Promoting cooperation of agents through aggregation of services in trust models

Abstract

In agent systems, cooperation plays a key role that can only be achieved trusting other agents. However, in many application areas of agents, the lack of enough frequent direct experiences is a major problem to find the right partner to cooperate with. The exchange of indirect information such as recommendations about third parties intends to address this problem. Since an increase of recommendation exchanges will produce more accuracy in trusting decisions, any way of promoting the exchange of recommendations would be welcome in order to extend the use of Trust models. In this paper, we present a method of exchanging recommendations to be used by such trusting agents which pursues an increase of cooperation (in the form of more recommendations exchanges). We used ART testbed domain as an illustrative example of application, where agents act as providers and consumers of painting appraisals, and we propose the pair of interacting agents to agree about a joint pair of reputation requests that would immediately satisfy both agents instead of expecting future benefits of sharing reputation in an altruistic way. We show under what circumstances (different parameter setups) an implementation of our proposal behaves better (in terms of quality and cost) compared to no-recommendations and the way recommendations take place in ART Testbed.