Value Model Agent: A Novel Preference Profiler for Negotiation with Agents

Abstract

As multi-agent systems become more complicated, situations arise were agents have different goals. Accomplishing their goals may require trading resources or compromising on the state of the system. In these situations cooperation can improve the utilities of all parties, but some agreements are preferred to others. Humans have reached these agreements for thousands of years through an art known as negotiation. In this paper we describe an agent that was developed for the ANAC2011 bilateral negotiation competition. Our main contribution is a novel approach to modeling the preference profile of the other agent. This preference profile is then used to improve exploration of the bid space and approximate the opponent’s concessions.