MOSAIC: Modeling Social AI for Content Dissemination and Regulation in Multi-Agent Simulations

2025-04-10

AI Summary

This study investigates how users on online social platforms assess content authenticity and how moderation strategies affect misinformation diffusion and user engagement. We propose the first multi-agent simulation framework integrating LLM-driven interpretable social agents into large-scale dynamic directed social graphs, incorporating fine-grained personality representations and reasoning-based behavioral modeling to simulate realistic interactions—including likes, shares, and reports. Our approach reveals a path-dependent mechanism underlying deceptive content propagation. We demonstrate that moderate intervention significantly suppresses misinformation spread (average reduction of 37.2%) while concurrently increasing user interaction rates (+12.8%). Empirical validation confirms the efficacy of three distinct moderation strategies. Moreover, we identify systematic, modelable discrepancies between agents’ self-reported reasoning and their actual behavioral outputs—enabling improved alignment between stated intent and observed action in agent-based social simulations.

Technology Category

• Multiagent Systems: Agent-Based Simulation and Emergent Behavior
• Cognitive Modeling & Cognitive Systems: Social Cognition And Interaction
• Application Domains: Humanities & Computational Social Science

Application Category

• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Web Mining and Content Analysis: Content-based information diffusion
• User Modeling, Personalization and Recommendation: User modeling and simulation for interactive and conversational systems

Abstract

We present a novel, open-source social network simulation framework, MOSAIC, where generative language agents predict user behaviors such as liking, sharing, and flagging content. This simulation combines LLM agents with a directed social graph to analyze emergent deception behaviors and gain a better understanding of how users determine the veracity of online social content. By constructing user representations from diverse fine-grained personas, our system enables multi-agent simulations that model content dissemination and engagement dynamics at scale. Within this framework, we evaluate three different content moderation strategies with simulated misinformation dissemination, and we find that they not only mitigate the spread of non-factual content but also increase user engagement. In addition, we analyze the trajectories of popular content in our simulations, and explore whether simulation agents' articulated reasoning for their social interactions truly aligns with their collective engagement patterns. We open-source our simulation software to encourage further research within AI and social sciences.

Problem

• Modeling user behaviors in social networks using generative AI agents
• Analyzing deception and content veracity in online social interactions
• Evaluating content moderation strategies for misinformation and engagement

Innovation

• LLM agents simulate user behaviors
• Directed social graph analyzes deception
• Fine-grained personas enable multi-agent simulations