Current Position & Biography
Kenji Doya is Professor of the Neural Computation Unit at the Okinawa Institute of Science and Technology (OIST), working across artificial intelligence, robotics and computational neuroscience. His research asks how people and animals learn skills and make decisions, and how those principles can inform learning machines.
Doya earned his bachelor’s, master’s and doctoral degrees at the University of Tokyo. His research career included work at the University of California, San Diego, the Salk Institute and Japan’s Advanced Telecommunications Research Institute (ATR).
His OIST profile also records a long-running role in the Okinawa Computational Neuroscience Course and his former co-editorship of Neural Networks. These educational and editorial activities sit alongside his research rather than replacing it.
Notable Contributions
His 1995 conference paper, “Temporal Difference Learning in Continuous Time and Space,” derived a continuous-time, continuous-state learning algorithm for control and neurobiological modeling. It tested algorithms on a pendulum swing-up task with limited torque. His later 2000 paper addressed reinforcement learning in continuous time and space. Together, these publications connect reward-driven learning with continuously changing physical systems.
His 2002 paper, “Metalearning and Neuromodulation,” proposed computational roles for brain systems that regulate learning. It approached neuromodulation as a way of coordinating distributed learning processes. This is a theoretical account, not a claim that a single model completely explains the brain or establishes a clinical treatment.
OIST describes his unit’s research as combining learning algorithms with investigations of their biological mechanisms. In March 2026, it announced his selection for the ninth Tateishi Special Prize, recognizing work linking reinforcement learning, neuroscience and robotics, as well as research education. The announcement also described ongoing work toward a unified account of perception and action. These are active research aims, not completed demonstrations of general intelligence.
Why Katharos Is Following Them
Doya offers a perspective on AI grounded in adaptation, embodiment and feedback. His career connects the mathematical design of learning systems with the experimental study of biological learning. That combination is useful when assessing claims about autonomous robots, brain-inspired AI and systems expected to operate beyond carefully controlled benchmarks.
Areas to Follow
Reinforcement learning for physical systems, probabilistic inference, computational theories of neuromodulation, learning and decision-making experiments, and interdisciplinary research training.
Reporting Scope
Coverage distinguishes computational hypotheses from established biological mechanisms, experimental findings from clinical applications, and research goals from dependable deployed systems. Individual contributions are placed alongside the work of laboratory members and collaborators.
