Advancing the State of the Art in Agent-Based Modeling

William Rand  

Wednesdays@NICO Seminar, Noon, October 25 2006, Chambers Hall, Lower Level

Dr. William Rand, Northwestern University

Abstract 

Agent-based modeling (ABM) is still a relatively nascent field, but already it holds promise for helping researchers, students, and the public understand complex systems in new and innovative ways. I will begin with an introduction to ABM and compare it to traditional modeling techniques. After that I will examine recent areas of research within the methodology of ABM. For instance, recently my colleagues and I have been worked on developing new methods which allow us to incorporate richer conceptions of pattern within ABM. By developing a new framework of communication we are working toward integrating social network analysis (SNA) and geographic information systems (GIS) within the overall ABM conception, and by doing so we are creating a common language for these disparate fields. Of course as one builds up more and more detailed ABMs it becomes important to make sure the models are accurate. Computational modeling requires a reassessment of standard scientific methodology surrounding replication, verification, and validation. We have taken advantage of a recent replication of Robert Axelrod's Ethnocentrism model to further develop new research into improving the accuracy of ABM. Finally, most traditional agent-based modeling has focused on virtual agents and their interactions. Recently we have been working on incorporating robotic and human agents into these modeling efforts. This requires the creation of not only new techniques but also new conceptions to begin the discussion of these techniques. I will finish with a demonstration of an early prototype that incorporates human and robotic agents.