Sunday, October 15, 2006

Agent based modeling of terrorists

This weekend I got a chance to read one of the IEEE papers titles ‘Modeling terrorists’ which was sent by my cousin Shanmuga.

Predicting and preventing terror attacks has become a very essential task for any country today. There is a traditional ‘analyst’ approach using which these terrorist attacks can be predicted. This analyst approach includes spying, tracking people using satellites, using detectives and tapping various signals like telephone, mobile phone and internet traffic. This analyst approach requires lots of people to perform these tasks and churn out huge amount of data to come up with predictions. What if we can model people’s behavior using software and predict their actions? How about using a simulation program which can help predict how a particular group of people are behaving at this moment? What about a computer model which can help intelligence agencies to unleash the terrorist networks? This is exactly what couple of Universities in the US is trying to achieve.

Basically Artificial Intelligence (AI) based on pre-defined facts and actions. Say for example a robot can be modeled with a set of commands and actions. This particular robot doesn’t learn anything on its own because of which its actions can be predicted. Human beings are far more complex where every human being reacts differently to the same situation. To put it in different terms for the same ‘Stimuli’ there will be different ‘Responses’ in the system. This is mainly because our actions are bound by individual personality, emotions and culture. These factors along with our memory (what we have learned previously) creates an expression inside us which results in a particular action or response. Since these factors vary from person to person the response is totally unpredictable.

Professor Barry Silverman’s group in University of Pennsylvania Kathleen Carley’s group at Cornell University is working on creating this model. The former has taken the ‘inside-out’ approach in which he has created a ‘simulated world’ with set of people living in a particular location using ‘intelligent’ agents and predicting their behaviors. Later taken an approach in which the ‘network’ of people is created and conclusions are drawn based on ‘collective’ approach. Of course these simulations cannot be always true but can provide significant insights into predicting terrorist attacks.

I really enjoyed reading this article!

1 comment:

Tom said...

Lars Erik Cederman ( http://www.icr.ethz.ch/people/cederman/ ) has always had interesting models. Here is a list of some publications by Cederman and others on the topic of wars and terrorism: http://www.icr.ethz.ch/publications/by_type .