A Media–Cognition–Interaction Framework for Group Emotion Dynamics: Dual Mechanisms on Multiple Social Platforms in the Era of Generative AI

Although understanding group emotion dynamics is vital, existing research lacks large-scale, parallel investigations within distinct platforms that measure emotional polarization through communicative connections and analyze its multi-level drivers. Focusing on Generative Artificial Intelligence (GenAI), this study explains why group emotions diverge on various platforms by integrating user-level cognitive foundations with platform-specific media mechanisms and interaction structures, drawing on group dynamics theory. Methodologically, we move beyond conventional sentiment aggregation by incorporating communication chains of user interactions, yielding more precise estimates of group emotion. To ensure comparability, we developed the MMIIC crawler, which unifies heterogeneous interaction structures on Weibo, Bilibili, and Douyin, producing a dataset of 860,000 comments. We identify a dual mechanism behind group emotion divergence: consistent cross-platform drivers rooted in user cognition (e.g., fragmented information consumption and low technical literacy among the young and less-educated), and platform-specific amplifiers (e.g., Weibo’s high rate of unanswered posts and celebrity-centric content) that intensify polarization through unique interaction designs. Theoretically, this study formalizes a Media-Cognition-Interaction (MCI) framework that integrates multi-platform constants with platform-specific amplifiers, extending group dynamics and field theory to systematically explain emotional polarization in multi-platform environments.