The rise of mass shootings and the influence of social media highlight the need for proactive measures. Social platforms amplify violence and may even validate potential perpetrators. In contrast to prevailing research, which often focuses on post-event analysis using sequential machine learning models, our proactive approach predicts Twitter users’ emotional responses to a mass shooting by inferring pre-event topics discussed in their tweets. We design a Topic Modeling Graph for forecasting post-event reactions, based on the premise that users discussing similar everyday topics exhibit similar reactions when confronted with mass shootings. We collected tweets from approximately 82,000 users following three separate mass shootings, creating novel datasets. Using our graph-based approach, we achieved a maximum Accuracy of 75%. Anticipating users’ reactions enables proactive intervention strategies, such as psychological support and targeted interventions, aimed at preventing potential new massacres fueled by widespread negativity.