Q-learn argumentation schemes for car sales dialogues

Abstract

Agents need to argue with other agents many times, developing persuasion strategies that are effective over repeated situations. Applying reinforcement learning (RL) to the design of argumentation policies is appealing to dialogues where the counterpart can be modelled as a probability distribution. The idea of this research is to apply RL to speech acts in order to learn which discourse pattern is best to be conveyed during an argumentation game. Empowered by this learning mechanism, the persuasive agents gradually become more skillful through repeated argumentation.