In computational neuroscience, the free-energy principle holds that organisms minimize free energy via perception and learning. It includes predictive coding, where the brain updates an internal model by reducing prediction errors, and active inference, where organisms select actions that lower expected free energy and unify perception and planning. Despite support, human research often separates sensory processing and motor control. We propose a unified sensorimotor model based on the free-energy principle. We will implement the model, run simulations, and compare outputs with empirical data to evaluate it. It aims to advance HCI, psychology, and applications in developmental disorders and robotics.
GRANTS
-
BOOST (young researchers), Japan Science and Technology Agency, 2025-2030.
PUBLICATIONS
-
Keigo Matsumoto. The Interface as a Markov Blanket: Reframing VR Interaction via the Free Energy Principle. 2026 IEEE Conference on Virtual Reality and 3D User Interfaces Abstracts and Workshops (VRW), pp. 634-638, 2026.
-
Hideaki Yoshii, Takuji Narumi, and Keigo Matsumoto. Modeling Hand Movements Under Pseudo-Haptic Feedback: A Comparative Analysis of Active Inference and Stochastic Optimal Feedback Control Models. Augmented Humans International Conference 2026, pp. 432-441, 2026.
-
Kohei Murayama, Shinnosuke Noguchi, Sotaro Yokoi, Takuji Narumi, Hideaki Kuzuoka, and Keigo Matsumoto. Modeling Multisensory Integration in Hand Redirection: A Bayesian Causal Inference Framework for Understanding Individual Variability. ACM Symposium on Applied Perception 2025, Article 13, 2025. [doi]