Thursday, November 3, 2011

Gaming LeT

The TerrorWonk has been busy lately, so I haven’t been blogging much (although I am going to try to turn that around for NaBloPoMo . One of the things I was busy with was a pair of papers on Lashkar-e-Taiba that were presented at EISIC/OSINT conference in Athens in September. One paper used SOMA to analyze the behavior of LeT. SOMA is a modeling system developed at UMD, we’ve gotten some interesting findings from it looking at other terrorist groups, such as Hezbollah. I’ve written a fair amount about SOMA and more is coming on LeT.

The other paper was a game theoretic analysis of LeT. This was a new area for yours truly and as the subject matter expert I developed the scenarios and payoff matrix (ie what are the different combinations of moves the different players could make and how happy (or unhappy) in each combination is each player. Basically the study found that situations where LeT’s best option was to disband its armed wing were situations in which the US and India double-teamed Pakistan so that the military cracked down hard on LeT. Several papers in India interviewed us and discussed our work (see the LCCD homepage for links), but of them The Telegraph of Calcutta article included this very nice graphic that nicely encapsulates the project.


Two particularly interesting things struck me (as a novice to game theory) about this project. First, in game theory the players seek the Nash equilibrium (named for the Nobel prize winning mathematician) in which no actor can increase their payoff without causing a decrease in some other player’s payoff. But our work included “mixed equilibria” in which players did not simply adopt one strategy but shifted between strategies. This better reflects how nations act. Sometimes, nations systematically switch between policy options. In other cases different components of the state pursue different strategies – some elements of the Pakistani military crackdown on LeT while others continue to provide support.

The other point is that one criticism is that it did not include several key players such as China or Pakistani public opinion. True enough – but in some regards that only strengthens the overall concept. That is, with five actors and 13 possible actions between them there were hundreds of possible combinations. More players and actions means even more combinations – more than a person can systematically analyze.

So, as I’ve written before, models can’t necessarily replace human judgments but by systematically analyzing enormous combinations of data and scenarios they can identify possibilities that humans might miss.

No comments: