Abstract
Modeling terrorism is both necessary and difficult. While the necessity comes from the all too obvious real-world pressures our society is facing, the difficulty stems from the underlying complexity of the phenomena itself - there are many variables to account for, they are hard to measure, and the relationships between them are confounding.
Since modeling terrorism is at its most onerous when it comes to predicting specific attacks, their timing and scale, we opted to work around this using observed probabilistic distributions and integrate power laws into our SD model. After evaluating thousands of simulations runs, this allows us to replicate historical data as well as produce prognostic scenarios, while maintaining what we believe to be authentic behavior.
Compromises need to be made, but we believe that this approach can be useful for systems highly dependent on events or parameters which we are unable to predict but whose distributions are known.