Human Decision-making Strategy Analysis during Large-scale Disaster

In recent decades, the frequency and intensity of natural disasters have increased significantly. Following these unexpected disasters, understanding and predicting human decision-making strategies during disasters will play a vital role in planning effective humanitarian relief, disaster management, and long-term societal reconstruction. However, such research is challenging due to the unavailability of reliable and large-scale human mobility data. In addition, these personal strategies during disasters become hard to predict due to the influences of multiple factors. In the decision-making process, what key (sometimes hidden) factors do people consider when making a particular mobility decision following a large-scale disaster? In this study, we construct a large human mobility database collected from mobile devices (GPS records of 66,598 users in the Greater Tokyo area), and analyze the human decision-making strategies following Japan’s 2011 Tohoku earthquake. Except for users’ location-based information, we also extract the grid-based stay areas from historical check-ins (normal days before disasters) and discover the functions of those areas combined with points of interest (POI) located in an area. In the final, we jointly model the users’ location-based factors and historical check-ins information, and develop an empirical prediction for decision-making strategies (i.e., destination, travel mode, and departure time from their workplaces) following the Tohoku earthquake. An explainable analysis is conducted to explore the fundamental laws that govern human mobility following disasters. The results are encouraging to support future interventions in analyzing decision-making strategies during disasters, and understanding the relationship between decision making and considered multiple factors/features (as well as the relationship between feature values and possible predicted values).