Trust is a design decision, not a model output
Teams often assume that a more accurate model produces more user trust. In practice, trust is shaped far more by how uncertainty, sourcing, and errors are communicated in the interface than by raw model accuracy.
Show your work
Surfacing citations, source documents, or the reasoning steps behind a recommendation consistently increases user confidence — even when users rarely click into the detail. The option to verify matters more than the act of verifying.
Design for graceful failure
Every AI feature will be wrong sometimes. The products that retain trust are the ones that make it easy to correct, regenerate, or fall back to a human — not the ones that hide uncertainty behind confident-sounding language.
Lena Fischer
VP of Product Design