child, right? Nor is mine invalid just because it is false that Georges are non-coincidentally connected to serial killing.

    The argument really breaks down as follows: Profiling for Middle Eastern men would be like profiling for Georges in a world in which this name had a non-random propensity to serially kill. This is not our world, maybe, but who cares? That doesn’t affect the logic of the argument.

    Logic aside, yes, I should have picked a better analogy from a stylistic-persuasive point of view. There are too many examples of similar statistical connections which hold in the present world to make shit up. I suggested a couple of them at the end of the essay (e.g. marrying late in life as connected to suicide rates) but did not expand these as with “George.”

    I don’t feel incredibly “strong” about my response here, man. I could be overlooking something. I’m not a prob-stat guy. And again, I agree with what I take to be the basic thrust or “spirit” of your point. So I don’t want to contest too much. But this is what I came up with. Also, per your last paragraph, I hope you caught that I too mentioned the effort of terrorists to work around Coulter’s profile by recruiting from other ethnic groups. I just didn’t make that the focus of my critique.


  3. If it walks like a duck……….You call it profiling. I call it a description. Wake up, sheeple.

  4. Duncan–Yes, “profiling” is a kind of “description.” (Duh.) So I don’t get your distinction there. Certainly, it isn’t clear how it contradicts anything I’ve said. (That is, it isn’t clear how it could be a response to anything I’ve said.) The question–rather–is whether applying this particular “description” in the way we call “profiling” is productive as opposed to counter-productive. I’ve done the work of providing an answer to that question. If you have a counter-argument, I’d love to hear it. But what you’ve provided so far is piss. Tell us how employing the kind of “description” you have in mind is actually worth the effort. Respond to the goddamned argument. Or fuck off.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s