Subjective $\textit{Isms}$? On the Danger of Conflating Hate and Offence in Abusive Language Detection

TL;DR

The paper critiques the use of annotator subjectivity in hate and abusive language detection, arguing that hate and offence are orthogonal and should be disentangled. It proposes a theoretical framing that treats isms as culturally defined societal norms while viewing offence as a subjective individual experience. The authors advocate a new formulation of isms as cultural formations of norms, along with actionable recommendations for annotation schemas and annotator recruitment that reflect lived experience. This reframing aims to produce more robust, policy-relevant NLP artifacts and to better reflect the experiences of marginalised communities in abusive language research.

Abstract

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.