Emotion-Model | Takato Horii | Osaka University

Affective Communication via Active Inference

Mon, 01 Jan 2024 00:00:00 +0000

Based on active inference and the free-energy principle, we computationally model multimodal affective communication between robots and humans. Integrating energy-minimization-driven active perception by the robot with emotion estimation and expression through mental simulation, we build frameworks in which the robot reads and responds to others’ emotions.

Co-Construction of Emotion through Social Interaction

Mon, 01 Jan 2024 00:00:00 +0000

We computationally model the process by which multiple agents co-construct emotion categories through social interaction, using collective predictive coding and Metropolis–Hastings naming games. Built on multimodal deep generative models, we aim to constructively clarify how emotion symbols emerge and are shared among embodied agents.

Emotion Development Models Centered on Tactile Properties

Mon, 01 Jan 2024 00:00:00 +0000

Focusing on tactile dominance in infancy, we computationally model how emotions differentiate and develop through multimodal (visual, auditory, tactile) interactions. Using probabilistic generative models, we reproduce the developmental trajectory of emotion categories that emerge through bodily and social interactions with caregivers, aiming at a constructive understanding of emotion.