Beyond content: Multimodal emotional responses predict online moral contagion across laboratory and real-world contexts

Abstract

Research on online moral contagion has highlighted the role of emotion in shaping the diffusion of moralized content. However, existing studies primarily focus on the emotional attributes of content rather than viewers’ actual emotional responses. How individuals’ actual reactions influence diffusion, especially whether positive or negative emotions facilitate spreading, remains unclear. This research employed multimodal measurements to capture college students’ subjective and physiological emotional experiences while viewing moralized short videos, along with their subsequent contagion behaviors. In Study 1, participants watched videos on a computer and reported their sharing intentions. Emotional responses were measured using EEG, ECG, eye tracking, and self-reports. In Study 2, participants viewed and shared videos on mobile devices to better reflect real-world contexts, with emotional responses captured via electrodermal activity, respiration, accelerometers, and gyroscopes. Machine learning models showed that actual emotional responses predicted sharing more accurately than content attributes (80% vs. 69% for intentions; 82% vs. 61% for behaviors). Short videos evoking high arousal or positive valence were more likely to be shared. Shapley value comparisons indicated that EEG, gyroscope, electrodermal activity, self-reports, and accelerometer contributed more to prediction accuracy than respiration, eye tracking, and ECG. EEG had the highest contribution, while the non-intrusive measures, gyroscope and accelerometer, may serve as substitutes for more intrusive methods. These findings extend the theoretical framework of online moral contagion by consistently demonstrating that high-arousal, positively valenced emotional responses facilitate diffusion, and also provide practical guidance for selecting effective modalities and implementing non-intrusive emotional assessments in future research and applications.