AI Glossary Term
What does “Trust Calibration” mean?
Aligning user confidence in AI outputs with the system's actual reliability in each task.
How Trust Calibration works
Example
A radiologist uses an AI tool that is strong on some scans and weak on others. Interface cues showing when the tool is less reliable help the doctor lean on it where it is strong and double-check elsewhere.
Why it matters
Too much trust lets errors through, and too little wastes a useful tool. Showing uncertainty and track record helps people judge.
Common questions
Trust calibration vs model calibration?
Model calibration is about the system's confidence scores. Trust calibration is about the human's confidence in the system.
What helps?
Clear indications of uncertainty, explanations, training, and feedback about the system's actual accuracy.
Why do people over-trust?
Fluent, confident output and automation bias, the tendency to defer to a machine's suggestion.