Using Joint Responsibility to Coordinate Collaborative Problem Solving in Dynamic Environments

Book One

Volume

Proceedings of the AAAI Conference on Artificial Intelligence, 10

Multi-Agent Coordination

Joint responsibility is a new meta-level description of how cooperating agents should behave when engaged in collaborative problem solving. It is independent of any specific planning or consensus forming mechanism, but can be mapped down to such a level. An application of the framework to the real world problem of electricity transportation management is given and its implementation is discussed. A comparative analysis of responsibility and two other group organisational structures, selfish problem solvers and communities in which collaborative behaviour emerges from interactions, is undertaken. The aim being to evaluate their relative performance characteristics in dynamic and unpredictable environments in which decisions are taken using partial, imprecise views of the system.