RSD framework for the detection of covert online harms
Abstract
AbstractIn this paper, I propose the RSD framework for computational detection of covert harmful online behaviors. The framework’s three dimensions of Repetition, Severity and Disregard are used together to assess the harmfulness of norm-violating behaviors. These are observable and operationalizable dimensions that aim to support computational detection, enabling evidence-based identification of harmful behavior in conjunction with community norms. While the computational detection of overt harmful behaviors like aggression and hate speech has received a lot of attention, covert harmful behaviors like trolling have been much more challenging to identify. I argue that covert behaviors can be traced by models that examine norm violations crossing the thresholds of the dimensions of RSD.