AAMAS-04 Workshop on Learning and Evolution in Agent Based Systems 1 Adapting Network Structure for Efficient Team Formation

The concept of team formation is central to a wide vari-ety of disciplines, including planning and learning in multi-agent systems, artificial social systems, and distributed arti-ficial intelligence. Typically, models of agent-based systems do not focus on the nature and structure of the interconnec-tion network that dictates agent interaction, although recent findings suggest that real-world networks of many different types have rich, purposeful, and meaningful structures. Us-ing a simple agent-based computational model of team for-mation, our previous work suggests that organizational net-work structure can have a significant effect on the dynamics of team formation. We present several strategies for locally adapting network structure in a simple team formation sce-nario, and give empirical results that show that such adap-tation methods are capable of significantly improving orga-nizational efficiency. These methods prove to be especially useful for adapting agent networks in the presence of at-tack faults that target the most highly connected nodes in the network