Turn management is one of the skills necessary for social interactions. In human-human interactions, the exchange of turns is naturally completed by interruption, a communicative act connoting “cooperation” or “competition”. Interruptions, which at first glance can be considered as discourteous, are inherent in interaction and must therefore be modelled to create new interfaces such as Embodied Conversational Agent. A challenge is then to manage these agents, represented graphically so that they communicate autonomously with humans both verbally and nonverbally. This article presents ongoing work whose goal is to animate an Embodied Conversational Agent capable of managing the interruptions in an interaction. In particular, we are interested in when and how to appropriately interrupt humans without offending them and how to respond appropriately when interrupted. In order to achieve this goal, we start by analyzing human-human interaction data.