Optimization of Conditions for Issuing Warnings Considering the Learning Process of Disaster Experience

抄録

Natural disasters have intensified in recent years due to climate change. Local governments issue disaster warnings to reduce casualties, but they often do not account for the "cry wolf " effect, in which is the phenomenon where residents lose trust in disaster warning when they frequently receive false warnings. As a result, casualties may increase, as residents may fail to evacuate due to their decreased trust in the accuracy of the warning. On the other hand, if local governments issue disaster warnings based on stringent conditions to avoid the “cry wolf” effect, some residents may fail to evacuate. We propose the resident’s evacuation choices model by considering the “cry wolf” effect. Specifically, we expressed the “cry wolf” effect by updating the subjective probability using the Inverse Bayesian inference. Further, we conduct Monte Carlo simulation by applying the proposed model to explore optimal warning conditions that minimize the total personal cost.