Tuesday, August 09, 2005

 

Agent-based disaster management

Penn State's Dr. John Yen and the other folks at the Laboratory for Intelligent Agents have come up with an interesting concept: using intelligent agents to improve team communication for high-pressure situations. The R-CAST system is designed to eliminate bottlenecks in information flow, according to the press release.
It's inspired by recognition-primed decision making (aka RPD), which at first glance looks interesting. RPD a way of modeling how people react to situations under stress. Basically, RPD figures that people try to fit the current situation into previously-experienced prototype situations by feature matching. If any anomalies come up, they try to fit it into another one.
It matches some of my observations of the mistakes we make under pressure: we inappropriately discard anomalies. The hardest part in the early stages of a is recognizing the shape of a situation. People have a limited set of previous experiences that they can use for comparison. Since disasters are, pretty much by definition, exceptional events, we're not very good at recognizing them in the very early stages-- the times when intervention can do the most good.

Comments: Post a Comment

<< Home

This page is powered by Blogger. Isn't yours?