Detecting Social Media Rumor Debunking Effectiveness During Public Health Emergencies: An Interpretable Machine Learning Approach.

This article investigates the effectiveness of social media rumor debunking during public health emergencies, focusing on COVID-19-related misinformation on the Chinese platform Sina Weibo. It introduces a novel Debunking Effectiveness Index (DEI) that quantifies rumor debunking outcomes by integrating both positive and negative audience interactions, addressing limitations of prior metrics that considered only interaction volume. Using an interpretable machine learning approach, the study identifies key factors influencing debunking effectiveness—including debunker credibility, content features, communication channels, contextual consistency, and rumor characteristics such as type, topic, and public involvement—and finds significant variation in effectiveness across these dimensions. The findings reveal that while debunking generally has beneficial effects, a substantial backfire effect exists, with many debunking efforts exhibiting weak overall impact, underscoring the need for tailored strategies in rumor governance on social media during health crises.