Yusuke Hayashi
Independent researcher — computational cognitive science & collective intelligence
I study how symbols, meaning, and agreement emerge in populations of learning agents — and how the same mathematics can help humans and machines deliberate together.
Research
Collective Predictive Coding (CPC)
Free-energy accounts of symbol emergence in multi-agent systems; phase transitions in shared representations.
Symbol emergence & emergent communication
Lewis games, VQ world models, and the information-theoretic limits of shared codes.
AI alignment
Alignment as collective inference: what agreement between humans and models must preserve, and what it may compress away.
Digital democracy
Deliberation platforms (Polis-style) as collective world models; mapping high-dimensional opinion spaces.
Full publication list
Recent Activity
Jun 2026 · Conference
Organizer of OS-49 “AI Alignment” at JSAI 2026, the 40th annual conference of the Japanese Society for Artificial Intelligence (Takasaki, Jun 8–12). Session list
May 2026 · Preprint
“Lost and Found in Translation: Variational Diagnostics for Neural Codebook Channels.” arXiv:2605.18846
Feb 2026 · Preprint
“Thermodynamic Limits of Physical Intelligence.” arXiv:2602.05463
2026 · Preprint
“Symbiotic Alignment via Collective Predictive Coding: A Theoretical Framework for Co-Creative Human–AI Ecosystems.” Zenodo
2026 · Invited article
“CPC の数理的理解——Shannon 理論から集合的予測符号化へ,” Journal of the Robotics Society of Japan, 44(5).