I can do better than your AI: expertise and explanations
summary
An experiment at UCSB conducts a quantitative study N=529 using an abstract trust game (the diner's dilemma) with a “dining guru” assistant that varies explanation presence, automation level, and injected errors. The assistant’s predictions outperform typical human moves, reported familiarity negatively predicts optimal moves, higher self-reported familiarity yields more stated trust but less adherence, explanations only help lowest quartile, and automation reduces awareness creating automation bias when explanations are shown to confident users.
AI-written summary · about
chapters (20)
- 0:00Introduction
- 0:46Outline
- 1:33Background
- 3:35Previous work
- 4:39The Kruger effect
- 6:06The Diners Dilemma
- 8:26Methodology
- 9:09Process flow
- 11:17Awareness questions
- 11:42Validation
- 12:11Research questions
- 12:40Results
- 13:26Selfreported familiarity
- 14:19Adherence
- 15:14Automation bias
- 16:03Conclusions
- 16:54Research Question
- 18:12Conclusion
- 19:01Question
- 20:54Evaluation