Physiological and Socio-Behavioral Determinants of Viral Video User Engagement

The use of social media platforms like TikTok, Facebook, and YouTube to consume videos has seen significant growth in recent years. Some videos can generate high levels of engagement - views, likes, shares, or comments at a rapid pace, making them “viral.” However, understanding what causes these videos to become viral or predicting virality is still largely a mystery. This study aimed to uncover if biometrics-based emotion and arousal data, such as facial muscle and skin conductance data, could predict user engagement and contribute to video virality. The experiment used 64 participants, who watched 13 videos and had their facial expressions and galvanic skin response (GSR) data recorded throughout the viewing experience. The study then used an XGBoost classifier and 42 collected features, including physiological data and socio-behavioral responses, to predict user engagement (willingness to like, share, or comment) with over 80% accuracy. The results indicate that a combination of facial expression, GSR, and socio-behavioral data can accurately distinguish between high and low user engagement without needing to ask viewers anything about the videos or analyze video content. This study elevates the role of viewers’ physiological and subconscious responses to video content across the viewing experience.