aceModerate evidence
Published recordScienceGood Space
More
When artificial intelligence apologizes: Dynamic trust updating and the role of personality.
- Truvace record
- TRV-2026-2518raw data (JSON)
- Version
- v1Published
- First published
- Evidence type
- Randomized controlled trial
- Sources
- 1 sourcePeer-reviewed
- sha256
- dde862c099b7…check it
Artificial intelligence (AI) systems are susceptible to error, and many are designed to apologize.Yet, it remains unclear which apologies restore user trust, a question requiring repeated interactions with an AI agent and models capturing how trust evolves across encounters.Drawing on models of apology from human-human interaction research, we examined whether apology components that repair interpersonal trust function similarly in AI.Participants (N = 245) solved five business problems with a chatbot preprogram…
The gain
In a randomized experiment with 245 participants, a chatbot expressing regret after an error was the only one of six apology types that softened the immediate drop in trust compared with no apology.
What this doesn’t fix
None of the apology conditions improved trust recovery once the chatbot resumed correct performance, and the effects were selective and context-dependent, depending on participants' evaluation profiles and personality.
Sources
1 on file- Technology, Mind, and Behaviordoi:10.1037/tmb0000230
Cite this record
v1Truvace Impact Record TRV-2026-2518, v1: “When artificial intelligence apologizes: Dynamic trust updating and the role of personality..” Truvace, 2026-11-12. /record/TRV-2026-2518 (accessed on the day you copy it). sha256 dde862c099b7d63b…
Change history
1 versionEvery change made to this record since it was first published, and why. Nothing has changed since it went up.
Certified into the record
Check this record hasn’t been altered
How to check it yourself, without trusting this page
Download any version from /api/record/TRV-2026-2518 and hash its canonical field yourself, for example with shasum -a 256. You should get its content_hash. Each version’s text also ends with prev:and the hash of the version before it (version 1 points to 64 zeros). If anyone had changed a single character of any version after it was saved, those hashes wouldn’t line up.