Just published! Madeleine's work on inferring internal states has been published in Frontiers in Robotics and AI (open access).
The paper investigates how partially hidden 'internal states' (like emotions, cooperativeness, etc) can be decoded from simple visible cues, like skeletons. Also demonstrates that social situations can be described along 3 simple dimensions: the imbalance of the interaction, its general valence, and the degree of engagement of the interactors.