Exploring Trust Calibration in XAI - The Impact of Exposing Model Limitations to Lay Users

In three linesPreregistered study (N=418) on trust calibration in explainable AI: disclosing model limitations slightly improves alignment between user trust and actual performance in skin-lesion classification, but direct experience outweighs onboarding manipulations.

Summary generated by Claude — human-verified