Confidence Calibration
Updated 2026-08-11
INTRODUCTION
English translation pending.
CORE DEFINITION
Confidence calibration is the degree to which a person's subjective confidence matches objective accuracy: a well-calibrated judge who says eighty per cent is right about eighty per cent of the time. Philip Tetlock's long-run study of expert political forecasting, published as Expert Political Judgment, found that the most famous and most confident experts were among the least accurate. Weather forecasters are the standard counter-example, because they publish explicit probabilities and are scored against outcomes immediately.
SCAFFOLDING EFFECT
Reduce cognitive load
- Numeric commitment: replace words like likely with an explicit probability before deciding - Personal ledger: record the forecast and the outcome so frequencies can be computed - Bucket review: group past forecasts by probability band and check the realised hit rate against your stated confidence
Anchor fast decisions
Confidence is an internal feeling, while accuracy can only be measured against external outcomes, and the two have no necessary connection, so confidence generally runs above the actual hit rate. Calibration improves judgement through a feedback loop: when judgements are written in checkable form and results are recorded, the subjective feeling is gradually corrected by the observed frequency. In domains without prompt feedback, such as long-horizon political prediction, that corrective pressure never arrives and overconfidence can persist indefinitely.
MINIMUM ACTION
In progress 0/1Practice this model in one real situation:
account_treeGenealogyexpand_more
menu_bookReferencesexpand_more
Source support: Explicit
- mungermodels.comhttps://mungermodels.com/models/confidence-calibrationverified
PRIVATE NOTES · Only visible to you
SAVED Q&A
ENTRY Q&A · Private saving available
Ask with a clear boundary
thinkingmodels answers from published entry context only.
Your question is sent to thinkingmodels. The answer uses public entry context only.
RELATED MODELS