An Argumentation Based Approach to Multi-Agent Learning

This paper addresses the issue of learning from communication among agents that work in the same domain, are capable of learning from examples, and communicate using an argumentative framework. We will present (1) an argumentation framework for Case-Based Reasoning agents and (2) an individual policy for agents to generate arguments and counterarguments (including counterexamples). We focus on argumentation between two agents, presenting an interaction protocol (AMAL2) that allows agents to learn from counterexamples and a preference relation to determine the joint outcome when individual predictions are in contradiction. The experimental evaluation shows that argumentation-based joint predictions and learning examples from communication both improve over individual predictions.