Moderating the Good, the Bad, and the Hateful: Moderators’ Attitudes Towards ML-based Comment Moderation Support Systems

Abstract

Comment sections have established themselves as essential elements of the public discourse. However, they put considerable pressure on the hosting organizations to keep them clean of hateful and abusive comments. This is necessary to prevent violating legal regulations and to avoid appalling their readers. With commenting being a typically free feature and anonymity encouraging increasingly daunting comments, many newspapers struggle to operate economically viable comment sections. Hence, throughout the last decade, researchers set forth to develop machine learning (ML) models to automate this work. With increasingly sophisticated algorithms, research is starting on comment moderation support systems that integrate ML models to relieve moderators from parts of their workload. Our research sets forth to assess the attitudes of moderators towards such systems to provide guidance for future developments. This paper presents the findings from three conducted expert interviews, which also included tool usage observations.