This article presents an approach to improve collaborative learning in terms of performance & satisfaction through the generation of non-verbal behavior of users displayed on their avatar in virtual reality. Various works have focused on the behavioral realism of avatars, which can considerably improve interactions. The purpose of this paper is to investigate the impact of displaying the facial expressions of a user in real time on the performance of the task, the satisfaction and the behavioral changes of the users interacting in a virtual environment. To evaluate this approach, we carried out a study where the users collaborated to build a TV stand in dyad including a novice (the participant) and an expert assistant (the experimenter).