Dynamical Systems to Account for Turn-Taking in Spoken Interactions

Mathieu Jégou (1) , Pierre Chevaillier (2, 1, 3) , Pierre de Loor (1, 2, 3)

1 Lab-STICC - Laboratoire des sciences et techniques de l'information, de la communication et de la connaissance
2 ENIB - École Nationale d'Ingénieurs de Brest
3 Lab-STICC_ENIB_CID_IHSEV

Turn management is considered as essential for an Embodied Conversational Agent (ECA) to increase user’s engagement. This article presents a dynamical model for turn management in dyadic interactions. The model is a system of differential equations that mixes two models from the cognitive sciences, the Drift Diffusion Model, and the Behavioral Dynamics. Decision-making and the control of actions are two coupled processes that modulate continuously the behavior of the interacting agent. This conceptual model accounts for the emer- gence of smooth transitions without using neither prediction nor planning of the agent’s behavior. The objective was not to obtain a fully realistic behavior, but to show how the model could account for the main qualitative properties of turn management, such as interrupting the current speaker, signaling its willingness to go on speaking, or yielding the turn to the next speaker.

Mots clés: Human-Computer Interaction; Spoken interaction; Turn-taking management

Domaines: Informatique [cs]; Interface homme-machine [cs.HC]

Mathieu Jégou, Pierre Chevaillier, Pierre de Loor. Dynamical Systems to Account for Turn-Taking in Spoken Interactions. Springer International Publishing. Fourteenth International Conference on Intelligent Virtual Agents (IVA 2014), Aug 2014, Boston, United States. , Intelligent Virtual Agents, 8637, pp.218-221, 2014, Lecture Notes in Computer Science. ⟨10.1007/978-3-319-09767-1_27⟩. ⟨hal-01165933⟩