Protective Actions in the Modern Information Environment: Agent-Based Modeling of Hurricane Evacuation Decisions

As forecasts of hazardous weather grow more accurate, natural and social scientists are increasingly interested in how weather information is communicated to populations at risk and used in protective action decisions (e.g., evacuation). To this end, social scientists are often deployed after major hurricanes to interview residents of impacted areas to document if, how, and why those residents took protective actions before the storm reached their area. In many areas of North America, forecasts about weather hazards are passed through many and diverse information conduits that span traditional broadcast media and new social media networks. Repeated re-publishing of weather information can introduce error in the content and can lag well behind current forecasts, contributing to both vulnerable residents failing to take protective actions and other residents taking protective actions unnecessarily.

To better understand how information moving through digital and non-digital social networks affects risk assessment and decisions to take protective actions, we have developed an agent-based model (ABM) implementation of Lindell and Perry's (2012) Protective Action Decision Model (PADM) in the context of an approaching tropical cyclone. Simulations of our ABM are run on the well-known NetLogo platform. Unlike previous ABMs of short-term behavioral responses to hurricanes -- which have studied, for example, the feasibility and logistics of evacuating specific locations -- this model instead focuses on the information acquisition and cognitive processes that influence agents' decisions to take protective actions. This model is focused on helping researchers (in both social science and meteorology) to explore decision-making in the modern information environment in a virtual laboratory that can simulate the days leading up to a hurricane landfall.

The current version is an abstract (rather than semi-realistic) ABM of social processes, but is designed to be integrated with computational models of tropical cyclones and related storm surges as part of the NSF-supported Communicating Hazard Information in the Modern Information Environment (CHIME) project. The model has been calibrated initially using published social science survey data about protective actions during hurricane events -- and will be further refined with input from social media studies (e.g., Twitter) and focus groups.