Researchers propose HALE framework to combine LLMs with agent-based modeling for epidemic simulation

Agent-based modeling (ABM) is widely used for policy-making because it can simulate millions of individuals and their interactions, but traditional ABMs rely on static priors that limit their ability to adapt to real-time changes. Researchers propose a novel approach to bridge this information gap by integrating large language models (LLMs) into ABMs to predict human decision-making. The team introduces the Hybrid Agent-based and Language-driven Epidemic (HALE) modeling framework, which leverages LLMs to enhance the adaptability of ABMs in dynamic environments.

As a proof-of-concept, the researchers apply HALE to simulate the spread and effects of COVID-19 in Salt Lake County, Utah. The framework demonstrates how LLMs can be used to inform agent behaviors in real time, potentially improving the realism and responsiveness of epidemic models compared to traditional static-prior approaches.

The work is presented as an arXiv preprint and highlights the scalability of HALE, positioning it as a generalizable method for integrating LLMs with ABMs beyond epidemic modeling. The authors include Sifat Afroj Moon, Dakotah Maguire, Adam Spannaus, Joe Tuccillo, Maksudul Alam, Sudip K. Seal, John Gounley, and Heidi Hanson.