I stole the idea of confidence tags from Gwern. A confidence tag on an essay gives my subjective credence in the central thesis of that essay, using this spectrum of estimative terms:
- “almost certain”
- “highly likely”
- “likely”
- “somewhat likely”
- “possible”
- “somewhat unlikely”
- “unlikely”
- “highly unlikely”
These do not correspond to exact probabilities, but you could loosely interpret them like this:
- “almost certain”: >90%
- “highly likely”: 80–90%
- “likely”: 65–80%
- “somewhat likely”: 55–65%
- “possible”: 45–55%
And similarly for the terms below “possible”.