Unveiling Emotional Signature in Conspiracy Topics on Social Media Through Vector Analysis

This study explores the emotional signature of online discussions within conspiracy, non-conspiracy, and real echo chamber communities on Reddit. Using a RoBERTA-based emotion recognition model, we create numerical representations of emotional intensities across 27 categories. Cosine similarity analysis reveals that real echo chambers exhibit the highest emotional homogeneity, with strong internal alignment and reinforced emotional tones. Conspiracy communities show moderate-to-high emotional similarity, suggesting partial echo chamber behavior, whereas non-conspiracy discussions display the lowest internal consistency, indicating greater emotional diversity. Furthermore, Principal Component Analysis (PCA) highlights a continuum of emotional reinforcement, where non-conspiracy subreddits demonstrate wider emotional dispersion, while conspiracy forums cluster closer to real echo chambers. These findings suggest that emotional vectorization serves as an effective method for identifying echo chamber dynamics and distinguishing conspiracy communities from more diverse online discussions. Our approach provides a foundation for further research into emotion-driven behavioral patterns in digital environments, with potential implications for misinformation detection and online discourse analysis.