The Role of Explanation Format and Certainty Expression in User Trust Towards AI-Assisted Healthcare Recommendations

As Artificial Intelligence (AI) becomes increasingly integrated into healthcare, how AI-based recommendations should be explained to enhance user trust remains an open question. This study investigated the effects of explanation format and certainty expression on users’ trust in an AI-assisted healthcare recommendation system. Twenty-four participants performed dosage adjustment tasks under four explanation formats and two certainty expression conditions. Subjective trust, task performance, and eye-tracking measures were recorded. The results revealed that explanation format had significant main effects on subjective trust and reliance, compliance rate, task performance, and total fixation duration within areas of interest (AOIs). Counterfactual explanations elicited the highest levels of trust and compliance but required longer decision times due to deeper cognitive processing. Certainty expression showed significant main effects on compliance rate, pupil diameter, blink count, and AOI fixation count. Compliance rate was significantly higher under the certain condition compared to the uncertain condition, as uncertain expressions increased cognitive load and thereby reduced decision efficiency. Overall, both counterfactual explanations and certain expressions were found to enhance user trust and decision quality. Eye-tracking measures further demonstrated their value as objective indicators of cognitive load and human–AI trust. These findings provide guidance for designing more interpretable and user-centered AI explanation mechanisms in healthcare.