mlsanityData to 7 Oct 2026

Image Representations Learned With Unsupervised Pre-Training Contain Human-like Biases

arXiv 2010.15052 ↗First posted 28 Oct 2020cs.CY cs.CVcited by 107

Not on the map. Why?

Abstract (arXiv)

Recent advances in machine learning leverage massive datasets of unlabeled images from the web to learn general-purpose image representations for tasks from image classification to face recognition. But do unsupervised computer vision models automatically learn implicit patterns and embed social biases that could have harmful downstream effects? We develop a novel method for quantifying biased associations between representations of social concepts and attributes in images. We find that state-of-the-art unsupervised models trained on ImageNet, a popular benchmark image dataset curated from internet images, automatically learn racial, gender, and intersectional biases. We replicate 8 previously documented human biases from social psychology, from the innocuous, as with insects and flowers, to the potentially harmful, as with race and gender. Our results closely match three hypotheses about intersectional bias from social psychology. For the first time in unsupervised computer vision, we also quantify implicit human biases about weight, disabilities, and several ethnicities. When compared with statistical patterns in online image datasets, our findings suggest that machine learning models can automatically learn bias from the way people are stereotypically portrayed on the web.

Key facts. Cited by 107 papers in the map (data through 7 Oct 2026). First posted on arXiv 28 Oct 2020.

Builds on

No reference list yet.

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2021: 14; 2022: 28; 2023: 21; 2024: 21; 2025: 16; 2026: 7.

107 papers.

Sort:

By area:

  1. 2022Photorealistic Text-to-Image Diffusion Models with Deep Language Understandingcited by 6,322
  2. 2022LAION-5B: An open large-scale dataset for training next generation image-text modelscited by 4,131
  3. 2022Video Diffusion Modelscited by 2,280
  4. 2021Vector-quantized Image Modeling with Improved VQGANcited by 656
  5. 2023Muse: Text-To-Image Generation via Masked Generative Transformerscited by 640
  6. 2022Safe Latent Diffusion: Mitigating Inappropriate Degeneration in Diffusion Modelscited by 501
  7. 2022DALL-Eval: Probing the Reasoning Skills and Social Biases of Text-to-Image Generation Modelscited by 230
  8. 2022Fake it till you make it: Learning transferable representations from synthetic ImageNet clonescited by 171
  9. 2023Sociotechnical Safety Evaluation of Generative AI Systemscited by 154
  10. 2022Can Machines Help Us Answering Question 16 in Datasheets, and In Turn Reflecting on Inappropriate Content?cited by 150
  11. 2022Deep Spectral Methods: A Surprisingly Strong Baseline for Unsupervised Semantic Segmentation and Localizationcited by 150
  12. 2023Fair Diffusion: Instructing Text-to-Image Generation Models on Fairnesscited by 149
38 more
  1. 2022A Prompt Array Keeps the Bias Away: Debiasing Vision-Language Models with Adversarial Learningcited by 104
  2. 2021Are Gender-Neutral Queries Really Gender-Neutral? Mitigating Gender Bias in Image Searchcited by 90
  3. 2021Extending the WILDS Benchmark for Unsupervised Adaptationcited by 85
  4. 2021A Survey on Bias in Visual Datasetscited by 84
  5. 2021Improved Transformer for High-Resolution GANscited by 59
  6. 2022Use All The Labels: A Hierarchical Multi-Label Contrastive Learning Frameworkcited by 59
  7. 2022Towards Intersectionality in Machine Learning: Including More Identities, Handling Underrepresentation, and Performing Evaluationcited by 59
  8. 2022Normalizing Flows for Human Pose Anomaly Detectioncited by 59
  9. 2023FACET: Fairness in Computer Vision Evaluation Benchmarkcited by 58
  10. 2021PASS: An ImageNet replacement for self-supervised pretraining without humanscited by 51
  11. 2021Mitigating Dataset Harms Requires Stewardship: Lessons from 1000 Paperscited by 49
  12. 2023Inspecting the Geographical Representativeness of Images from Text-to-Image Modelscited by 44
  13. 2022Contrastive Language-Vision AI Models Pretrained on Web-Scraped Multimodal Data Exhibit Sexual Objectification Biascited by 43
  14. 2023A Survey on Intersectional Fairness in Machine Learning: Notions, Mitigation, and Challengescited by 41
  15. 2022American == White in Multimodal Language-and-Image AIcited by 39
  16. 2023T2IAT: Measuring Valence and Stereotypical Biases in Text-to-Image Generationcited by 39
  17. 2021Supervised Transfer Learning at Scale for Medical Imagingcited by 38
  18. 2023Overwriting Pretrained Bias with Finetuning Datacited by 32
  19. 2024Improving Diffusion-Based Image Synthesis with Context Predictioncited by 32
  20. 2021Tractable Density Estimation on Learned Manifolds with Conformal Embedding Flowscited by 30
  21. 2021MMChat: Multi-Modal Chat Dataset on Social Mediacited by 30
  22. 2022Evidence for Hypodescent in Visual Semantic AIcited by 30
  23. 2022Markedness in Visual Semantic AIcited by 29
  24. 2022Diagnosing and Fixing Manifold Overfitting in Deep Generative Modelscited by 28
  25. 2024Identifying Implicit Social Biases in Vision-Language Modelscited by 27
  26. 2021Improving Fractal Pre-trainingcited by 26
  27. 2023Factoring the Matrix of Domination: A Critical Review and Reimagination of Intersectionality in AI Fairnesscited by 26
  28. 2023Survey of Social Bias in Vision-Language Modelscited by 26
  29. 2024Position: Measure Dataset Diversity, Don't Just Claim Itcited by 26
  30. 2022Fairness Indicators for Systematic Assessments of Visual Feature Extractorscited by 24
  31. 2022Gender Bias in Word Embeddings: A Comprehensive Analysis of Frequency, Syntax, and Semanticscited by 24
  32. 2023MultiModal Bias: Introducing a Framework for Stereotypical Bias Assessment beyond Gender and Race in Vision Language Modelscited by 23
  33. 2023Multimodal Composite Association Score: Measuring Gender Bias in Generative Multimodal Modelscited by 21
  34. 2022FairCLIP: Social Bias Elimination based on Attribute Prototype Learning and Representation Neutralizationcited by 20
  35. 2023Mitigating stereotypical biases in text to image generative systemscited by 19
  36. 2023Ethical Considerations for Responsible Data Curationcited by 17
  37. 2024Joint Vision-Language Social Bias Removal for CLIPcited by 17
  38. 2022Emblaze: Illuminating Machine Learning Representations through Interactive Comparison of Embedding Spacescited by 16

50 most cited of 107 All 107 (JSON)