Adaptive Negotiation Based on Rewards and Regret in a Multi-Agent Environment

The paper presents a model of heuristic negotiation between self-interested agents which allows negotiation over multiple issues of the negotiation object, comprises different types of negotiation primitives, including argument based ones, and a set of rules to conduct negotiation. In order to negotiate strategically and to adapt negotiation to different partners, the agents use rewards associated to negotiation objects and the notion of regret to compare the achieved outcomes with the best possible results that could have been obtained both in a particular negotiation and in selecting the partner agent.