Empirical Evaluation of Strategies for Multiparty Argumentative Debates
Résumé
Debating agents have often different areas of expertise and conflicting opinions on the subjects under discussion. They are faced with the problem of deciding how to contribute to the current state of the debate in order to satisfy their personal goals. We focus on target sets, that specify minimal changes on the current state of the debate allowing agents to satisfy their goals, where changes are the addition and/or deletion of attacks among arguments. In this paper, we experimentally test a number of strategies based on target sets, and we evaluate them with respect to different criteria, as the length of the debate, the happiness of the agents, and the rationality of the result.
Domaines
Intelligence artificielle [cs.AI]
Origine : Fichiers produits par l'(les) auteur(s)
Loading...