Investigating Non-Verbal Behaviors Conveying Interpersonal Stances

Interpersonal stances are expressed by non-verbal behaviors on a variety of different modalities. The perception of these behaviors is influenced by the context of the interaction, how they are sequenced with other behaviors from the same person and behaviors from other interactants. In this paper, we introduce a framework considering the expressions of stances on different layers during an interaction. This framework enables one to reason on the nonverbal signals that an Embodied Conversational Agent should express to convey different stances. To identify more precisely humans’ non-verbal signals conveying dominance and friendliness attitudes, we propose in this paper a methodology to automatically extract the sequences of non-verbal signals conveying stances. The methodology is illustrated on an annotated corpus of job interviews.