Dynamics of local interactions and evacuation behaviors in a social network

Abstract This paper examines how social interaction affects the choice to evacuate during a network formation process. The primary objective is to evaluate in detail the influence of other individuals by considering how influencers and entire network structures affect one-to-one interactions. This study proposes an analytical framework for the simultaneous evaluation of local interaction and face-to-face communication network formation. This framework is appropriate for dynamic scenarios, because local interactions strongly influence human decision-making and because network formation changes over time. Our detailed local interaction model, which is based on Brock and Durlauf (2001), analyzes the differences in the interaction weights related to pairs and behaviors. We estimate the utility and the asymmetric weight parameters in the local interaction model by using the nested pseudo-likelihood approach. Our proposed network formation model, which is based on the discrete choice model, evaluates the probability of face-to-face communication and the spatial correlations of the interaction pairs in an area. Our case study validates the introduction of (1) asymmetric weights of interaction in the evacuation departure choice and (2) spatial correlation in the network formation model, using the behavioral data collected during a no-notice disaster in a devastated settlement. In numerical simulations, the proposed evaluation framework can effectively illustrate the impact of network structures on the choice probabilities influenced by interactions. Additionally, the framework is useful for evaluating the critical measures for prompt evacuation.