# calibrated doubt?

**Does “calibrated doubt keeps a system sound” have a precise form for AI models?**

A probabilistic version of self-trust seems to survive if it allows a margin of error. How far that reaches for AI models is open.

Connects to:
- [what is a belief?](https://mapideas.org/i/h_q1)

Source: https://mapideas.org/i/h_q2 · open on the map: https://mapideas.org/#h_q2