martes, 27 de marzo de 2012

Validating agent-based behavioral geography


Like all models, agent-based simulations are used to abstract from a more complicated reality, or to develop synthetic laboratories for exploring systems that are not amenable to investigation on the ground. Unlike most models, however, agent-based simulations allow for an unprecedented degree of detail to be introduced to a simulation. This makes it difficult to assess the fit of simulations to the real-world, to plans, to policies, or to theories.
We are developing a set of methods, and related tools, for validating agent-based models of behavioral geography, based on the spatio-temporal properties of information flow in models of complex systems, and of the spacing and timing of agent behavior in simulation. This builds on our efforts to couple agent-based models and space-time GIS, and related projects to examine ways to accelerate agent-based computation.