RAPID: Evacuate or Not? Modeling the Decision Making of Individuals in Impending Disaster Areas

点击翻译按钮获取中文摘要

A category 4 hurricane is approaching. Should a potentially affected individual follow the official orders and evacuate, or stay in place? Millions of individuals situated in vulnerable areas face this grave question as imminent disaster threatens. Many choose to leave, whereas some do not. Numerous interviews with such persons clearly convey their conviction in having made the right choice. This RAPID project will identify the variables that significantly influence the decision making of individuals in impending disaster areas, and it will contribute to the understanding of how the variables are utilized differently by different individuals. These insights will help to build new computational models of the individual's decision making under uncertainty, in extreme situations such as hurricanes and other natural disasters. The focus disasters will be the impacts of Hurricane Harvey on the Texas coast and Hurricane Irma on Florida and Georgia. Outcomes could augment evacuation efforts with actions on the ground that target those most likely to ignore official recommendations. Furthermore, such modeling will likely help relief-and-rescue efforts to better coordinate and provide faster relief with increased precision. Outcomes from this research will be integrated into the classroom instruction of courses taught by the PIs, which will provide students with exposure to how decision-making science can have real-world impact even under the most extreme circumstances.The technical approach begins with characterizing the affected classes of individuals of interest. Next, various types of data about them will be gathered. In particular, interviews of affected individuals before the impending disaster and after, as reported by various news agencies, relevant social media messages originating from disaster areas, government data on evacuees and their demographics, and other survey instruments will be used to build a comprehensive data set for analysis. These data will be sifted to infer the significant variables and how they interact in individual decision making. The analysis and data will be used to build empirically-informed decision making models, which will combine principled agent-based modeling with parametric human judgment and choice models. The exploratory nature of this research makes model evaluation particularly important. Performance of the various models on the data will be compared based on their fits and qualitative assessments. This research plan is expected to yield validated models of the decision-making processes of several affected individuals for government use and further study.