The ART of IAM: The Winning Strategy for the 2006 Competition

In many dynamic open systems, agents have to interact with one another to achieve their goals. Here, agents may be self-interested, and when trusted to perform an action for others, may betray that trust by not performing the actions as required. In addition, due to the size of such systems, agents will often interact with other agents with which they have little or no past experience. This situation has led to the development of a number of trust and reputation models, which aim to facilitate an agent's decision making in the face of uncertainty regarding the behaviour of its peers. However, these multifarious models employ a variety of different representations of trust between agents, and measure performance in many different ways. This has made it hard to adequately evaluate the relative properties of different models, raising the need for a common platform on which to compare competing mechanisms. To this end, the ART Testbed Competition has been proposed, in which agents using different trust models compete against each other to provide services in an open marketplace. In this paper, we present the winning strategy for this competition in 2006, provide an analysis of the factors that led to this success, and discuss lessons learnt from the competition about issues of trust in multiagent systems in general. Our strategy, IAM, is Intelligent (using statistical models for opponent modelling), Abstemious (spending its money parsimoniously based on its trust model) and Moral (providing fair and honest feedback to those that request it).