HALE Framework Combines LLMs and Large-Scale Epidemic Simulations

Researchers presented the HALE framework (Hybrid Agent-based and Language-driven Epidemic), which combines classical agent-based modeling with large language models. The system allows LLMs to predict decisions of millions of simulated people in real time, adapting to real changes. Using COVID-19 in Salt Lake County as an example, scientists showed that the hybrid approach can become a tool for more accurate policy forecasting.

Problem: Static Models

Traditional agent-based models (ABM) can simulate behavior of millions of people and their interactions—invaluable for policy development. However, they rely on static assumptions: agent behavior rules are fixed in advance and don't change.

During an epidemic, human behavior changes radically. People start wearing masks, avoid crowds, change work schedules. A static model cannot account for these shifts in real time—it is locked into the initial scenario and cannot learn from new data.

How HALE Solves the Problem

HALE inserts an LLM into the agent-modeling cycle. Instead of hard-coding behavior rules, the system uses a neural network to dynamically predict each agent's decisions. At each simulation step, the LLM analyzes the current context—number of infected in the area, public fear level, vaccine availability—and predicts a person's decision: will they go to the store, will they maintain distance, will they agree to vaccination.

• Framework uses LLM to forecast behavior at each simulation step
• HALE tested on COVID-19 example in Salt Lake County, Utah
• System adapts to real data instead of following a prescribed scenario
• Scales to millions of agents

Why This Matters for Policy

Epidemiological models are used by governments to make decisions about quarantine, vaccination, and intervention. If the model is inaccurate, policy can cause unnecessary harm or prove ineffective.

HALE increases accuracy through adaptability. The system doesn't just follow a plan but changes as new data appears and population reacts. Researchers showed using COVID-19 in Salt Lake County that HALE reflects real dynamics of infection and behavioral shifts better than static models.

This is especially important for rapid response: if policymakers get more accurate forecasts, they can adjust measures earlier than an infection wave spreads.

What This Means

The hybrid approach shows that LLMs can be not just a tool for analysis, but an active component of simulations themselves. If HALE works on epidemics, it can be applied to other crises: economic shocks, social conflicts, natural disasters. This will open a new class of models—living, adaptive, and driven by neural networks.