Journal of the Society of Disaster Information ISSN:1976-2208(Print) 2671-5287(Online) 한국재난정보학회논문집

Purpose: This study proposes criteria for documenting disaster response decisions informed by AI risk predictions. Method: Korean legislation and relevant literature were analyzed, and the criteria were applied to a hypothetical inundation scenario through a cumulative comparison varying record content and linkages. Result: Records should connect the handling of important risk information with grounds for evaluating predictions and selecting, maintaining, modifying, or lifting measures. Decisions not to take particular measures should be distinguished from pending or transferred review, confirmed omissions, and unknown review status. Essential information and reasons should be recorded without obstructing emergency response. When even minimal contemporaneous recording is infeasible, the circumstances should be documented once recording becomes practicable. Later supplements should identify their sources and preparation times. Technical logs alone do not establish substantive review or explain response choices. Conclusion: The proposed operational criteria apply existing recordkeeping rules to disaster response. Record completeness does not establish decision legality or predictive accuracy. The hypothetical comparison does not isolate the effects of content from those of linkages. Effects on recording workload and reconstruction accuracy require further examination through exercises and operational data.