Application of Artificial Intelligence Information Prompts in Endoscopic Navigation Tasks: The Effect of Prompt Type and Visual Occlusion

作者：Mu Tong, Long Cui, Sumin Zhu, Xuanfeng Zhang, Haobo Liu · 发表于：International Journal of Human-Computer Interaction · 年份：2026 · DOI：10.1080/10447318.2026.2703334 · 研究领域：Surgical Simulation and Training、Artificial Intelligence in Healthcare and Education、Augmented Reality Applications

Disorientation during endoscopy is a major challenge for novice operators. This study examined how no prompt, visualization prompts, and AI-assisted prompts affect simulated endoscopic navigation under normal and visually occluded conditions, while also assessing dependence on AI guidance. 32 medical students and interns with basic endoscopic knowledge completed navigation tasks. Accuracy, completion time, situation awareness, cognitive load, user preference, and errors induced by AI false alarms were measured. AI prompts improved navigation accuracy, whereas visualization prompts better supported for situation awareness. Although participantsgenerally preferred AI prompts, false-alarm responses revealed a clear risk of overreliance. The study also introduced a confidence-ring design to communicate AI confidence and uncertainty. Structured AI prompts can support directional decision-making in novice endoscopy training, but their benefits should be balanced against possible reductions in situation awareness and increased dependence. These findings inform the design of intelligent endoscopic navigation support systems for novice users.