Agent-Based Computational Laboratories
Dr. Catherine Dibble
Boston University, 8-10 August 2003
Abstract
This workshop is for anyone interested in agent-based research
strategies. The workshop emphasizes computational laboratory modeling
and science. Research thinking and interests are far more important
than programming skills for this workshop. Nontechnical researchers are
welcome and encouraged to attend.
After a brief historical overview of agent-based approaches and
platforms, the workshop will focus on RePast (from U Chicago Social
Science Research Computing, a 2nd generation clone of Swarm that has
especially strong support for social science modeling), GeoGraphs (which
allow for agent simulations on network landscapes, including small-world
and scale-free networks), and Genetic Algorithms (which mimic natural
evolutionary processes to find adaptive solutions to highly complex
problems, and which work as complements to agent-based simulations at
several levels), with an emphasis on model design, experimental design,
and on how agent-based computational laboratories complement other
approaches to research. Application examples will illustrate the
utility of this research approach in exploring topics such as the
evolution of organizations, epidemiology, settlement patterns,
globalization processes, the effects of social and spatial structures on
the evolution of conflict and cooperation, and the evolution of
inequality. Handouts will be provided, along with a web site from which
to download and install computational laboratory software and related
development software.
The instructor, Dr. Catherine Dibble, is an economic geographer
at the University of Maryland. She has decades of experience and
training in evolutionary systems and scientific method, formal economic
theory and game theory, computer science, theoretical geography, and
especially computational laboratories. She has been working
professionally with simulation models of many types since 1980, with
genetic algorithms since 1993, and with agent-based simulations since
1995. Her publications so far cover new designs for handling spatial
structure and solving location-allocation problems using genetic
algorithms, for representing absolute and relative space and time in
genetics based machine learning, and for genetic evolution of optimal
organizational designs under disparate conditions. She is the inventor
of the general purpose GeoGraph library for Swarm and RePast, which
supports the construction of agent-based models on richly structured
network landscapes such as organizations, social institutions, and
geographic landscapes. Dr. Dibble teaches advanced PhD seminars in
Computational Laboratories and leads an active Computational Laboratory
research group at the University of Maryland.
|
