Skip to content
Loom
All threads

Question

Does showing an AI's reasoning actually make people trust it more, according to the research?

Literature synthesis

3 sources

Does showing the reasoning build trust? What the research says

Transparency raises trust, but only up to a point, and only when it is honest about uncertainty. Showing the sources moves trust more reliably than narrating the thinking.

Transparency raises trust, to a point

Across studies, surfacing some reasoning raises trust over a bare answer - people prefer a system that shows part of its work.1

Over-explaining backfires

Past a threshold, more explanation produces over-trust: users accept wrong answers because the narration sounds thorough.2

Trust that does not track accuracy is the failure mode; honest uncertainty calibrates it better than confident prose.2

Provenance beats narration

Naming the sources moves perceived credibility more reliably than narrating a chain of thought, and it gives the user something to check.3

What this implies

Show the structure and the sources, keep the narration short, and be explicit about uncertainty. Let people verify rather than asking them to be impressed.3

woven in 1.73s · 3 sources · 372 tokens