Abstract:
This paper studies multimodal communication between a teacher and a learner. Our long term goal is to specify the behavior of an animated pedagogical agent in a tutoring system. Our methodological approach is based on video corpora annotation. Dyadic interactions between teachers and learners were recorded. We worked with the Pepite tutoring system that assesses the level in algebra of students aged 15-16. We defined an annotation schema applied to a part of the videos. Statistical analysis of the annotations allows for capturing different strategies in the way the teacher uses gaze depending on students’ level and on pedagogical acts.