Closing the Affective Loop: Multimodal Speaker-Listener Emotion-Dynamics-Aware Empathetic Social Robots
AffectLoop gives the robot its own modeled affective state, then uses it alongside the user’s changing emotion to shape speech and behavior.
The paper describes a Misty II spoken-dialogue system that tracks verbal and facial affect from the speaker while estimating the robot listener’s verbal and behavioral affect. Those streams condition an LLM-generated short empathetic reply and matching embodied behavior. In a five-person within-subject pilot, it rated higher than an utterance-only baseline, particularly on empathetic response and satisfaction. The authors also report stronger affective alignment and valence-based distress recovery in post-hoc logs. ArXiv · AI/CL/LG's note
The paper describes a Misty II spoken-dialogue system that tracks verbal and facial affect from the speaker while estimating the robot listener’s verbal and behavioral affect. Those streams condition an LLM-generated short empathetic reply and matching embodied behavior. In a five-person within-subject pilot, it rated higher than an utterance-only baseline, particularly on empathetic response and satisfaction. The authors also report stronger affective alignment and valence-based distress recovery in post-hoc logs. ArXiv · AI/CL/LG's note
score 4