Reading the Digital Pulse: Context-Aware AI for Sensing Engagement Drivers in Health Forums

Pouria Rad, Gianluca Zanella, Mohsen Jozani

Confirmation: I have read and agree with the IEEE BSN 2025 conference submission's policy on behalf of myself and my co-authors.

Keywords: digital health, emotion analytics, online health communities, decision support, adaptive intervention

TL;DR: Analyzing 1M+ health posts with an AI model reveals that the emotional drivers of community engagement are not universal but depend on the forum's topic, highlighting the need for context-aware moderation.

Abstract: Online Health Communities (OHCs) are vital for patient support, but the massive scale of user-generated content makes manual monitoring of user well-being and engagement drivers impossible. This study integrates Self-Determination Theory with a large-scale computational analysis of 1.03M questions and 6.0M replies, applying a fine-tuned DistilRoBERTa affect model ($\kappa$=0.74 vs. human) and Poisson regression (pseudo-R2=0.03) to quantify how emotional tone and informational intent predict engagement of over one million posts from the MedHelp platform. Our results show that, while informational contributions consistently predict higher engagement, the impact of emotional expression is highly context-dependent. For instance, expressions of sadness and anger, which have a minor effect on engagement overall, are strongly associated with increased community response in high-stakes forums like the Cancer community. These findings demonstrate the need for context-aware moderation tools and provide a scalable framework for designing more responsive and effective digital health interventions.

Track: 3. Signal processing, machine learning, deep learning, and decision-support algorithms for digital and computational health

NominateReviewer: Pouria Rad, [e-mail]

Submission Number: 33