A Study on Information Requirements for Evacuees in Augmented Reality (AR)-Based Evacuation Route Guidance During Flood Disasters

This study presents a mixed-methods framework that integrates latent Dirichlet allocation (LDA)-based topic modeling and content analysis to identify the information requirements for augmented reality (AR)-assisted evacuation guidance systems during flood disasters. A corpus of academic papers and operations manuals related to disaster evacuation was compiled and processed using a seven-stage preprocessing pipeline, and an LDA model was constructed to extract latent topics. The extracted topics were assigned to five theoretically grounded categories with confirmed inter-rater reliability (Cohen's κ = 0.72), and the category distributions between the two source types were explored through a comparative analysis. The route and shelter location information, AR interface and information delivery mechanisms, behavioral decision support information, and hazardous environment information were identified as the core information requirements for AR evacuation guidance systems. In conclusion, these findings provide an empirical foundation for the information design of AR-based evacuation systems and highlight the research–practice gap that must be bridged to provide effective real-time guidance for evacuees in flood disaster scenarios.