EMA-Anchored Mental Health Surveillance: Calibrating Digital Traces with Redundancy/Uniqueness Decomposition

Montag & Hall (2023) call for stronger real-time digital surveillance to guide mental health policy, but many systems rely on biased self-report or platform-specific artifacts. The Smart Platform studies (Katapally et al., 2024 PLOS; 2024 medRxiv) show EMAs reduce recall bias vs retrospective surveys. We propose rolling, small EMA panels recruited as digital citizen scientists to anchor and recalibrate mental health classifiers trained on social media/search streams. A multilevel Bayesian calibration maps EMA-derived symptom trajectories to platform features, adjusting for concept drift. To avoid overweighting redundant signals, we bring in Gaussian partial information decomposition with finite-sample bias correction (Venkatesh et al., 2023) to quantify how much unique vs redundant information Twitter, Reddit, Google Trends, etc., contribute about EMA-measured states. This is different from standard fusion in Madamalla (2025): we explicitly measure and penalize redundancy while prioritizing unique, validated signal components. The approach promises more stable surveillance indices, better cross-platform interpretability, and clearer guidance on which streams actually add value for mental health monitoring.

References:

1. Digital Epidemiology in Action: A Cross-Platform Review of Social Media and Internet-Based Surveillance for Infectious Disease Outbreaks. Esha Madamalla (2025). International Journal of Innovative Science and Research Technology.
2. Enhancing real-time digital surveillance can guide evidence-based policymaking to improve global mental health. C. Montag, Brian J. Hall (2023). Nature Mental Health.
3. How can digital citizen science approaches improve ethical smartphone use surveillance among youth: traditional surveys versus ecological momentary assessments.. T. Katapally, S. Alakshar, S. Tolulope Ibrahim (2024). medRxiv.
4. How can digital citizen science approaches improve ethical smartphone use surveillance among youth: Traditional surveys versus ecological momentary assessments. Sarah Al-akshar, S. Tolulope Ibrahim, T. Katapally (2024). PLOS Digital Health.
5. Gaussian Partial Information Decomposition: Bias Correction and Application to High-dimensional Data. Praveen Venkatesh, C. Bennett, Samuel D. Gale, Tamina Ramirez, Greggory Heller, S. Durand, Shawn R. Olsen, Stefan Mihalas (2023). Neural Information Processing Systems.