Effect of AI system data modality on user trust

The dataset comprises responses from optometrists who participated in a trust questionnaire. The questionnaire was designed to assess their perceptions and levels of trust when collaborating with two different types of AI systems—unimodal and multimodal—in the context of glaucoma diagnosis.

The unimodal AI system relied solely on a single type of data, such as Fundus imaging, to provide diagnostic assistance, while the multimodal AI system utilized a combination of multiple data sources, potentially including Fundus imaging, patient history, and other diagnostic tests, to offer a more comprehensive analysis.

The questionnaire gathered responses on several dimensions related to trust, including the optometrists' confidence in the AI systems’ input data, the accuracy of the AI-generated outputs, the explainability of the system’s recommendations, and the overall system quality. The optometrists were asked to evaluate both the unimodal and multimodal AI systems based on these criteria, providing insights into how different modalities of AI systems influence their trust, reliance, and decision-making during glaucoma diagnosis. The data collected offers valuable information for understanding AI technology in clinical practice.