Meno

Meno is a research network, not an organization or company. Our members hold their primary appointments at universities and labs, and collaborate here on shared questions about AI and human reasoning. The network is a fiscally sponsored project of Anti Entropy, a 501(c)(3) nonprofit.

Facilitating the development of human ideas, helping people reason carefully, explore alternatives, and reach beliefs they would genuinely endorse on reflection. Preventing lock-in, so that AI feedback loops do not freeze today's errors in place, making them self-reinforcing and hard to escape.

AI systems now play a role in how people learn, conduct research, and form beliefs. In principle, they could serve as co-learners, co-scientists, and co-thinkers whose optimization objectives are grounded in the informed wills of individuals. In practice, the same systems are already associated with disinformation, deskilling, cognitive distortions, and AI-induced psychosis, and raise longer-term concerns about polarization and the entrenchment of today's values.

We develop training interventions, formal statistical tools, and empirical methods that do not require controlled laboratory conditions. Our work includes Martingale training for the removal of belief entrenchment and coherence optimization as a computable formalism for reflective equilibrium. It also includes, to our knowledge, the first empirical characterization of the human-AI feedback loop at real-world scale and over long time horizons.