mlsanityData to 14 Oct 2026

When the Majority is Wrong: Modeling Annotator Disagreement for Subjective Tasks

arXiv 2305.06626 ↗First posted 11 May 2023cs.CL cs.AIcited by 90

Not on the map. Why?

Abstract (arXiv)

Though majority vote among annotators is typically used for ground truth labels in natural language processing, annotator disagreement in tasks such as hate speech detection may reflect differences in opinion across groups, not noise. Thus, a crucial problem in hate speech detection is determining whether a statement is offensive to the demographic group that it targets, when that group may constitute a small fraction of the annotator pool. We construct a model that predicts individual annotator ratings on potentially offensive text and combines this information with the predicted target group of the text to model the opinions of target group members. We show gains across a range of metrics, including raising performance over the baseline by 22% at predicting individual annotators' ratings and by 33% at predicting variance among annotators, which provides a metric for model uncertainty downstream. We find that annotator ratings can be predicted using their demographic information and opinions on online content, without the need to track identifying annotator IDs that link each annotator to their ratings. We also find that use of non-invasive survey questions on annotators' online experiences helps to maximize privacy and minimize unnecessary collection of demographic information when predicting annotators' opinions.

Key facts. Cited by 90 papers in the map (data through 14 Oct 2026). Builds on 9 of its 26 references that are in the map. First posted on arXiv 11 May 2023.

Builds on

9 of 26 references.

References · 9

  1. 2019RoBERTa: A Robustly Optimized BERT Pretraining Approachcited by 13,372
  2. 2019Social Bias Frames: Reasoning about Social and Power Implications of Languagecited by 380
  3. 2021Dealing with Disagreements: Looking Beyond the Majority Vote in Subjective Annotationscited by 314
  4. 2021Annotators with Attitudes: How Annotator Beliefs And Identities Bias Toxic Language Detectioncited by 264
  5. 2021On Releasing Annotator-Level Labels and Information in Datasetscited by 126
  6. 2022Jury Learning: Integrating Dissenting Voices into Machine Learning Modelscited by 118
  7. 2021Designing Toxic Content Classification for a Diversity of Perspectivescited by 65
  8. 2023Everyone's Voice Matters: Quantifying Annotation Disagreement Using Demographic Informationcited by 52
  9. 2022Unifying Data Perspectivism and Personalization: An Application to Social Normscited by 22

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2023: 6; 2024: 32; 2025: 26; 2026: 26.

90 papers.

Sort:

By area:

  1. 2024A Roadmap to Pluralistic Alignmentcited by 184
  2. 2024Quantifying the Persona Effect in LLM Simulationscited by 169
  3. 2024Personalizing Reinforcement Learning from Human Feedback with Variational Preference Learningcited by 155
  4. 2023Distributional Preference Learning: Understanding and Accounting for Hidden Context in RLHFcited by 133
  5. 2024Personalized Language Modeling from Personalized Human Feedbackcited by 132
  6. 2023Sensitivity, Performance, Robustness: Deconstructing the Effect of Sociodemographic Promptingcited by 81
  7. 2024Modeling Future Conversation Turns to Teach LLMs to Ask Clarifying Questionscited by 76
  8. 2024Fairness in Large Language Models: A Taxonomic Surveycited by 58
  9. 2024Are Large Language Models Consistent over Value-laden Questions?cited by 58
  10. 2024Can Unconfident LLM Annotations Be Used for Confident Conclusions?cited by 49
  11. 2025Neither Valid nor Reliable? Investigating the Use of LLMs as Judgescited by 39
  12. 2024The Perspectivist Paradigm Shift: Assumptions and Challenges of Capturing Human Labelscited by 38
38 more
  1. 2025Beyond Demographics: Fine-tuning Large Language Models to Predict Individuals' Subjective Text Perceptionscited by 33
  2. 2024WildFeedback: Aligning LLMs With In-situ User Interactions And Feedbackcited by 31
  3. 2023Disentangling Perceptions of Offensiveness: Cultural and Moral Correlatescited by 26
  4. 2024Under the Surface: Tracking the Artifactuality of LLM-Generated Datacited by 24
  5. 2026Rethinking Failure Attribution in Multi-Agent Systems: A Multi-Perspective Benchmark and Evaluationcited by 24
  6. 2024Improving Context-Aware Preference Modeling for Language Modelscited by 22
  7. 2025Game Theory Meets Large Language Models: A Systematic Survey with Taxonomy and New Frontierscited by 16
  8. 2025Hire Your Anthropologist! Rethinking Culture Benchmarks Through an Anthropological Lenscited by 14
  9. 2026Beyond Consensus: Perspectivist Modeling and Evaluation of Annotator Disagreement in NLPcited by 13
  10. 2026Counting on Consensus: Selecting the Right Inter-annotator Agreement Metric for NLP Annotation and Evaluationcited by 13
  11. 2024AdvisorQA: Towards Helpful and Harmless Advice-seeking Question Answering with Collective Intelligencecited by 12
  12. 2025MiCRo: Mixture Modeling and Context-aware Routing for Personalized Preference Learningcited by 11
  13. 2025Modeling Annotator Disagreement with Demographic-Aware Experts and Synthetic Perspectivescited by 10
  14. 2024Subjective textitIsms? On the Danger of Conflating Hate and Offence in Abusive Language Detectioncited by 9
  15. 2024Voices in a Crowd: Searching for Clusters of Unique Perspectivescited by 9
  16. 2024Corpus Considerations for Annotator Modeling and Scalingcited by 8
  17. 2024Re-examining Sexism and Misogyny Classification with Annotator Attitudescited by 8
  18. 2025Capturing Individual Human Preferences with Reward Featurescited by 8
  19. 2023Active teacher selection for reward learningcited by 7
  20. 2024Venire: A Machine Learning-Guided Panel Review System for Community Content Moderationcited by 7
  21. 2024Robustness and Confounders in the Demographic Alignment of LLMs with Human Perceptions of Offensivenesscited by 7
  22. 2025NUTMEG: Separating Signal From Noise in Annotator Disagreementcited by 7
  23. 2024Whose Preferences? Differences in Fairness Preferences and Their Impact on the Fairness of AI Utilizing Human Feedbackcited by 6
  24. 2025HESEIA: A community-based dataset for evaluating social biases in large language models, co-designed in real school settings in Latin Americacited by 6
  25. 2026Tears or Cheers? Benchmarking LLMs via Culturally Elicited Distinct Affective Responsescited by 6
  26. 2024Beyond Accuracy: Investigating Error Types in GPT-4 Responses to USMLE Questionscited by 5
  27. 2023A Taxonomy of Rater Disagreements: Surveying Challenges & Opportunities from the Perspective of Annotating Online Toxicitycited by 4
  28. 2023First Tragedy, then Parse: History Repeats Itself in the New Era of Large Language Modelscited by 4
  29. 2025AfroXLMR-Social: Adapting Pre-trained Language Models for African Languages Social Media Textcited by 4
  30. 2025Perspectives in Play: A Multi-Perspective Approach for More Inclusive NLP Systemscited by 4
  31. 2026Interactive Narrative Analytics: Bridging Computational Narrative Extraction and Human Sensemakingcited by 4
  32. 2024Cross-cultural Inspiration Detection and Analysis in Real and LLM-generated Social Media Datacited by 3
  33. 2024Let Guidelines Guide You: A Prescriptive Guideline-Centered Data Annotation Methodologycited by 3
  34. 2024Accurate and Data-Efficient Toxicity Prediction when Annotators Disagreecited by 3
  35. 2024A Collaborative Content Moderation Framework for Toxicity Detection based on Conformalized Estimates of Annotation Disagreementcited by 3
  36. 2024Beyond Dataset Creation: Critical View of Annotation Variation and Bias Probing of a Dataset for Online Radical Content Detectioncited by 3
  37. 2025Dealing with Annotator Disagreement in Hate Speech Classificationcited by 3
  38. 2025Multimodal Large Language Models Predict Urban Safety Perception but Encode Non-Neutral Demographic Priorscited by 3

50 most cited of 90 All 90 (JSON)