Modeling The Impact Of Action Tendency On An Agent Interrupting Behavior.

Expression of emotions and interruption management are two major capa-bilities that should be displayed by Embodied Conversational Agents. Yet, no work have linked interruption management and emotions. In this paper, we introduce a new model for interruption management that considers one component of emotions: the action tendency. We explain how the action tendency influences the motivation that the agent has to change role (e.g. the motivation to speak if it is currently listening to the user), which impacts its behaviors related to interruptions. We then present an implementation of our model in an existing agent architecture, AgentSlang, and illus-trate the behavior of our model with examples of interactions between the agent and one user. We explain for each case how motivation to change role and turn-taking behaviors are computed for each case.