Dispute Mediation: A Computer Model

Abstract

This paper integrates Artificial Intelligence techniques with decision theoretic methods to address the problem of finding compromise solutions to multi-agent conflicts through mediation/negotiation. This is a difficult problem since conflict resolution is an ill-defined and complex process, the compromise choices that a problem solver has for continuum-valued issues are infinite, and the agents need to be persuaded to shift their positions during problem solving. Previous approaches have been based on quantitive models that are inflexible and inaccesible to the practitioners. The proposed conflict resolution model integrates reasoning from past cases similar to the current conflict, the use of multi-attribute utility theory, and reasoning to accomodate idiosyncratic behavior of the agents. The model has been implemented in a computer program, the PERSUADER, that functions as mediator in hypothetical labor negotiations. It suggests appropriate settlements to this disputants. If a suggested compromise is rejected, the PERSUADER either improves the compromise or generates persuasive arguments to change the opposing party’s “view” of the settlement.