Subjective $\textit{Isms}$? On the Danger of Conflating Hate and Offence in Abusive Language Detection
This addresses a critical methodological issue in abusive language detection for NLP researchers, highlighting a potential flaw in current approaches.
The paper argues that treating annotator subjectivity as equally valid in hate speech detection is problematic because it conflates hate with offense, which can invalidate research findings, and calls for future work to ground such detection in theory to separate these concepts.
Natural language processing research has begun to embrace the notion of annotator subjectivity, motivated by variations in labelling. This approach understands each annotator's view as valid, which can be highly suitable for tasks that embed subjectivity, e.g., sentiment analysis. However, this construction may be inappropriate for tasks such as hate speech detection, as it affords equal validity to all positions on e.g., sexism or racism. We argue that the conflation of hate and offence can invalidate findings on hate speech, and call for future work to be situated in theory, disentangling hate from its orthogonal concept, offence.