Encouraging Emotion Regulation in Social Media Conversations through Self-Reflection

Digital Emotion Regulation (DER) is the practice of employing digital technologies, such as smartphones, to influence one's emotional state. However, online rage, hate speech, or toxicity that stem from ineffective Emotional Regulation (ER) often harms online well-being. Earlier research has focused on identifying ER with multi-modal sensors and studying daily DER practices, but the contextual dynamics of DER remain unexplored. This work provides an intervention for supporting DER in online conversations. It introduces a graph-based framework to identify the need for ER in these conversations. Additionally, it informs users of their emotional influence on a conversation, prompting self-reflection. Through interaction design, this work introduces a technique to de-anonymize online users, encouraging accountability and responsible online behaviour to maintain digital well-being. We collected 2.5K tweets from major news outlets and Australian state Premiers over a year, analysing emotions at individual and group levels in Twitter conversations. Compared to Google's Perspective API, our framework reduced toxicity by 10% on test data, surpassing Google's API by 3%.