Modeling Non-Cooperative Dialogue: Theoretical and Empirical Insights

Abstract Investigating cooperativity of interlocutors is central in studying pragmatics dialogue. Models conversation that only assume cooperative agents fail to explain the dynamics strategic conversations. Thus, we investigate ability identify non-cooperative while completing a concurrent visual-dialogue task. Within this novel setting, study optimality communication strategies for achieving multi-task objective. We use tools learning theory develop theoretical model identifying and apply analyze different strategies. also introduce corpus conversations about images GuessWhat?! dataset proposed by De Vries et al. (2017). reinforcement implement multiple context find empirical results validate our theory.