Research Projects | Takato Horii | Osaka University

Human–AI Symbiosis and Co-Creation: Self Mirroring Twins for Active Social Co-Creation

Wed, 01 Oct 2025 00:00:00 +0000

This project builds an AI system foundation that enables active social co-creation by having humans live in symbiosis with “Self Mirroring Twins” — agents that learn and imitate the user’s behavior, preferences, and values — thereby deepening self-understanding and prompting behavior change.

Drawing on cognitive developmental robotics and active inference, we model long-term interactions between Self Mirroring Twins and humans, the dynamics by which both parties update their beliefs and predictions, and the computational understanding of how the boundary between the self and the Twin is formed. We aim to demonstrate, on real-world robots and dialogue systems, the effects on behavior change brought about through interaction with the Twin.

AI Cobots for Next-Generation Shipyards

Tue, 01 Apr 2025 00:00:00 +0000

Shipyards demand collaborative environments in which humans and robots can work in close proximity without safety barriers. This project develops AI-driven motion acquisition and task execution algorithms for collaborative robots (cobots), contributing to next-generation shipyards that flexibly adapt to complex and diverse shipbuilding processes.

We combine imitation learning, reinforcement learning, and foundation models such as Large Language Models and Vision-Language Models to study frameworks by which robots acquire skills while observing and understanding human work. The goal is to implement general-purpose cobots that can safely cooperate with on-site workers while handling tasks in the long tail of the distribution.

Comforting Social Touch — Individual Differences and Cognitive-Scientific Understanding

Tue, 01 Apr 2025 00:00:00 +0000

Being touched by another — stroking, gentle contact, or a handshake — gives us a strong sense of comfort and security. Yet the neural and cognitive mechanisms that underlie this pleasantness, and the origins of individual differences in how it is felt, remain largely unclear.

This project combines psychophysics, neuroscience measurement, and computational modeling to systematically clarify the affective value generated by social touch and its individual differences. Drawing on affective human-robot interaction via touch, we build computational frameworks based on active inference and generative models, and aim at a foundation that allows robots to adaptively respond to individual tactile preferences.

Elucidating Infant Learning Principles Based on Predictive Coding

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

Human infants acquire the structure of the world and the intentions of others remarkably quickly from limited experience. This project aims to formulate the principles of infant learning as computational models grounded in predictive coding and active inference.

We model the integration of multimodal sensory information and the generative processes of expectations and prediction errors in social interactions with caregivers. From the perspective of cognitive developmental robotics, we bridge real-robot experiments and developmental psychology experiments. By describing infant learning principles as a reproducible computational theory, we aim to apply these insights to the understanding of developmental disorders and to early-learning algorithms for robots.

Fundamental R&D on Adaptive CA Control Technology (Moonshot)

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

Under Moonshot Goal 1 “ ,” we conduct research and development on the foundational technologies that enable Cybernetic Avatars (CAs) to interact freely with environments from remote locations. We design control architectures that seamlessly integrate AI-driven autonomy and human teleoperation, and validate them toward societal deployment.

We took over this project as the new PI in January 2024, and continue the work as a follow-on project (through November 2030) starting in December 2025. For details, see the .

Qualia Structurology — Bridging Subjective Conscious Experience and Objectivity

Sat, 01 Apr 2023 00:00:00 +0000

This project aims to bring subjective conscious experiences (qualia) — such as the redness of red or the painfulness of pain — onto the scientific stage as structures. The research area builds methodologies that measure, describe, and compare the “structure” of qualia at the intersection of psychophysics, neuroscience, information theory, and machine learning.

We participate in the planned research project “Correspondence between Qualia Structures and Information Structures” (PI: Masafumi Oizumi), and use deep generative models and representation learning to model the correspondence between the structures of human subjective experience and the information structures inherent in neural activity and behavioral data. In particular, we quantitatively verify the consistency between latent representations obtained via self-supervised learning and similarity structures derived from psychophysics, taking on the challenge of a constructive understanding of “the structure of inner experience” in robots and AI.

Embodied Knowledge of Music — Enhancing Musical Emotion through Biofeedback

Sat, 01 Oct 2022 00:00:00 +0000

The “moving” experience of music — including chills — is closely tied to the listener’s physiological and bodily responses. This international joint project builds a new framework for experience design that measures physiological signals such as heart rate, electrodermal activity, and respiration in real time, and feeds the resulting changes back into musical performance and composition.

From the perspectives of active inference and predictive coding, we model the cycle of expectation, prediction error, and bodily response that underlies the experience of being “moved,” and contribute to designing bidirectional interactions between robot/AI performers and listeners through their bodily responses.

Plantar Science Opened up by Insole Sensors and Plantar Big Data

Fri, 01 Apr 2022 00:00:00 +0000

Standing, walking, and running all connect us to the ground through the soles of our feet. This project develops highly wearable insole sensors and accumulates long-term individual plantar information at scale to establish a new scientific foundation for movement, health, and skill.

We contribute by designing machine-learning and deep-learning models for the collected plantar time-series data, and by developing algorithms that automatically estimate motor skills and physical states. The outcomes are extended to motor learning support, computational understanding of human motion, and motion control models in robotics.

Research Projects | Takato Horii | Osaka University

Human–AI Symbiosis and Co-Creation: Self Mirroring Twins for Active Social Co-Creation

関連する研究費

AI Cobots for Next-Generation Shipyards

関連する研究費

Comforting Social Touch — Individual Differences and Cognitive-Scientific Understanding

関連する研究費

Elucidating Infant Learning Principles Based on Predictive Coding

関連する研究費

Fundamental R&D on Adaptive CA Control Technology (Moonshot)

関連する研究費

Qualia Structurology — Bridging Subjective Conscious Experience and Objectivity

関連する研究費

Embodied Knowledge of Music — Enhancing Musical Emotion through Biofeedback

関連する研究費

Plantar Science Opened up by Insole Sensors and Plantar Big Data

関連する研究費