The article discusses a new simulation framework developed by researchers from the University of Murcia, aimed at understanding the spread of misinformation online. The framework generates synthetic online social networks populated by agents with distinct personality traits and predictable behaviors, allowing for detailed analysis of information flow. A significant feature is the inclusion of a system to simulate coordinated disinformation campaigns, validating its effectiveness through rigorous analysis of network structure, agent behavior, and the language of generated posts.
Central to this research is the integration of Large Language Models (LLMs), such as Gemini, to create realistic social media content that aligns with each agent’s profile. The agents behave in a controlled manner, allowing researchers to explore social phenomena on platforms like Twitter and Mastodon. With the ability to simulate up to one million agents, the framework offers customizable environments for examining the impact of disinformation and analyzing social media interactions.
The study yields synthetic social networks that reflect real-world characteristics, including patterns of homophily and triadic closure, as well as structural and behavioral realism. The framework features a visualization layer for real-time monitoring, making it a robust testbed for evaluating the dynamics of online harm and the effectiveness of intervention strategies against misinformation and disinformation campaigns. The research highlights the potential of this framework for further studies in computational social science, bot detection, and natural language processing.
