Andrew Kydd has a “short paper”:http://www.stanford.edu/group/ir_workshop/kydd.pdf that uses game theory to analyze the benefits and disadvantages of ethnic profiling at airports etc. He starts with analysis of a simple Matching Pennies game (which I’ve always had in the back of my head as the most obvious simple way of modeling this process; I suspect the same is true of many people who have (a) taken a course in basic game theory, and (b) thought about profiling tactics), and moves on to a more complex specification. On the basis of a Colonel Blotto game, Kydd argues that some degree of ‘seemingly unfair profiling’ is likely to result from rational strategies employed by actors on both sides, except under relatively unlikely parameter values. He concludes that:
Profiing in the context of strategic terrorist groups is rational, when considered solely on the basis of what will stop more terrorist attacks. Other factors may constrain the United States from profiling, but the logic of interaction with terrorist groups suggests a powerful dynamic in its favor.
While the paper is still a work in progress, I worry that this is one of those cases where game theorists reach confident conclusions on the basis of _ceteris paribus_ conditions, where the ceteris are anything but paribus if you look at them closely. Entirely apart from the ethical implications (which he seems interested in exploring in future iterations of the piece), the conclusion that profiling will stop terrorism rests on the assumption that increases in profiling of specific ethnic groups won’t have any consequences for future recruitment of terrorists from those groups. There’s an array of circumstantial evidence that the perceived humiliation that targeted searches and other such measures involve have significant implications for relations between the targeted group and other members of the population.
Jack Knight and I have some unpublished work where we try to tease this out in the context of trust relations – we argue that those who feel targeted by measures of this sort are rationally going to be more distrustful of those who belong to more privileged sections of the population for a variety of reasons. This may obviously have consequences for the ease of recruiting those aggrieved members to terrorist organizations. Obviously, there’s no good way of measuring the extent of this effect, assuming it exists. But equally obviously, this and related interaction effects mean that one should be quite careful before drawing broad conclusions about the merits of policies on the basis of simple game theoretic models – the unmeasured knock-on consequences may very likely swamp the specific effects that your model is designed to single out.