EAGER: Generative AI for Synthetic Population Generation for Disaster Impact Assessment

The objective of this EArly-concept Grant for Exploratory Research (EAGER) project is to support research on developing an open source, AI-enabled synthetic population inventory. Such an inventory facilitates disaster impact studies by eliminating the need for complex data generation and downstream simulation tasks to generate behavior insights. A comprehensive and accurate synthetic population dataset that integrates people, the built environment, and behavior plays a critical role in disaster impact assessment. The resulting inventory serves as a crucial data source for training and testing resilience tools as well as informing disaster policies. The advancements in knowledge, models, and algorithms from this project lay the groundwork for leveraging generative AI and multi-domain data fusion to generate synthetic data. This research project puts forward a novel, untested paradigm for synthetic population generation. A new generative AI framework is introduced to learn intricate patterns in microdata to produce diverse and realistic synthetic populations. Explicit household and building relationships are also revealed through machine learning applied to language-embedded household and building features. The project deepens understanding of post-disaster household adaptation through comprehensive surveys and data fusion with large language models. This paradigm has the potential to transform how synthetic populations are used in various fields, making disaster planning and policy development more effective and accessible. 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.