Strategies for Software Agent Based Multiple Issue Negotiations

Abstract

This paper summarizes the results of a research project that yielded a prototype for agent-based multiple issue negotiations along with a package of strategies those agents use to accomplish their goals. Whereas many simple negotiation protocols for software agents have already been developed, multiple issue negotiations are seldom addressed. Here, a bilateral negotiation protocol is outlined and a model allowing autonomous software agents to balance out their diverging interests is presented. The model outlines strategies for the agents to compromise on issues based on utility calculations. To calculate its next move, an agent does not only rely on user preferences (importance of issues, attitude towards risk,...) but also considers information from its dynamic surroundings, e.g. the current market situation. The agents also get to know and recognize each other, strategically applying their experiences from negotiations formerly completed or aborted. To ensure applicability, the algorithms designed for modeling the agents’ behavior rest upon certain pragmatic assumptions like bounded rationality, limited information and self-interest.