The Epi-LLM Framework: probing LLM behavioral priors through epidemiological agent-based models

2026-06-01 · Source: Artificial Intelligence · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Robotics & Autonomous Systems, Emerging Technologies & Innovation · Depth: Expert, quick

Summary

The Epi-LLM framework, introduced on 2026-06-01, integrates agent-based modeling, real-life epigames, and large language models to simulate human behavior during epidemics. This novel system creates a synthetic society of agents that reason and adapt dynamically within an outbreak contact network. Across four different LLM architectures, the framework demonstrated that LLM agents reduced peak active infections, with quarantine compliance reaching 58-65% on day six of a 15-day simulation. A binomial generalized linear model identified perceived health severity ($β= 0.33, p = 0.002$) as the strongest predictor of quarantine behavior, yielding a pseudo-$R^2$ of 0.055, comparable to the 0.072 observed in human trials. The study also found that LLM architecture significantly influences epidemic dynamics, suggesting low-variance models for testing behavioral rules and high-variance models for representing real-world decision-making. Explicit attitudinal parameterization, not just geographic labels, is crucial for culturally differentiated behavior. This proof-of-principle work aims to establish Epi-LLM as a scalable, risk-free simulation environment for pandemic preparedness research.

Key takeaway

For research scientists developing pandemic preparedness models, the Epi-LLM framework offers a scalable, risk-free simulation environment. You should consider integrating LLM agents to model complex human behavioral dynamics, noting that perceived health severity is a strong predictor of quarantine compliance. Carefully select your LLM architecture based on whether your goal is internal validity for rule testing or realistic decision-making representation. Explicitly parameterize attitudinal factors to achieve culturally differentiated behaviors in your simulations.

Key insights

The Epi-LLM framework uses LLMs in agent-based models to simulate epidemic behavior, showing LLMs can reduce infection peaks.

Principles

• LLM agents can reduce peak epidemic infections.
• Perceived health severity predicts quarantine compliance.
• LLM architecture choice impacts simulation validity.

Method

The Epi-LLM framework integrates agent-based modeling, real-life epigames, and LLMs to simulate a synthetic society's dynamic reasoning and adaptation over an outbreak contact network.

In practice

• Use Epi-LLM for risk-free pandemic preparedness simulations.
• Parameterize LLM agents with explicit attitudes for cultural nuance.
• Select LLM architecture based on simulation goal (validity vs. realism).

Topics

• Epi-LLM Framework
• Agent-Based Modeling
• Large Language Models
• Epidemic Simulation
• Pandemic Preparedness
• Behavioral Modeling

Best for: AI Scientist, Research Scientist

Related on AIssential

• Modeling Pathology-Like Behavioral Patterns in Language Models Through Behavioral Fine-Tuning — Behavioral fine-tuning can induce specific, stable, pathology-like biases in LLMs, shifting their fundamental interpretive priors.
• ClinicalMC: A Benchmark for Multi-Course Clinical Decision-Making with Large Language Models — Evaluating LLMs in evolving multi-course clinical scenarios is crucial for real-world healthcare deployment.
• Beyond Static Responses: Multi-Agent LLM Systems as a New Paradigm for Social Science Research — A six-tier framework maps LLM agent evolution from static tools to complex adaptive systems for social science research.
• Week Ending 2.8.2026 — AI research is advancing model robustness, efficiency, and real-world applicability across diverse domains.
• Can Large Language Models Revolutionize Survey Research? Experiments with Disaster Preparedness Responses — Theory-constrained retrieval and demographic anchoring enable LLMs to improve survey data imputation and analysis, especially for vulnerable groups.
• Evaluating LLMs as Human Surrogates in Controlled Experiments — LLMs can reproduce directional behavioral effects but often miscalibrate effect magnitudes compared to human data.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Artificial Intelligence.