Bilateral Negotiation with Incomplete and Uncertain Information: A Decision-Theoretic Approach Using a Model of the Opponent

The application of software agents to e-commerce has made a radical change in the way businesses and consumer to consumer transactions take place. Agent negotiation is an important aspect of e-commerce to bring satisfactory agreement in business transactions. We approach e-commerce and negotiation in the context of a distributed multiagent peer help system, I-Help, supporting students in a university course. Personal agents keep models of student preferences and negotiate on their behalf to acquire resources (help) from other agents. We model negotiation among personal agents by means of influence diagram, a decision theoretic tool. To cope with the uncertainty inherent in a dynamic market with self-interested participants, the agents create models of their opponents during negotiation, which help them predict better their opponents' actions. We carried out experiments comparing the proposed negotiation mechanism with influence diagram, one using in addition a model of the opponent and one using a simple heuristic approach (as a base for comparison). The results show some of the advantages and disadvantages of the proposed negotiation mechanisms.