Disentangling Hate Across Target Identities
Not on the map. Why?
Abstract (arXiv)
Hate speech (HS) classifiers do not perform equally well in detecting hateful expressions towards different target identities. They also demonstrate systematic biases in predicted hatefulness scores. Tapping on two recently proposed functionality test datasets for HS detection, we quantitatively analyze the impact of different factors on HS prediction. Experiments on popular industrial and academic models demonstrate that HS detectors assign a higher hatefulness score merely based on the mention of specific target identities. Besides, models often confuse hatefulness and the polarity of emotions. This result is worrisome as the effort to build HS detectors might harm the vulnerable identity groups we wish to protect: posts expressing anger or disapproval of hate expressions might be flagged as hateful themselves. We also carry out a study inspired by social psychology theory, which reveals that the accuracy of hatefulness prediction correlates strongly with the intensity of the stereotype.
Key facts. No citing papers in the map yet (data through 7 Oct 2026). Builds on 32 of its 49 references that are in the map. First posted on arXiv 14 Oct 2024.
Builds on
32 of 49 references.
References · 32
- 2023GPT-4 Technical Reportcited by 18,132
- 2024The Llama 3 Herd of Modelscited by 14,192
- 2020DeBERTa: Decoding-enhanced BERT with Disentangled Attentioncited by 2,157
- 2023Llama Guard: LLM-based Input-Output Safeguard for Human-AI Conversationscited by 1,124
- 2020Beyond Accuracy: Behavioral Testing of NLP models with CheckListcited by 979
- 2017Automated Hate Speech Detection and the Problem of Offensive Languagecited by 724
- 2023Bias and Fairness in Large Language Models: A Surveycited by 626
- 2019BLiMP: The Benchmark of Linguistic Minimal Pairs for Englishcited by 553
- 2020GoEmotions: A Dataset of Fine-Grained Emotionscited by 479
- 2020Dataset Cartography: Mapping and Diagnosing Datasets with Training Dynamicscited by 423
- 2020HateXplain: A Benchmark Dataset for Explainable Hate Speech Detectioncited by 403
- 2019Social Bias Frames: Reasoning about Social and Power Implications of Languagecited by 377
20 more
- 2019Don't Take the Easy Way Out: Ensemble Based Methods for Avoiding Known Dataset Biasescited by 279
- 2018Large Scale Crowdsourcing and Characterization of Twitter Abusive Behaviorcited by 265
- 2020HateCheck: Functional Tests for Hate Speech Detection Modelscited by 215
- 2019Racial Bias in Hate Speech and Abusive Language Detection Datasetscited by 208
- 2019SemEval-2019 Task 6: Identifying and Categorizing Offensive Language in Social Media (OffensEval)cited by 206
- 2020HateBERT: Retraining BERT for Abusive Language Detection in Englishcited by 206
- 2020Adversarial Filters of Dataset Biasescited by 159
- 2018Reducing Gender Bias in Abusive Language Detectioncited by 152
- 2020Directions in Abusive Language Training Data: Garbage In, Garbage Outcited by 139
- 2019Multilingual and Multi-Aspect Hate Speech Analysiscited by 133
- 2023NLPositionality: Characterizing Design Biases of Datasets and Modelscited by 91
- 2020Hate Speech Detection and Racial Bias Mitigation in Social Media based on BERT modelcited by 87
- 2022Language Generation Models Can Cause Harm: So What Can We Do About It? An Actionable Surveycited by 80
- 2022Multilingual HateCheck: Functional Tests for Multilingual Hate Speech Detection Modelscited by 51
- 2021Understanding and Countering Stereotypes: A Computational Approach to the Stereotype Content Modelcited by 23
- 2020The POLAR Framework: Polar Opposites Enable Interpretability of Pre-Trained Word Embeddingscited by 15
- 2021An Information Retrieval Approach to Building Datasets for Hate Speech Detectioncited by 9
- 2022Analyzing Hate Speech Data along Racial, Gender and Intersectional Axescited by 7
- 2024Hate Speech Detection with Generalizable Target-aware Fairnesscited by 7
- 2023Towards Weakly-Supervised Hate Speech Classification Across Datasetscited by 2
Built on by
Not cited yet.