Using contextual knowledge for trust strategy selection

Abstract

In open multiagent systems in which some of the agents may be self-interested, it is vital that an agent be able to make trust decisions about its peers to determine which of them may be untrustworthy and how to behave in response. The agent’s context, including the environment, observed and inferred actions and motives of other agents, and properties of the MAS as a whole, is critical to making an informed trust decision and especially to choosing a strategy for taking actions. However, most prior work in the area has ignored context or only treated it implicitly. In this paper, we present an implemented approach that explicitly represents the agent’s context, informed by known contexts, and that uses that contextual knowledge to select the best strategy, even in the presence of untrustworthy agents.