Agent-Based Models

Scott de Marchi, Scott E. Page
2014 Annual review of political science (Palo Alto, Calif. Print)  
The most pronounced impact of agent based modeling on economics may be that it enlarges the set of questions we can explore. (Anderson, Arrow, Pines 1988 , Arthur, Durlauf, Lane 1997 . Agent based modeling allows us to consider richer environments with greater fidelity than do existing techniques (Tesfatsion 1997) . This increased fidelity results from the inductive nature of the modeling enterprise. When constructing an agent based model, we are constrained only by our imagination and
more » ... In contrast, when constructing a mathematical model, we must always be aware of what we can prove of what is tractable. This constrains our endeavors. The set of models that one believes to be tractable is small when compared against the set of models worth exploring. By freeing us from considerations of provability, agent based models allow us to focus on those aspects of the world that we believe most relevant. These assumptions are then encoded in a computer program and allowed to iterate their logic. Some models produce a chaotic mangle, and the assumptions need to be rethought. But often, agent based models produce interesting results. And often, these results can be understood analytically, even proved. Given that it is far easier to prove a result when the answer is known than when the answer is not known, agent based models are a powerful engine for generating insights. But they are far more. * I would like to thank Ken Kollman and Rick Riolo for comments on earlier drafts.
doi:10.1146/annurev-polisci-080812-191558 fatcat:mxwyihzap5hj5beregqfkneffy