Trust Isn’t a Number: Designing AI Confidence People Can Actually Verify

Nabanita Banerjee on YouTubeWatch on YouTube

SaveMark complete

Sign in to save videos and track what you’ve finished.

IntermediateUX7 min

About this video

A product designer builds a fictional fraud-review tool to compare an AI confidence score shown alone with one that shows the supporting evidence and its limits. It is aimed at designers working on AI products who want to make model confidence verifiable.

What you’ll learn

  • Compare a bare confidence score with an interface that shows its supporting evidence
  • Show when a model lacks enough evidence to be sure
  • Tie AI confidence displays to facts users can check themselves