Skip to Content

Search: {{$root.lsaSearchQuery.q}}, Page {{$root.page}}

Complex Systems Seminar<br>Query-Based Model Exploration: Parameters and Paradigms<br><b>Speaker: Forrest Stonedahl (Centre College)</br></b>

Tuesday, November 15, 2011
12:00 AM
411 West Hall, Ehrlicher Room

Speaker: Forrest Stonedahl (Centre College)

In the past few decades, agent-based modeling (ABM) has emerged as a powerful computer simulation technique in which many agents interacting according to simple rules can give rise to complex aggregate-level behavior. However, as ABM is increasingly employed in both the natural and social sciences, the methods and tools for understanding, exploring, and analyzing the behavior of agent-based models have not kept pace. I see this as a large outstanding problem (and challenge) for the ABM research community.

In this seminar, I will provide an overview of my dissertation research, which addresses one aspect of this challenge: the exploration of model behavior as model parameters are varied. Specifically, I will report on a comprehensive investigation of the use of genetic algorithms (and other metaheuristic search algorithms) for exploring the range of behaviors produced by agent-based models. This investigation is comprised of a series of in-depth case studies (including models of collective animal motion, viral marketing in social networks, ancient Puebloan civilizations, and online news story consumption), as well as the development of a broader benchmark suite of exploration tasks. More generally, I will argue for the benefits of a paradigm shift in exploration methodology. However, the success of this paradigm shift depends on new tools. Thus, I will also discuss the design of BehaviorSearch, a software tool that I have developed to support this methodology. This work has important implications for the calibration, verification, and validation of agent-based models, and I hope that this and similar projects will nudge (or entice) ABM practitioners toward more robust model analysis, and ultimately result in better science.