mlsanityData to 7 Oct 2026

Counterfactually Augmented Data and Unintended Bias: The Case of Sexism and Hate Speech Detection

arXiv 2205.04238 ↗First posted 9 May 2022cs.CLcited by 11

Not on the map. Why?

Abstract (arXiv)

Counterfactually Augmented Data (CAD) aims to improve out-of-domain generalizability, an indicator of model robustness. The improvement is credited with promoting core features of the construct over spurious artifacts that happen to correlate with it. Yet, over-relying on core features may lead to unintended model bias. Especially, construct-driven CAD -- perturbations of core features -- may induce models to ignore the context in which core features are used. Here, we test models for sexism and hate speech detection on challenging data: non-hateful and non-sexist usage of identity and gendered terms. In these hard cases, models trained on CAD, especially construct-driven CAD, show higher false-positive rates than models trained on the original, unperturbed data. Using a diverse set of CAD -- construct-driven and construct-agnostic -- reduces such unintended bias.

Key facts. Cited by 11 papers in the map (data through 7 Oct 2026). Builds on 18 of its 36 references that are in the map. First posted on arXiv 9 May 2022.

Builds on

18 of 36 references.

References · 18

  1. 2018BERT: Pre-training of Deep Bidirectional Transformers for Language Understandingcited by 40,204
  2. 2020Beyond Accuracy: Behavioral Testing of NLP models with CheckListcited by 979
  3. 2020Language (Technology) is Power: A Critical Survey of "Bias" in NLPcited by 951
  4. 2019Learning the Difference that Makes a Difference with Counterfactually-Augmented Datacited by 428
  5. 2019Nuanced Metrics for Measuring Unintended Bias with Real Data for Text Classificationcited by 396
  6. 2021Polyjuice: Generating Counterfactuals for Explaining, Evaluating, and Improving Modelscited by 219
  7. 2020Learning from the Worst: Dynamically Generated Datasets to Improve Online Hate Detectioncited by 217
  8. 2020HateCheck: Functional Tests for Hate Speech Detection Modelscited by 215
  9. 2021Causal Inference in Natural Language Processing: Estimation, Prediction, Interpretation and Beyondcited by 177
  10. 2019Build it Break it Fix it for Dialogue Safety: Robustness from Adversarial Human Attackcited by 176
  11. 2020Contextualizing Hate Speech Classifiers with Post-hoc Explanationcited by 84
  12. 2020Removing Spurious Features can Hurt Accuracy and Affect Groups Disproportionatelycited by 52
6 more
  1. 2019A Just and Comprehensive Strategy for Using NLP to Address Online Abusecited by 49
  2. 2020"Call me sexist, but...": Revisiting Sexism Detection Using Psychological Scales and Adversarial Samplescited by 48
  3. 2021Causal Direction of Data Collection Matters: Implications of Causal and Anticausal Learning for NLPcited by 21
  4. 2021How Does Counterfactually Augmented Data Impact Models for Social Computing Constructs?cited by 19
  5. 2022Fact Checking with Insufficient Evidencecited by 11
  6. 2021Text as Causal Mediators: Research Design for Causal Estimates of Differential Treatment of Social Groups via Language Aspectscited by 3

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2022: 2; 2023: 4; 2024: 4; 2025: 0; 2026: 1.

11 papers.

  1. 2024A Survey on Multilingual Large Language Models: Corpora, Alignment, and Biascited by 82
  2. 2022Perturbation Augmentation for Fairer NLPcited by 51
  3. 2022Probing Classifiers are Unreliable for Concept Removal and Detectioncited by 40
  4. 2024Hate Speech Detection with Generalizable Target-aware Fairnesscited by 7
  5. 2023Hate Speech Detection via Dual Contrastive Learningcited by 3
  6. 2023Unlock the Potential of Counterfactually-Augmented Data in Out-Of-Distribution Generalizationcited by 2
  7. 2024Divided by discipline? A systematic literature review on the quantification of online sexism and misogyny using a semi-automated approachcited by 2
  8. 2023Causal Effect Regularization: Automated Detection and Removal of Spurious Attributescited by 1
  9. 2023On the Challenges of Building Datasets for Hate Speech Detectioncited by 0
  10. 2024NLPGuard: A Framework for Mitigating the Use of Protected Attributes by NLP Classifierscited by 0
  11. 2026Training-Time Explainability for Multilingual Hate Speech Detection: Aligning Model Reasoning with Human Rationalescited by 0