Modeling Epidemiological Dynamics Under Adversarial Data and User Deception

A team of researchers has published a novel paper on arXiv that applies game theory to a critical flaw in modern epidemiology: deceptive self-reported data. The work, 'Modeling Epidemiological Dynamics Under Adversarial Data and User Deception,' addresses the growing reliance on real-time behavioral data—like vaccination status or mask usage—which is often strategically misreported by individuals seeking to avoid penalties, access benefits, or express distrust.

The core innovation is a formal game-theoretic framework that models this interaction as a signaling game. In this setup, individuals (the 'senders') choose how to report their true health behaviors, while the public health authority (the 'receiver') must update its epidemiological models based on these potentially distorted signals. The researchers focus specifically on deception around masking and vaccination. They analytically characterize the game's equilibrium outcomes, evaluating how much deception a system can tolerate while still maintaining epidemic control through non-pharmaceutical interventions (NPIs).

Key technical findings show that in 'separating equilibria,' where signals truthfully reveal behavior, infections can be driven to near zero. More remarkably, the analysis demonstrates that even under 'pooling equilibria' characterized by widespread dishonesty, well-designed strategies for both senders and receivers can still maintain effective control of an outbreak. This work, categorized under Computer Science and Game Theory (cs.GT) and Artificial Intelligence (cs.AI), moves beyond traditional statistical corrections by formally incorporating strategic human behavior into the modeling process. It provides a mathematical foundation for building public health models that are robust to the adversarial data environments increasingly seen in real-world crises.