N°24-78: Quantification of the Self-Excited Emotion Dynamics in Online Interactions

Emotions are essential for guiding human behavior, particularly in social interactions. In modern societies, a growing share of human interactions are taking place online which has been shown to amplify and distort the expression and perception of emotions. However, the entanglement across different emotions is not fully understood. We use a multivariate Hawkes self-excited point process to model and calibrate the temporal expressions of six basic emotions in YouTube live chats. This allows us to understand interdependencies among emotions, but also to disentangle the influence from the video content and social interactions with peers. Positive emotions are found to be more contagious, while negative emotions tend to leave a longer-lasting impression on users' memories. Furthermore, we quantify the endogeneity of online emotion dynamics and find that peer interactions drive user emotional expressions 3-5 times more than passive content consumption. This underscores the powerful incentives of social interactions and the potential risk of emotional manipulation through the use of modern chatbots.