Automated Identification of Verbally Abusive Behaviors in Online Discussions

Abstract

Discussion forum participation represents a crucial support for learning and often the only way of supporting social interactions in online settings. However, learner behavior varies considerably in these forums, including positive behaviors such as sharing new ideas or asking thoughtful questions, but also verbally abusive behaviors, which could have disproportionate detrimental effects. To provide means for mitigating potential negative effects on course participation and learning, we developed an automated classifier for identifying communication that show linguistic patterns associated with hostility in online forums. In so doing, we employ several wellestablished automated text analysis tools and build on common practices for handling highly imbalanced datasets and reducing sensitivity to overfitting. Although still in its infancy, our approach shows promising results (AUC ROC=0.74) towards establishing a robust detector of abusive behaviors. We provide an overview of the classification (linguistic and contextual) features most indicative of online aggression.