Probabilistic UI
Probabilistic UI is interface design for software that can produce different outputs for the same input—LLMs, recommenders, generative media—not deterministic CRUD.
It requires states for uncertainty, regeneration, verification, and graceful failure.
What it means
UX that assumes variation and error: streaming, confidence cues, edit/regenerate, citations, and rollback—not a single fixed happy path.
Why designers should care
Applying deterministic-app patterns to AI creates false certainty. Probabilistic UI makes variation and risk visible so users can steer.
Example
Same prompt returns three draft variants in a carousel; user picks one, edits, and rates helpfulness to improve the next session.
Common mistakes
- • Single submit → spinner → immutable result with no retry.
- • No empty or “I don’t know” states.
- • Punishing users for regenerating (“Are you sure?” every time).