Managing such complex systems as the stock market or a battlefield is particularly challenging because each agent in the system chooses how to behave, not only by following the rules of the game but also by predicting how all the other agents around them will behave and adjusting their own actions accordingly.
“Decentralized Control in a System of Strategic Actors,” an August 16-18 workshop at SFI, will draw on experts from a number of fields to explore ways to manage multiple-player systems.
“The agents in the system are all using ‘I know what you know what I know’ kinds of reasoning, and we have to try to control behavior starting from that,” says NASA senior computer scientist David Wolpert, who is collaborating with SFI Professor Eric Smith to organize the conference.
Creating something that can take so many factors into account with as little error as possible is a worthwhile undertaking. Beyond the stock market, he says, it could have broad uses for the management of complex systems like the power grid, air traffic, and national economies.