Simulating team formation in social networks

This research examines the problem of team formation in social networks. Agents, each possessing certain skills, are given tasks that require particular combinations of skills, and they must form teams to complete the tasks and receive payoffs. However, agents can only join teams to which they have direct connections in the social network. We find that a simple, locally-rational team formation strategy can form team configurations with near-optimal earnings, though this greedy hill-climbing search does converge to suboptimal local maxima. Under this strategy, a variety of random graph topologies not only achieve earnings competitive with complete graphs, but also are much more efficient, achieving these results in less time and with far fewer connections between agents. Several variations were tested; the best results for average earnings and equality occurred when groups were allowed to merge and expel agents, and when groups were fully connected during formation.