Network-Based Modeling for Characterizing Human Collective Behaviors during Extreme Events

Modeling and predicting human dynamic behaviors in the face of stress and uncertainty can help understand and prevent potential irrational behavior, such as panic buying or evacuations, in the wake of extreme events. However, in terms of the types of events and the distinct human psychological factors, such as risk perception (RP) and emotional intensity (EI), human dynamic behaviors exhibit heterogeneous spatiotemporal characteristics. For example, we can observe different collective responses to the same events by people in different regions, with distinct trends unfolding over time. To provide a computational means for understanding the spatiotemporal characteristics of human behaviors during different types of extreme events, here we present a network-based model that enables us to characterize dynamic behaviors. This model assumes the perspective of a dynamic system, whose behavior is driven by human psychological factors and by the network structure of interactions among individuals. By making use of the available data from Twitter and GoogleTrends, we conduct a case study of human dynamic behavioral and emotional responses to the Japanese earthquake in 2011 in order to examine the effectiveness of our proposed model. With this model, we further assess the impacts of an event by evaluating the interrelationships of human RP and levels of EI in terms of observed collective behaviors. The results demonstrate that human behaviors are subjected to personal observations, experiences, and interactions, which can potentially alter perceptions and magnify the impacts of an event.