Document Type

Article

Abstract

Multi-issue negotiation may produce mutual beneficial results to both negotiators while single-issue negotiation can not. However, there are difficulties in automating a multi-issue negotiation, since the search space grows dramatically as the number of issues increases. Although many concession strategy learning mechanisms have been proposed to deal with the problem, recent research uncovered that the fixed strategy of concession and the fixed-pie bias are the two major interferences in the automation of multi-issue negotiation. It is suggested that the lack of communication between agents may have impeded information sharing and joint-problem solving possibilities.

In this paper, we show that the fixed-pie bias can interfere with the negotiation outcome if there are non-conflicting issues. We propose a new negotiation model and an innovative algorithm that not only allows information to be shared in a controlled way, but also allows the information shared to be effectively used for conducting a systematic search over the negotiation problem space. The combined mechanism is capable of using strategies learned from counter-offers and is immune to the fixed-strategy limitation and the fixed-pie bias. It contributes to the automation of multi-issue negotiation in the context of open and dynamic environments.

Share

COinS