Coordinated inauthentic behavior and information spreading on Twitter

2022-06-01

Decision Support Systems

AI Summary

This work investigates how coordinated inauthentic accounts (e.g., bot clusters) on Twitter manipulate information diffusion algorithms and amplify misinformation through temporally consistent anomalous behaviors. To address the limitation of existing methods—namely, their failure to model multi-level propagation dynamics—we propose the first systematic framework for modeling coordinated misinformation behavior. Our approach integrates graph neural networks, temporal behavioral clustering, causal inference over propagation paths, and multi-source feature fusion. We design a detection model grounded in behavioral temporal consistency, achieving 89.7% accuracy in identifying coordinated inauthentic accounts on a real-world Twitter dataset. Furthermore, we uncover three novel information manipulation topologies—previously unreported—revealing positional preferences of coordinated actors within propagation chains, characteristic delay patterns, and mechanisms by which coordination amplifies influence.

Technology Category

• Natural Language Processing: Fact-Checking / Misinformation Detection (NLP Focus)
• Application Domains: Misinformation & Fake News
• Multiagent Systems: Coordination and Collaboration

Application Category

• Social Networks and Social Media: Influence propagation, information diffusion, and the prediction on networks
• Web Mining and Content Analysis: Content-based information diffusion
• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs

Problem

• Analyzing coordinated users' impact on Twitter information spread.
• Identifying differences in behavior between coordinated and non-coordinated accounts.
• Quantifying coordinated accounts' influence on cascade metrics and user interaction.

Innovation

• Quantify coordinated users' tactics efficacy
• Introduce new measures for infectivity and interaction
• Identify saturation-like interaction pattern threshold