Foundation Models brought general capability. Decades of cognitive science describe the structure by which minds, records, teams, and cultures actually reason. The discipline by which they multiply is what this Institute is here to build.
How we work
“Build us a customer‑facing AI.”
Redesign how engagement itself is engineered.
The iteration principle. Not a one‑off product.
“Build us a document AI.”
Reframe what trustworthy reading actually means.
The provenance discipline. Not a black‑box model.
“Build us strategic AI for the executive room.”
Engineer the room where disagreement survives.
The disagreement protocol. Not a forced consensus.
“Build us a culture analytics dashboard.”
Build an instrument for collective reasoning.
The living instrument. Not a frozen dashboard.
What we ship is the strategy. What our clients ship is the system.
Four programmes · Four cognitive contexts
Human intent over algorithm.
AI that reads a mind under pressure — distortion, trait, affect — and answers with the timing and tone a person can actually use.
Enter the programme →Every claim, traceable.
Reading contracts, filings, and financial records at a standard an auditor can defend — provenance built into the system, not bolted on.
Enter the programme →Tension as method.
Strategic decisions held in their tension by a room of agents — never collapsed into a single confident answer the room cannot defend.
Enter the programme →Cognition lives between people.
Multi‑agent simulation as the instrument for studying how teams, institutions, and cultures actually reason together — the discipline distributed cognition has waited for.
Enter the programme →Track record
What the Director has been working on — in operating companies and in print — so the verification a careful reader wants to do is reachable from this page.
Founded
Cirius Technologies
Naked Technology
Nexus FrontierTech
Cinnamon AI
Invested (picked up)
Agna Capital
Momos
OpenCommerce Group
Cotobox
Cynthialy
Nishika
SVSS
Featured publications
CMTD: Cognitive Modeling with Traits and Distortions for Multimodal Emotion Recognition · ACL 2026 (accepted)
From Private Beliefs to Public Silence: A Multi‑Agent LLM Simulation of Psychological Safety · CogSci 2026 (accepted)
Metamo: Empowering Large Language Models with Psychological Distortion Detection for Cognition‑aware Coaching · EMNLP 2025 (Demo Track)
Automatic Prompt Selection for Large Language Models · PAKDD 2025
Advisory engagements
Active engagements include leading Japanese enterprises across telecommunications, financial services, and strategic consulting. Specific partnerships are protected by mutual agreement.
The Institute
Hajime Hotta, Ph.D.
Director · Principal Investigator
A core team of researchers and engineers across Tokyo and Hanoi, a wider collaborator network, and the conditions for work that takes a decade rather than a quarter.
First conversations
A talk you heard. A paper you read. A name someone mentioned. A passing thought about cognition × AI that you would like to explore aloud.
Any of these is reason enough. No formal proposal is required, no research plan, no decision about which working mode it might become. A short note — “here is what I am thinking about, would you talk?” — is the most useful place to start.
The Director responds personally, and treats first conversations as exploratory.
Correspondence
Whether you have read the Letter, or just want to share a passing thought — a note is the best place to start. There is no shape an inquiry has to take. The Director responds personally.