Designing Trust
How should AI communicate confidence?
Trust is not created by making AI sound certain. It is created by helping people understand how much confidence they should place in an output.
Trust is a communication problem
AI products often present answers with too much confidence. The interface can make something probabilistic feel finished, which quietly shifts responsibility onto the user.
Good trust design helps people understand what the system knows, what it is unsure about, and what should be checked before acting.
Confidence is not certainty
A confidence score is only useful if people understand what it means. Otherwise it becomes decoration, or worse, false reassurance.
I prefer designs that communicate confidence through evidence, source visibility, missing information and clear next steps.
- What evidence is the AI using?
- What is missing from the answer?
- How easy is it to check or challenge?
- What happens if the output is wrong?
What good looks like
A trustworthy AI experience should make uncertainty legible without making the product feel unusable.
The goal is not to make people trust the system blindly. It is to help them decide when to trust, when to check and when to ignore.
The best AI interfaces do not just answer. They help people understand how much weight that answer should carry.