Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Simulating Influence Dynamics with LLM Agents

Mehwish Nasim, Syed Muslim Gilani, Amin Qasmi, Usman Naseem

TL;DR

The paper addresses how AI-enabled LLM agents influence public opinion within social networks and how counter-misinformation strategies can be evaluated in a wargame-inspired setting. It introduces a simulator that merges opinion-dynamics ABMs with LLM-based agents (Red and Blue) operating on a population of Green Nodes, where messages have potency and costs, and opinions update within a bounded range $[0,1]$ under a convergence parameter $\mu$ and confidence bound $\epsilon$. Key contributions include modeling adversarial information environments, resource-constrained messaging, and providing open-source access with network-upload or generation options, plus metrics like network alignment distribution and polarisation. The framework enables researchers to study influence dynamics, misinformation mitigation, and policy interventions in a reproducible environment.

Abstract

This paper introduces a simulator designed for opinion dynamics researchers to model competing influences within social networks in the presence of LLM-based agents. By integrating established opinion dynamics principles with state-of-the-art LLMs, this tool enables the study of influence propagation and counter-misinformation strategies. The simulator is particularly valuable for researchers in social science, psychology, and operations research, allowing them to analyse societal phenomena without requiring extensive coding expertise. Additionally, the simulator will be openly available on GitHub, ensuring accessibility and adaptability for those who wish to extend its capabilities for their own research.

Simulating Influence Dynamics with LLM Agents

TL;DR

The paper addresses how AI-enabled LLM agents influence public opinion within social networks and how counter-misinformation strategies can be evaluated in a wargame-inspired setting. It introduces a simulator that merges opinion-dynamics ABMs with LLM-based agents (Red and Blue) operating on a population of Green Nodes, where messages have potency and costs, and opinions update within a bounded range under a convergence parameter and confidence bound . Key contributions include modeling adversarial information environments, resource-constrained messaging, and providing open-source access with network-upload or generation options, plus metrics like network alignment distribution and polarisation. The framework enables researchers to study influence dynamics, misinformation mitigation, and policy interventions in a reproducible environment.

Abstract

This paper introduces a simulator designed for opinion dynamics researchers to model competing influences within social networks in the presence of LLM-based agents. By integrating established opinion dynamics principles with state-of-the-art LLMs, this tool enables the study of influence propagation and counter-misinformation strategies. The simulator is particularly valuable for researchers in social science, psychology, and operations research, allowing them to analyse societal phenomena without requiring extensive coding expertise. Additionally, the simulator will be openly available on GitHub, ensuring accessibility and adaptability for those who wish to extend its capabilities for their own research.

Paper Structure

This paper contains 6 sections, 3 figures.

Table of Contents

  1. Introduction
  2. Scenario
  3. Simulation Dynamics
  4. Evaluation
  5. Conclusions and Future Work
  6. Ethics Statement

Figures (3)

  • Figure 1: Architecture of the model. Each agent (red/blue), broadcast a message with a potency to affect the population. The nodes in the network receive those messages. They also interact with their direct neighbors. During the interaction they may change their opinion.
  • Figure 2: User is prompted to confirm the settings and enter a topic.
  • Figure 3: Polarisation in the network over time.