Agent-Based Simulation of Social Networks for Disinformation Research

This paper addresses the challenges of studying online social networks (OSNs) by proposing a simulation framework. The framework's key strength lies in its realism and explainability, achieved through agent-based modeling with demographic-based personality traits, finite-state behavioral automata, and an LLM-powered generative module for context-aware posts. The integration of a disinformation campaign module (red module) and a Mastodon-based visualization layer further enhances the framework's utility for studying information dynamics and the effects of disinformation. This is a valuable contribution because it provides a controlled environment to study complex social phenomena that are otherwise difficult to analyze due to data limitations and ethical concerns.