Modeling Multiagent Deliberation from an Abstract Standpoint

Simply put, a multiagent system can be understood as a collection of autonomous agents able to accomplish as a whole goals beyond the capabilities of any of its members. The traditional example depicts a heavy armchair that can be easily lifted by coordinating the effort of a group of persons despite that none of them would have been able to pick it up alone. Thus, one might argue that precisely the agent interaction is boosting the system performance. Since this interaction comes in several flavors, the literature has already explored notions such as agent coordination, cooperation, and collaboration in the context of multiagent systems. This extended abstract outlines our own understanding on this matter, summarizing the evolution of an abstract model for the particular kind of agent interaction known as deliberation. A group of agents deliberate whenever they need to come to a mutually accepted position about some issue. This interaction among agents has drawn our attention given its ubiquity: we believe that complex interactions such as coordination or cooperation might be attained as a result of accruing one or more deliberations. Our proposal is inspired after the novel trend of reinterpreting agent interaction as if it were the result of an argumentation process. For instance, several authors [2, 3, 5, 13, 14] have recently considered reformulating the main aspects of multiagent negotiation in terms of defeasible argumentation. We follow a like approach in developing our model after a set of dialectical concepts borrowed from that same area. Our approach also strives for generality, mainly after Dung’s ample success with his notion of argumentative framework due to its abstract nature. In consequence, we too have decided to pursue an abstract model.