Alternative socialwelfare definitions for multiparty negotiation protocols

Abstract

Multiagent negotiation protocols, understood as a group decision making process, try to reach an agreement among all the negotiating agents. Traditionally, this agreement is an unanimous agreement. This consensus as unanimity may be quite difficult to achieve in practice or even undesirable in some situations. We propose a framework to incorporate alternate consensus definitions to multiagent negotiations in terms of utility sharing among the agents. The consensus definition is enforced by a mediator, which implements a linguistic-expressed mediation rule based on Ordered Weighted Averaging Operators (OWA). In each step of the mediation process, agents send offers to the mediator. To avoid zones of no agreement, the mediator applies Hierarchical Clustering (HC) to the offers to form group of agents. Then, the mediator computes a social contract, taking into account the desired consensus and the distance from an ideal consensus. The social contract is submitted as a feedback to the agents that explore locally the negotiation space using of a variation of the Generalized Pattern Search (GPS) nonlinear optimization technique to generate new offers that into account the social contract. Finally, We show how these mechanisms are able to reach agreements according to different consensus policies while avoiding zones of no agreement.