Social Structure in Artificial Agent Societies: Implications for Autonomous Problem-Solving Agents * Sascha Ossowski 1 and Ana García-Serrano 2 1 School of Engineering, Rey Juan Carlos University at Madrid, Campus de Móstoles s/n, 28933 Móstoles (Madrid), Spain S.Ossowski@escet.urjc.es 2 Department of Artificial Intelligence, Technical University of Madrid, Campus de Montegancedo s/n, 28660 Boadilla del Monte (Madrid), Spain agarcia@dia.fi.upm.es Abstract. In Distributed Problem-solving systems a group of purposefully de- signed computational agents interact and co-ordinate their activities so as to jointly achieve a global task. Social co-ordination is a decentralised mechanism, that sets out from autonomous, non-benevolent agents that interact primarily to improve the degree of attainment of their local goals. One way of ensuring the effectiveness of social co-ordination with respect to global problem solving is to rely on self-interested agents and to coerce their behaviour in a desired di- rection. In this paper we model the notion of social structure for a particular class of multiagent domains, and determine its functionality with respect to so- cial co-ordination. We show how social structure can be used to bias macro- level properties in the frame of multiagent system design, and discuss micro- level implications respecting the architecture of autonomous problem-solving agents. 1. Introduction Distributed Problem-Solving (DPS) relies on a purposefully designed architecture of computational agents that interact in order to achieve jointly a desired global func- tionality. The traditional DPS design philosophy of reductionism, that relies on a top- down decomposition of the global task, the assignment of subtasks to agents and co- ordination based on pre-established interaction patterns among benevolent agents, often turns out to be too rigid for large-scale agent systems [8]. Instead, a construc- tionist approach, based on the metaphor of societies of autonomous problem-solving agents, has become popular: agents are primarily interested in their local goals and interact to increase the degree of their attainment. This decentralised interaction [5] is termed social co-ordination. In order that the DPS system copes with the global task, * This work was partially supported by the Human Capital and Mobility Program (HCM) of the European Union, contract ERBCHBICT941611. J.P. Muller et al. (Eds.): ATAL’98, LNAI 1555, pp. 133-148, 1999. ..  Springer-Verlag Berlin Heidelberg 1999