Auteurs : Bassiliades N., Spanoudakis N.I., Kakas A.C.
Publication : In Proceedings of 10th Hellenic Conference on Artificial Intelligence (SETN18), Rio Patras, Greece, July 2018

In this paper, we develop a novel computational argumentation framework for resolving conflicts that arise in a community of multiple stakeholders where each one of them bears a private policy/strategy for shared and inter-related decisions. Decisions taken individually by stakeholders can be contradicting, so there is a need for an arbitration service that will resolve the conflict and conclude on a single decision. Centralized mediation approaches gather all relevant context information and decide on the prevailing decision option as suggested individually by multiple stakeholders. There is high complexity on resolving all possible competing option conflicts among all competing stakeholders, thus, usually centralized solutions do not scale. Our approach avoids this complexity because it is based on defining an arbitration meta-policy for deciding on the priorities among stakeholders, which are few, and not among competing decisions of stakeholders. Then, this meta-policy is automatically rewritten into a full meta-policy about conflicting options, but without user intervention. Thus, human arbitrators can seamlessly define their arbitration meta-policies without a heavy cognitive load.