Network Hawkes Auditing Model: A Causal Framework for Detecting Algorithmic Influence in Social Media

The pervasive influence of algorithmic filtering on social media platforms raises critical concerns about misinformation, polarization, and democratic integrity, yet existing auditing methods struggle to causally attribute behavioral changes to platform interventions. Current approaches typically fail to account for the complex interplay between social network effects, temporal dependencies, and algorithmic influence, leading to biased estimates and spurious detections. This paper introduces the Network Hawkes Auditing Model (NHAM), a novel causal inference framework that combines multivariate Hawkes processes with network theory to isolate algorithmic effects from organic social dynamics. Our approach models user actions as self-exciting processes where engagement intensities depend on baseline preferences, social influence from network neighbors, and algorithmic content stimulation. The core methodological contribution is the Hawkes Difference-in-Intensities Test, which compares observed user behavior under the platform's actual filtering against a carefully constructed counterfactual baseline under a neutral reference algorithm. We establish theoretical guarantees for parameter consistency and asymptotic normality, and demonstrate through extensive simulations that NHAM achieves substantially higher statistical power than existing methods, particularly in scenarios with strong network confounders. The framework scales efficiently to large user populations and provides interpretable measures of algorithmic influence that can inform evidence-based platform regulation. By bridging causal inference with point process theory and network science, NHAM represents a significant advancement in algorithmic accountability methodology with important implications for platform governance and digital policy.