Terrorist Profiling - When is it really ethnic profiling?

In the inimitable words of that great philosopher Forrest Gump, "Stupid is as stupid does.”

An excellent example of this is found in the following statement:

Between 1970 and 2005, there were about 60…notable examples of terrorism perpetrated by Middle Eastern male Islamists between the ages 17 and 40, but we think you get the point. Singling out " Middle Eastern male Islamists between the ages 17 and 40" is not ‘ethnic profiling,’ it's ‘terrorist profiling’ acting on prolific evidence.

[An internet search of this statement will find it copied and pasted onto multiple web sites.  We first saw it on the web site for a radio show.]

Can you figure out what’s wrong with it?  We hope so.

But for those who may be a little hazy about the fallacy in the above statement, here’s some help. The “profile” is far too limited by being way too general and inclusive. And since two out of the three criteria apply to an ethnic group, it does indeed turn out to be ethnic profiling.

Ask yourself, how many people fit all three (age, region of origin, and religious affiliation)?  Lots. Probably millions. So it’s useless. It functions merely as a stereotype and an ethnic profile—it doesn’t profile a single terrorist.  Maybe a visual aid can help make this clear:

 What can make a profile useful?

Profiling, in and of itself, is not the problem here.  What's a problem is when people use inappropriate criteria in a profile, in this case criteria that are way too broad.  Here's what's necessary for a valid and useful profile:

For a terrorist profile (or any profile) to be acceptable and beneficial, a statistical link must be shown between the defined characteristic(s) and the risk of terrorism. The stronger the statistical link, the more valid and discriminative the profile becomes.

The statistical link increases when:

a. There are more characteristics.  Three is far too few.

b. The characteristics are narrower.  The three used above are way too broad.

So, while there is evidence that many acts of terrorism have been perpetrated by Middle Eastern male Islamists between the ages of 17 and 40, these individuals (the ones who were terrorists) represent such a small portion of that very large population of similar individuals that the statistical link for risk is nearly nonexistent (as in, possibly less than one in 100 thousand).

To make a comparison, consider than 99% of rapes are made by males, but it would be pointless (some might even say “stupid”) to use gender alone as a “rapist” profile, since around half the population consists of males, very few of whom rape anyone. The statistical link between gender alone and risk of rape is negligible. So, yes, gender may be ONE characteristic in a rapist profile, but there are obviously several more that help narrow the field and increase predictive validity.

The same must be done with trying to identify those who might be terrorists. The three suggested above do not narrow the field sufficiently to be of any use all by themselves, except as an ethnic profile, which is why it is legitimate to question the potential for racism to underlie their use (as in, “I just plain don’t like, or trust any of those male Islamists between the ages of 17 and 40 anyway, so singling all of them out for scrutiny doesn’t bother me one one teeny, tiny bit”).

Done appropriately, profiling can be a useful tool

People of good sense do not question whether it is appropriate or valuable to arrive at some sort of picture of individuals who might be in the process of preparing for and/or committing terrorist acts (or any criminal activity, for that matter).  But they know better than to base it upon three such characteristics.

Are the additional characteristics already available.  Sure.  Do an online search for "characteristics of terrorists," and you'll find lots.  The Los Angeles Police Department lists nine.

As noted above, the more characteristics that can be developed and statistically linked to risk of terrorism, the more useful the profile becomes—it achieves what statisticians call “discriminative validity” (it can descriminate between those who might be terrorists and those who might not) which increases the potential “predictive validity” (being better able to predict who might commit a terrorist act).

For those who might be interested In statistics, there are several tools that can be used to put together a more valid profile.  For example, the use of multiple variables to predict the probability of a particular outcome is known as multiple regression analysis, and the use of multiple variables to classify (in this case into "probable terrorist" and "not probable terrorist") is called discriminate function analysis. The point being, profilers have some pretty sophisticated statistical models for putting together a profile with multiple characteristics (variables).

While the general public is not familiar with the more abstract aspects of stochastic (i.e., probabilistic) models of analysis and prediction, it does not take excessive brain-power, or knowledge, to recognize that using broad variables (such as the three suggested above) has minimal discriminative, or predictive validity.

And that’s been the objection to such “profiling.” Anyone who is serious about understanding what the issues are with “profiling” will take a few minutes to inquire, instead of engaging in such a ludicrous “straw man” argument (e.g., arguing against those who oppose ethnic profiling as if they are opposed to any and all attempts to identify those at risk to commit terrorist acts).

 To go to the brief introduction to International Issues