Invited Speakers

Dr. Gary An, M.D.
Director of the Burn Intensive Care Unit
Cook County Hospital
Chicago, IL

Dr. Gary An is a trauma surgeon in Chicago currently working as the Director of the Burn Intensive Care Unit at Cook County Hospital.  He has been interested in the application of complex systems analysis to sepsis and inflammation since 1999, and worked primarily with using agent based modeling to create mechanistic models of various aspects of the acute inflammatory response.  He is a founding member of the Society of Complexity in Acute Illness, and is also a faculty member of the Center for Inflammation and Regenerative Modeling at the McGowan Institute of Regenerative Medicine at the University of Pittsburgh. He has been a member of the Swarm community since 1999.

Morphology and Modularity: ABM Approaches to Biomedical Modeling
Gary An
Friday, 9:00am, June 23, 2006
McKenna Hall
University of Notre Dame
Notre Dame, IN

Prior to the advent of molecular biology a primary focus of traditional biology was classification based on observed morphological and structural differences.  The melding of biochemistry, cell biology and genetics in the mid-20th century led to a switch in the emphasis of biology towards more formal analysis along the lines of the Newtonian physics-based reductionist paradigm.  While this approach has been, and continues to be, extremely successful in the acquisition of mounds of detailed information, it is now recognized that there are significant limitations to this method.  One is the sheer volume of information that needs to be analyzed and integrated, another is the recognition of the importance of systems-level approaches needed to re-integrate the connectivity lost in the reductionist process.  In the medical field this has led to difficulty in translating the results of basic science research into effective clinical regimens.  I believe that Agent Based Modeling is particularly well suited to this translational role.  The traditional emphasis on classification/morphology is directly applicable to ABM construction, and the inherent modularity of ABMs makes them good platforms for the collaborative efforts necessary in a widely dispersed and compartmentalized research community.  I present here an ABM framework with a modular, multi-scale approach that will hopefully become a prototype platform for a community-wide, open-source, dynamic-functional data base of biomedical information.

Steve Bankes
Evolving Logic/RAND

Steve Bankes is Chief Technology Officer for Evolving Logic Inc., a software firm developing decision support systems for complex and deeply uncertain problems. He is also Professor of Information Science at the RAND Graduate School, where he teaches courses in Agent Based Modeling and Artificial Societies, and Policy Analysis for Complex Systems.  Dr. Bankes is the originator of the computational research methodologies known as "exploratory modeling", which provide a basis for studying complex, adaptive, and incompletely understood systems through computational experiments. And he is the main designer of the Computer Assisted Reasoning system (CARs), a technology that facilitates robust decision support for many important problems in government and industry. 

Dr. Bankes has over 50 publications in a variety of areas including computer science, operations research, global climate policy, sustainable development, computational social science, and neurophysiology.  He received his B.S in Engineering from Caltech, and a Ph.D. in Computer Science from the University of Colorado.  He holds two patents for software technology with two others pending.  He is a recipient of the Barchi Prize, awarded by the Military Operations Research Society.  He serves on the board of directors for the Center for Computational Social Science and the Center for Governance, and is a member of the Center for the Study of the Origins and Evolution of Life, all at UCLA.  His current research interests includes computational science, modeling and simulation theory and practice, complex adaptive systems, machine learning and self-organizing systems, and agent based simulation of social systems.

Robust Inference in Computational Social Science
Steve Bankes
Friday, 12:00pm, June 23, 2006
McKenna Hall
University of Notre Dame
Notre Dame, IN

The world faces profound social, economic, environmental, and technological transitions.  How we choose to meet our challenges stemming global terror, halting the spread of AIDS and other infectious diseases, achieving sustainable development, managing new genetic technologies, etc. -- will resonate throughout the 21st century.  Models of social-political behavior can be informative in understanding these problems, and can be used to anticipate possible future developments and the possible implications of contemplated actions.  Their value can be greatly enhanced by the use of robust inference in the design and analysis of computational experiments using them.  Techniques have proven their utility including the use of co-evolutionary mechanisms to seek in parallel robust conclusions and cases that maximally stress them, and the use of machine learning techniques to infer human interpretable generalizations from the results of thousands or millions of experiments.  These methods harness computation not to solve the intractable problem of predicting the long-term future, but instead to enable a fundamentally different, more sensible question:  Given what we know today, how should we act to best shape the future to our liking?  Our greatest potential influence for shaping the future may often be precisely over those time scales where our gaze is most dim.

Steve Railsback
Lang, Railsback & Associates
Department of Mathematics, Humboldt State University
Arcata, CA

Steve Railsback is an environmental scientist who has been developing individual-based models and applying them to ecological problems---especially, river management---for over a decade. He is a research consultant and adjunct professor with the natural resource modeling program at Humboldt State University (www.humboldt.edu/~ecomodel), and formerly was on the research staff at Oak Ridge National Laboratory. Last year, Steve and Volker Grimm published "Individual-based Modeling and Ecology", the first major monograph on agent-based simulation in ecology

What makes a good individual-based model (and has there ever been one)?
Steve Railsback
Saturday, 9:00am, June 24, 2006
McKenna Hall
University of Notre Dame
Notre Dame, IN

Agent-based (or "individual-based") models are now one of the most commonly used type of model in ecology, yet their contribution to theory and real-world problem solving remains limited. There certainly have been plenty of "bad" IBMs, but what makes a good one?  One criterion for a good model is that it has high "fitness": the model becomes well-known and widely understood, and is reproduced often. Craig Reynold's Boids clearly fits this criterion. But Boids fails a second criterion: a good model lets us learn something important about the individuals and system. Several IBMs, some very similar to Boids, clearly meet this second criterion. But many other models fail this criterion, usually because a cycle of testing alternative theories for individual behavior is not conducted. A third criterion that appears even more difficult to meet is that a good model is used successfully to support management decisions. Some models (including one that the speaker is partly responsible for, which may be the most complex Swarm model ever built) fail this criterion by trying to represent a whole system instead of being focused on a specific management problem. But whether an IBM is used for real-world problems appears to be heavily influenced by history: it is much harder to displace existing models than to fill a an empty modeling niche.

[Home] [CallForPapers] [Registration] [Housing] [Travel] [InvitedSpeakers] [Tutorial] [Schedule] [Committee]