Social Networks and Multi-agent Organizational Performance.pdfVIP

Social Networks and Multi-agent Organizational Performance Matthew E. Gaston and Marie desJardins Department of Computer Science University of Maryland Baltimore County mgasto1, mariedj@ Abstract dynamics (Strogatz 2001; Watts Strogatz 1998). The con- tribution of this paper is an empirical demonstration, in sev- In studying the dynamical behavior of processes in artificial eral different domains, that network structure has a signif- or natural social systems, a key factor is the topology of the network structure. It has been shown that real-world social icant effect on the dynamics of agent systems. The three networks tend to have non-random network structure with systems examined are innovation diffusion in an artificial properties such as short average path length, excess clus- agent organization, opinion formation among networked ar- ´ tificial agents, and on-line multi-agent team formation. Our tering, and skewed degree distributions (Albert Barabasi 2002; Newman 2003). We show in this paper that the nature findings are relevant for both artificial societies (Carley of the network structure in a social network of artificial or Gasser 1999; Epstein Axtell 1996) and multi-agent sys- simulated agents has a significant effect on the performance tems (Huhns Stephens 1999; Lesser 199