Role evolution in open multi-agent systems as an information source for trust

Abstract

In Open Multi-Agent Systems (OMAS), deciding with whom to interact is a particularly difficult task for an agent, as repeated interactions with the same agents are scarce, and reputation mechanisms become increasingly unreliable. In this work, we present a coordination artifact which can be used by agents in an OMAS to take more informed decisions regarding partner selection, and thus to improve their individual utilities. This artifact monitors the interactions in the OMAS, evolves a role taxonomy, and assigns agents to roles based on their observed performance in different types of interactions. This information can be used by agents to better estimate the expected behaviour of potential counterparts in future interactions. We thus highlight the descriptive features of roles, providing expectations of the behaviour of agents in certain types of interactions, rather than their normative facets. We empirically show that the use of the artifact helps agents to select better partners for their interactions than selection processes based only on agents' own experience. This is especially significant for agents that are newcomers to the OMAS. Copyright © 2010, International Foundation for Autonomous Agents and Multiagent Systems (www.ifaamas.org). All rights reserved.