Yunfeng Zhang, Q. Vera Liao, & Rachel K. E. Bellamy (2020)

Proceedings of the Conference on Fairness, Accountability, and Transparency (FAT* '20), 295-305.

DOI: https://doi.org/10.1145/3351095.3372852

Abstract. Empirical study finding that displaying AI confidence scores helps users calibrate trust — but only when the confidence is itself well-calibrated. When confidence is mis-calibrated, showing it actively harms human-AI team performance.

Tags: ai-usability trust calibration

Textbook of Usability

Effect of confidence and explanation on accuracy and trust calibration in AI-assisted decision making