I recently presented a paper at the GISrUK conference about using a genetic algorithm to calibrate a model and to better understand the behaviour of simulated burglars.
The aim of the work is to use an artificial intelligence algorithm to calibrate the behaviour of virtual burglars, bringing model results more in line with observed burglary patterns. As well as improving the accuracy of the model, the process might tell us something about the underlying human behaviour. The model is built on the ideas of Routine Activity Theory, the Rational Choice Perspective and the Geometric Theory of Crime. If the genetic algorithm finds that certain types of agent behaviour do not help to improve the accuracy of the burglary model, then this might tell us something about the applicability of the theories to the real world.