The system lets its confidence show — in wording, ranges, or marks on specific spans — so the actor knows which parts of the output stand on firm ground and which are guesses.
Consequences
- confidence is part of the output's surface — hedged phrasing, ranges, marks on the doubtful spans — so the actor's scrutiny can land where it is actually needed
- a number implies calibration that may not exist: "80% confident" is itself a generated claim, and precision in the mark overstates precision in the ground
- hedging everything is as useless as asserting everything; the value is in the difference between the marked and the unmarked
Resources & references
- aiuxdesign.guide — Confidence Visualization
- AI UX Playground — Confidence Score, Confidence Indicators
- Shape of AI — Caveat
Appears in
Calibrating reliance
- Making the output checkable, step 2 — Let the confidence show
Related patterns
Precedes
- Cognitive forcing functions — confidence is part of the output's surface — hedged phrasing, ranges, marks on the doubtful spans — so the actor's scrutiny can land where it is actually needed
Serves
- Assistance — expressed uncertainty covers assistance's signal-raising step — qualifying output so attention goes where confidence is thin
Enacts
- Agency — the actor keeps the judgement that uniform fluency would otherwise take
Complements
- Suggestion — a proposal's weight depends on how firm the ground under it is
- Citation — how sure it is versus where it came from
Related to
- Explanation — the factors behind a prediction are one way of saying how much to trust it