CAREER: An Integrated Trustworthy AI Research and Education Framework for Modeling Human Behavior in Climate Disasters

天气和气候灾害的频率和严重程度的增加对人类生命、民用基础设施和经济造成了更大的危害。建设抗灾能力的一个重要步骤是对灾难发生之前、期间和之后的人类行为有深刻的理解和健全的建模。该学院早期职业发展（Career）项目支持可信赖的人工智能（AI）方法的研究，以改变天气和气候灾害中的行为建模，并推进STEM教育和应急管理实践。这种基于理论的高分辨率模型是可以解释的，并且在支持时间和安全关键决策时，它们的偏差被最小化。该项目有助于培养下一代科学家和工程师，以建立和利用尖端的人工智能工具进行灾害管理。利用人工智能增强灾害中人类行为模型的研究正在兴起，但仍处于起步阶段。值得注意的是，目前的许多模型都是一个黑匣子，没有彻底考虑可信度、偏见和公平性等关键问题。该项目通过开发一个可信赖的人工智能研究和教育框架来填补知识空白。主要有五项工作：1)将基础理论与数据驱动的人工智能方法结合起来，提高对天气和气候灾害中人类行为的预测；2)描述基于人工智能的行为模型中存在的公平问题，减少固有的偏见和不公平；3)解释模型输入和输出，以获得独特的见解和推进理论；4)将研究成果转化为高中和本科工程课程；5)为应急管理人员和交通运输专业人员提供培训机会，以提高他们的技术能力。总的来说，这个职业项目有可能显著改善应急管理、社会公平和社区复原力。该奖项反映了美国国家科学基金会的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

Increase in frequency and severity of weather and climate disasters has resulted in greater harm to human lives, civil infrastructures, and the economy. An essential step toward building disaster resilience requires deep understanding and robust modeling of human behaviors before, during, and after catastrophic events. This Faculty Early Career Development (CAREER) project supports research on trustworthy Artificial Intelligence (AI) methodologies to transform behavioral modeling in weather and climate disasters and advances STEM education and emergency management practices. Such theory-informed, high-resolution models are explainable, and bias in them is minimized in support of time- and safety-critical decisions. The project contributes to training of the next-generation scientists and engineers to build and utilize cutting-edge AI tools for disaster management. Research on harnessing AI to enhance human behavioral modeling in disasters is emerging but still in its infancy. Notably, many of the current models function as a black box and have not thoroughly considered critical issues such as trustworthiness, bias, and fairness. This project fills the knowledge gap by developing an integrated trustworthy AI research and education framework. Five main tasks are pursued: 1) Coupling fundamental theories with data-driven AI approaches to improve the prediction of human behaviors in weather and climate disasters; 2) characterizing fairness issues present in AI-based behavioral models, and reducing inherent bias and inequity; 3) interpreting model inputs and outputs to gain unique insights and advance theories; 4) transferring research outcomes into high school and undergraduate engineering curricula; and 5) providing training opportunities to emergency managers and transportation professionals to enhance their technological competency. Collectively, this CAREEER project has the potential to yield significant improvement to emergency management, social equity, and community resilience.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.