HateCheck: Functional Tests for Hate Speech Detection Models

TL;DR

HateCheck is a suite of functional tests designed to diagnose specific weaknesses in hate speech detection models, addressing limitations of traditional evaluation metrics.

Contribution

The paper introduces HateCheck, a novel set of 29 targeted tests for hate speech models, validated through structured annotation and interviews with stakeholders.

Findings

Revealed critical weaknesses in transformer models

Identified systematic gaps in hate speech datasets

Demonstrated utility of functional testing for diagnostics

Abstract

Detecting online hate is a difficult task that even state-of-the-art models struggle with. Typically, hate speech detection models are evaluated by measuring their performance on held-out test data using metrics such as accuracy and F1 score. However, this approach makes it difficult to identify specific model weak points. It also risks overestimating generalisable model performance due to increasingly well-evidenced systematic gaps and biases in hate speech datasets. To enable more targeted diagnostic insights, we introduce HateCheck, a suite of functional tests for hate speech detection models. We specify 29 model functionalities motivated by a review of previous research and a series of interviews with civil society stakeholders. We craft test cases for each functionality and validate their quality through a structured annotation process. To illustrate HateCheck's utility, we test near-state-of-the-art transformer models as well as two popular commercial models, revealing critical model weaknesses.