The Perfect Swarm

Computerworld - Biologists and political scientists are just part of the growing brigade of natural and social scientists who are going against convention and embracing the concept of agent-based modeling.
While it's still far from the scientific mainstream, agent-based modeling is rapidly gaining acceptance in disciplines as diverse as economics, ecology and defense. And software - most notably a program called Swarm - has been fueling its development.
Agent-based modeling allows researchers to create artificial worlds that model activity in the natural world.
Using Swarm, researchers design a series of rules to govern those worlds and then send players, called agents, to live under those guidelines. The researchers in essence create their own games, then use those games to draw conclusions and test theories.
Planned Chaos
The scenarios are often chaotic and unpredictable, but they also give insight into the behavior of societies, economic markets or natural activities. Such insight may be used to help an economist determine how a market will behave, help a political scientist predict voting patterns or help a biologist simulate bacterial growth
"What's really exciting about Swarm is that it's useful to people in a lot of different disciplines. The cross-platform nature of it is fun," says Nelson Minar, a member of Swarm's original two-man design team and the founder of Popular Power Inc., a San Francisco-based technology company. "It's a major paradigm shift in how people understand our world. It provides a cross-disciplinary understanding of artificial phenomena."
Minar says Swarm has led the agent-based modeling charge because it has cross-disciplinary applications. And it has cross-disciplinary applications because it was developed at the Santa Fe Insitute (SFI) in New Mexico, a private, nonprofit, research and education center that aims to foster interdisciplinary study.
SFI began developing Swarm in 1994 to give scientists a practical tool for creating artificial worlds. The software is still considered experimental, but it has been useful to scientists who are scratching the surface of agent-based modeling.
Those charged with developing Swarm are still learning its capabilities. For example, Marcus Daniels, the executive director of SFI's Swarm Development Group, has spent two years trying to make Swarm accessible to what he calls "a scientific community that isn't comfortable with computers."
"Swarm is still considered very hard to use," says Daniels, who has created a CD-ROM version of the program and is creating a system that will allow users to access the Swarm model through a Web browser. "There is an infrastructure that people have to get used to."
Less Than Accessible
Users say Swarm's