The benefits of opponent models in negotiation

Information about the opponent is essential to improve automated negotiation strategies for bilateral multi-issue negotiation. In this paper we propose a negotiation strategy that exploits a technique to learn a model of opponent preferences in a single negotiation session. An opponent model may be used to achieve at least two important goals in negotiation. First, it can be used to recognize, avoid and respond appropriately to exploitation, which differentiates the strategy proposed from commonly used concession-based strategies. Second, it can be used to increase the efficiency of a negotiated agreement by searching for Pareto-optimal bids. A negotiation strategy should be efficient, transparent, maximize the chance of an agreement and should avoid exploitation. We argue that the proposed strategy satisfies these criteria and analyze its performance experimentally.