Decoding Echo Chambers: LLM-Powered Simulations Revealing Polarization in Social Networks

2024-09-28

International Conference on Computational Linguistics

AI Summary

Social media platforms face structural risks such as echo chambers and opinion polarization, yet existing numerically simplified models fail to capture text-driven opinion dynamics. To address this, we propose the first large language model (LLM)-driven social opinion network simulation framework, integrating multi-agent modeling with Bounded Confidence and Friedkin-Johnsen models to enable text-level opinion interaction and dynamic evolution. We innovatively design active and passive textual nudges—semantic-level interventions that modulate information exposure and opinion updating. Experiments demonstrate that our framework faithfully reproduces polarization phenomena, significantly outperforming conventional numerical models in fidelity. Both nudge types effectively mitigate echo chamber effects, reducing their average intensity by 37%. This work bridges the gap between high-fidelity linguistic representation and formal opinion dynamics modeling, offering a scalable, interpretable, and intervention-aware simulation paradigm for computational social science.

Technology Category

• Cognitive Modeling & Cognitive Systems: Social Cognition And Interaction
• Natural Language Processing: (Large) Language Models
• Multiagent Systems: Agent-Based Simulation and Emergent Behavior

Application Category

• Social Networks and Social Media: Media and governance, opinion dynamics, filter bubbles, polarization
• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Web Mining and Content Analysis: Sentiment analysis and opinion mining

Abstract

The impact of social media on critical issues such as echo chambers needs to be addressed, as these phenomena can have disruptive consequences for our society. Traditional research often oversimplifies emotional tendencies and opinion evolution into numbers and formulas, neglecting that news and communication are conveyed through text, which limits these approaches. Hence, in this work, we propose an LLM-based simulation for the social opinion network to evaluate and counter polarization phenomena. We first construct three typical network structures to simulate different characteristics of social interactions. Then, agents interact based on recommendation algorithms and update their strategies through reasoning and analysis. By comparing these interactions with the classic Bounded Confidence Model (BCM), the Friedkin Johnsen (FJ) model, and using echo chamber-related indices, we demonstrate the effectiveness of our framework in simulating opinion dynamics and reproducing phenomena such as opinion polarization and echo chambers. We propose two mitigation methods, active and passive nudges, that can help reduce echo chambers, specifically within language-based simulations. We hope our work will offer valuable insights and guidance for social polarization mitigation.

Problem

• Simulating social opinion networks to evaluate polarization effects.
• Addressing limitations of traditional models in capturing text-based communication.
• Proposing methods to mitigate echo chambers in language-based simulations.

Innovation

• LLM-based simulation for social opinion networks
• Agents interact via recommendation algorithms
• Active and passive nudges reduce echo chambers