A listening agent exhibiting personality traits Elisabetta Bevacqua, Maurizio Mancini, and Catherine Pelachaud University Paris 8, 140 rue de la Nouvelle France 93100, Montreuil, France INRIA Rocquencourt, Mirages BP 105, 78153 Le Chesnay Cedex, France Abstract. Within the Sensitive Artificial Listening Agent project, we propose a system that computes the behaviour of a listening agent by encompassing the notion of personality. In our system agent’s behaviour tendencies are defined by a set of parameters. The system selects the signals to be displayed by the agent depending on its behaviour tendency and its communicative intentions. In this paper we focus on the listener’s behaviour. 1 Introduction The work presented in this paper is set within the Sensitive Artificial Lis- tening Agent (SAL) project. It is part of the EU STREP SEMAINE project (http://www.semaine-project.eu). This project aims to build an autonomous talking head able to exhibit appropriate behaviour when it plays the role of the listener in a conversation with a user. Four characters, with different person- alities, invite the user to chat trying to induce her/him in a particular mood. Within SAL, we aim to build a real-time Embodied Conversational Agent (ECA) endowed with recognizable personality traits while interacting with users in the role of the listener. We propose a model that provides a static definition of such an ECA on the base of: (i) the preference the agent has in using each available communicative modality (e.g., head orientation, eyebrows movements, voice and so on) and (ii) a set of parameters that affect the qualities of the agent’s be- haviour (e.g. wide vs. narrow gestures). We call such a definition of the agent the baseline and we fix it depending on the agent’s personality traits. The pro- posed work incorporates a pre-existing system for the generation of distinctive behaviour in ECAs [MP07,MP08]. The result is a system capable of determining the agent’s communicative intention in real-time, according to the speaker’s be- haviour, and computing the verbal and non-verbal behaviours that the agent has to perform on the basis of both its baseline and its communicative intentions. For example, an extrovert agent can show its agreement through wide and fast head nods and an open smile on its face. 2 Background 2.1 Personality A big challenge that must be faced in the design of virtual agent is the issue of credibility, not only in the agent’s aspect but also in its behaviour. Users