HateCheck: Functional Tests for Hate Speech Detection Models

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.

arXiv commentAccepted at ACL 2021 (Main Conference)

References in corpus (6)

• Decoupled Weight Decay Regularization
• Directions in Abusive Language Training Data: Garbage In, Garbage Out
• The Hateful Memes Challenge: Detecting Hate Speech in Multimodal Memes
• Tackling Online Abuse: A Survey of Automated Abuse Detection Methods
• Evaluating Models' Local Decision Boundaries via Contrast Sets
• Detecting the Hate Code on Social Media

Cited by in corpus (4)

• Confronting Abusive Language Online: A Survey from the Ethical and Human Rights Perspective
• Anticipating Safety Issues in E2E Conversational AI: Framework and Tooling
• Dynaboard: An Evaluation-As-A-Service Platform for Holistic Next-Generation Benchmarking
• A General Model for Aggregating Annotations Across Simple, Complex, and Multi-Object Annotation Tasks