Arguments and Couterexamples in Case-based Joint Deliberation

Abstract. Multiagent learning can be seen as applying ML techniques to the core issues of multiagent systems, like communication, coordination, and competition. In this paper, we address the issue of learning from communication among agents circumscribed to a scenario with two agents that (1) work in the same domain using a shared ontology, (2) are capable of learning from examples, and (3) communicate using an argumentative framework. We will present a two fold approach consisting of (1) an argumentation framework for learning agents, and (2) an individual policy for agents to generate arguments and counterarguments (including counterexamples). We focus on argumentation between two agents, presenting (1) an interaction protocol (AMAL2) that allows agents to learn from counterexamples and (2) a preference relation to determine the joint outcome when individual predictions are in contradiction. We present several experiment to asses how joint predictions based on argumentation improve over individual agent’s prediction.