Key points are not available for this paper at this time.
Explainable AI (XAI) has been designed to render AI comprehensible to humans by providing explanations of its processes. However, an excessive amount of explanation can lead to cognitive overload in users and the development of inappropriate trust in AI. To explore the appropriate amount of explanation, this study investigated the effects of explanation type (AI attention heatmap, AI goal, and AI reliability) on trust in XAI for a traditional visual search task in Experiment 1, and the effects of presenting adjusted explanations on task performance for an applied task in Experiment 2. As a result, displaying AI results alone increased trust and task performance for a low-complexity task, and displaying AI results with AI attention heatmaps that have high interpretability increased trust and task performance for a high-complexity task. This study demonstrated the importance of adjusting the amount of AI explanation to develop trust appropriately and improve task performance.
Maehigashi et al. (Sat,) studied this question.