Intelligent trust evaluation and self-organizing multi-agent community
Wan Huang, David W. Cordes · 2004
This research addresses the trust relationships that are established between agents during their interactions in peer-to-peer electronic trading. In such situations, centralized control is not available and decisions must be made regarding whether or not to commit a transaction. The issue of trust arises from the uncertainty of results generated by interactions in the electronic trading community, where agents are expected to carry out financially critical tasks by cooperating with other agents. This differs from a conventional environment where trading parties meet face to face, buyers can check the quality of the product before buying it, and sellers can check the buyer's personal credit before delivering the goods. A solution is developed where agents have the ability to evaluate the trustworthiness of other agents and to adaptively find helpful information sources (other agents) to assist in this decision-making process. We refer to these helpful information sources as the agent's “neighborhood.” This neighborhood is expected to provide an effective and efficient evidence retrieval system, simulating a group of friends in a social circle. We develop a fuzzy reasoning scheme that helps agents evaluate the trustworthiness of interacting agents based on the agent's own past experiences and the evidence solicited from the neighborhood. We also introduce a heuristic search strategy designed to allow the agent to adaptively select the best possible neighborhood for the decision-making process, resulting in a self-organizing multi-agent community.