Signaling Game Framework for Adversarial Behavioral Data in Epidemiology
Epidemiological models that depend on crowdsourced behavioral data—such as vaccination status, mask usage, and social distancing adherence—face a fundamental challenge: the data is strategically reported, not passively sampled. A new framework, presented in arXiv preprint 2602.20134v2, models this interaction as a signaling game between individuals (senders) and a public health authority (receiver). Participants may misreport to avoid penalties, secure benefits, or express distrust, resulting in systematically corrupted inputs. The paper proposes both a generative model of such distorted data and a mechanism for the receiver to extract reliable signal from it. This approach offers a path to more robust epidemiological modeling under adversarial reporting.
Key facts
- Epidemiological models rely on crowdsourced, self-reported behavioral data.
- This data is strategically reported, making it adversarial to data mining pipelines.
- Individuals misreport to avoid penalties, access benefits, or express distrust.
- The framework casts the interaction as a signaling game.
- It provides a generative model of strategically-corrupted behavioral data.
- It provides a mechanism for the receiver to recover reliable signal.
- The preprint is identified as arXiv:2602.20134v2.
- The announcement type is replace-cross.
Entities
—