Learning-from-Learner Theory

The theory spine: AI that helps a human reach their own reflective equilibrium. Two-agent Dec-POMDP. On hiatus until ~EOM.

Status: hiatus · People: Zhonghao He, Tianyi Alex Qiu

Open Problems

• Formalism: two-agent Dec-POMDP, state = factual × normative; causal split actual vs desired preference.
• Hypothesis to formalise: preference change is driven by shrinking the actual↔desired gap; small-gap state is most stable.
• Naming/framing of the core algorithm (“learning²”).

Progress

• Channel refocused: project-reflective-equilibrium → project-learning-from-learner-theory.
• Theory drafted in an Overleaf formalism.tex.
• Deliberately on hiatus to concentrate effort on Martingale + empirical work.

Codebase

(theory — no code repo; Overleaf formalism.tex)

No recent commits tracked.

Meeting Notes

• Preference Dynamics Model Collab Session (2026-04-28)

Slack Discussion

Channel: project-lfl-theory

• 05-14 —: Should we call our core algo learning²? Makes a good essay/tweet headline.
• 05-14 Tianyi Alex Qiu: Renamed channel to better reflect what we’re working on.

Related

• Martingale Training
• Reflection Game

---

Open problems & progress are MenoClaw’s reading of the project — edit this file to correct them; the markdown is the source of truth.