calibration
making a model's stated confidence match how often it is right
A model often gives a probability with each answer. It is well calibrated when answers given 80% confidence are right about 80% of the time. Good calibration lets a business set safe thresholds. For example, low-confidence answers can go to a person.
Think of it as
A weather forecaster whose "70% chance of rain" days really do rain seven times in ten.
Ideas that use this term