The Role of Emotions in Informational Support Question-Response Pairs in Online Health Communities: A Multimodal Deep Learning Approach
Mohsen Jozani, Jason A. Williams, Ahmed Aleroud, Sarbottam Bhagat
TL;DR
This paper examines how emotions influence informational support in online health communities by labeling ISSQ-ISR pairs from MedHelp and applying a multimodal deep learning pipeline that fuses text, emotional cues, and numeric metadata. Using large language models and SHAP explanations, the authors show that emotional signals meaningfully contribute to predicting ISSQs and ISRs, and that informational support robustly predicts perceived helpfulness. The work demonstrates strong predictive performance (ISSQ accuracy ~95.3%, ISR ~90.7%) and transferability to new sub-communities (up to 87.6% accuracy), refining social support theory and offering practical avenues for emotion-aware decision aids and AI chatbots in healthcare contexts. The findings highlight the intertwined nature of emotional and informational support in OHCs and underscore the value of explainable multimodal models for understanding and enhancing information exchange online.
Abstract
This study explores the relationship between informational support seeking questions, responses, and helpfulness ratings in online health communities. We created a labeled data set of question-response pairs and developed multimodal machine learning and deep learning models to reliably predict informational support questions and responses. We employed explainable AI to reveal the emotions embedded in informational support exchanges, demonstrating the importance of emotion in providing informational support. This complex interplay between emotional and informational support has not been previously researched. The study refines social support theory and lays the groundwork for the development of user decision aids. Further implications are discussed.