Search papers, labs, and topics across Lattice.
This study develops a spatially grounded, agent-based simulation framework that utilizes large language model (LLM) decision-making to model individual behaviors during infectious disease outbreaks. By integrating LLM-generated decisions into a synthetic population based on real-world census data, the research captures the spatial distribution of demographic groups and their reporting behaviors. Key findings indicate that income and education significantly influence self-reporting rates of influenza-like illness, while geography and message framing also play a role, highlighting the importance of socio-economic factors in public health interventions.
Income and education drive self-reporting rates of illness more than geography or messaging, revealing critical insights for public health strategies.
Modelling individual decision-making during infectious disease outbreaks is crucial for understanding behavioural dynamics and informing effective public health interventions. Prior work has shown that large language models can simulate realistic human behaviour by generating agent decisions based on demographic prompts and situational context. We build on this foundation with a spatially grounded, agent-based simulation framework that integrates LLM-generated decisions about self-reported influenza-like illness into a census-based synthetic population of agents. Location is treated as a central feature: agents are assigned to spatial units within cities, capturing the spatial distributions of different demographic groups using real-world census data and enabling geographically diverse behavioural modelling. We implement and compare three decision scenarios, independent reasoning, household influence, and message framing, and simulate self-reporting outcomes in San Francisco and Atlanta. Results reveal that income and education are the dominant drivers of reporting rate variation, with smaller but consistent effects from geography, LLM model choice, and message framing. Our framework generates synthetic data that captures both social and geographic heterogeneity, supporting spatial epidemiological modelling and bias-aware behavioural analysis.
