A Large Language Model-Driven Agent-Based Modeling Framework with Multi-Round Communication for Simulating Vaccine Opinion Dynamics

2025-05-19 · Source: cs.MA updates on arXiv.org · Field: Science & Research — Artificial Intelligence & Machine Learning, Robotics & Autonomous Systems, Social Sciences & Behavioral Studies · Depth: Expert, extended

Summary

A new framework integrates the Qwen3-8B Large Language Model into agent-based modeling to simulate vaccine opinion dynamics, investigating how specific cognitive modules influence individual decisions and macro-level social phenomena. The study initializes 95 agents with heterogeneous demographic and socioeconomic profiles, connected by household, workplace, and social media networks. It evaluates four scenarios—Baseline, Memory, Prompt Diversity, and Combined—over 10 independent batch runs, each with 10 time steps. Results show that the prompt diversity module significantly increases vaccination rates and average opinion, leading to 52.9% vaccinated agents. Conversely, the memory module fosters resistance to opinion change, resulting in the lowest vaccination rate at 32.9% and amplifying repulsive social influence from 31.0% (baseline) to 38.6%. The framework successfully reproduces non-linear social influence patterns, including both assimilative and repulsive dynamics, and demonstrates potential for Level 3 validation of agent-based models.

Key takeaway

For research scientists designing agent-based models for social simulations, particularly opinion dynamics, you should integrate LLM-driven agents with multi-round communication and cognitive modules. This approach captures non-linear social influence, like repulsion and threshold effects, which traditional rule-based models miss. Be mindful that modules like memory can increase resistance to opinion change, while prompt diversity can enhance communication effectiveness, leading to complex, opposing macro-level outcomes.

Key insights

LLMs enable agent-based models to simulate complex, non-linear social influence and cognitive processes in opinion dynamics.

Principles

• Cognitive modules profoundly shape emergent opinion dynamics.
• Social influence is non-linear, showing assimilation and repulsion.
• Multi-round dialogue is key for realistic opinion formation.

Method

An LLM (Qwen3-8B) drives multi-round dialogues and reflections among agents with heterogeneous profiles and social networks. Opinions update mathematically based on these interactions.

In practice

• Simulate opinion polarization and consensus formation.
• Evaluate cognitive module impact on social phenomena.
• Model realistic decision-making in public health.

Topics

• Large Language Models
• Agent-Based Modeling
• Opinion Dynamics
• Social Simulation
• Cognitive Modules
• Vaccination Opinions
• Qwen3-8B

Best for: AI Scientist, Research Scientist

Related on AIssential

• Modeling U.S. Attitudes Toward China via an Event-Steered Multi-Agent Simulator — Event-Steered Multi-Agent Simulators (ES-MAS) with dual-stream data and dynamic news-driven interactions accurately model complex, real-world opinion evolution.
• Position: AI Agents Are Not (Yet) a Panacea for Social Simulation — LLM agents for social simulation require explicit environment modeling and rigorous evaluation beyond mere plausibility.
• Using Cognitive Models to Improve Language Model Simulation of Human Persuasion Games — Cognitive models and RL can guide LLMs to simulate diverse human decision-making, improving training environments.
• ScioMind: Cognitively Grounded Multi-Agent Social Simulation with Anchoring-Based Belief Dynamics and Dynamic Profiles — ScioMind combines structured opinion dynamics with LLM reasoning for more realistic social simulation.
• Simulation of Language Evolution under Regulated Social Media Platforms: A Synergistic Approach of Large Language Models and Genetic Algorithms — The framework simulates adversarial language evolution on social media using LLM-driven multi-agent systems and Genetic Algorithms.
• Network Effects and Agreement Drift in LLM Debates — LLM agent opinion dynamics are shaped by network structure, group imbalance, and a systematic "agreement drift" bias.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by cs.MA updates on arXiv.org.