A new category of business negotiation primitives for bilateral negotiation agents and associated algorithm to find Pareto optimal solutions

Abstract

How to conduct automated business negotiation over the Internet is an important issue for agent research. In most bilateral negotiation models, two negotiation agents negotiate by sending proposals and counterproposals back and forth. Proposals and counterproposals usually use a set of negotiation primitives to define the contents. These primitives deal with complete proposals or the whole negotiation process. With these primitives, a negotiation agent can issue a call-for-proposal (CFP), reject or accept the received proposal, propose a new proposal, or terminate the whole negotiation process. However, there is no primitive about individual attributes within proposals. In this paper, we propose a new category of negotiation primitives that have finer granularities than existing primitives do. They are in the attribute level instead of the proposal level, as other primitives are. These messages can help negotiation agents to find Pareto optimal solutions if both sides agree to use them. In order to eliminate non-Pareto optimal solutions, these primitives should be used before the actual negotiation (bargaining) begins. Pareto optimal solutions help both sides to reduce the negotiation effort and time, and these solutions are theoretically preferable to non-Pareto optimal solutions.