Disagreeing Rationales: Rethinking Classification and Explainability Evaluation in Hate Speech Detection
This work addresses the challenge of evaluating human labels and rationales in subjective NLP tasks like hate speech detection, which is important for researchers and practitioners developing explainable AI systems.
This paper explores the impact of human disagreement on classification and explainability evaluation in hate speech detection, focusing on token-level human rationales. It systematically re-implements and evaluates diverse models and metrics across different label and rationale representation spaces, finding that softer representations are more effective in capturing human variation.
Human disagreement is ubiquitous and well-known in labeling. However, variation in explanations, captured through token-level human rationales, remains far less explored. At the same time, it is unclear how to best evaluate human labels and rationales -- or even how to best aggregate rationales beyond majority vote -- in light of this variation. Yet, rationales may provide additional insights into the richness of human reasoning, that may differ in style, values and interpretations -- especially in subjective NLP tasks like hate speech detection. In this work, we unify diverse models, training strategies, loss functions, and existing evaluation metrics under a single protocol by systematically re-implementing them across different label and rationale representation spaces. Classification metrics are organized around two key properties -- predictive and distributional -- while explainability metrics through three complementary dimensions: plausibility, faithfulness, and complexity. In this unified supervision framework, we evaluate model behavior across classification and explainability metrics, as well as metric sensitivity to the choice of label (hard and soft) and rationale representation space (hard, intermediate and soft). Results show that both hard and soft metrics favor softer representations, highlighting their effectiveness in capturing variation and the need to rethink evaluation in subjective NLP.