Sunday, July 24, 2005

On profiling and game theory

Two sources I've encountered this weekend were this paper, which uses 'quality control' logic to show (among other points) why profiling is effective, and Bruce Schneier's post which raised the legitimate point that profiling based on some attributes (say, ethnicity) will motivate terrorists to masquerade those attributes (e.g. dye their hair blonde). However, this possibility does not mean that profiling is ineffective. Since the opponents have finite resources, and since masquerading has an associated cost (for example, the supply of likely terrorists might be reduced if they would need to take accent-reduction classes, or learn to convincingly masquerade their origin.
It is easy to see that while the approach of ‘only inspect people with suspect attributes’ is ineffective, equilibrium (at max probability of screening) is reached when the screening strategy chooses people of ‘profiled’ attributes more often than people without it.

This is really common sense, however.

Possibly in a future post I will run some simulations with fictitious numbers to demonstrate how the numbers work out.


Anonymous akthe47 said...

fascinating. You have some deep posts which are quite insightful. Good stuff, man.

10:46 AM  

Post a Comment

Subscribe to Post Comments [Atom]

<< Home