mlsanityData to 7 Oct 2026

Measuring Representational Harms in Image Captioning

arXiv 2206.07173 ↗First posted 14 Jun 2022cs.CY cs.CVcited by 36

Not on the map. Why?

Abstract (arXiv)

Previous work has largely considered the fairness of image captioning systems through the underspecified lens of "bias." In contrast, we present a set of techniques for measuring five types of representational harms, as well as the resulting measurements obtained for two of the most popular image captioning datasets using a state-of-the-art image captioning system. Our goal was not to audit this image captioning system, but rather to develop normatively grounded measurement techniques, in turn providing an opportunity to reflect on the many challenges involved. We propose multiple measurement techniques for each type of harm. We argue that by doing so, we are better able to capture the multi-faceted nature of each type of harm, in turn improving the (collective) validity of the resulting measurements. Throughout, we discuss the assumptions underlying our measurement approach and point out when they do not hold.

Key facts. Cited by 36 papers in the map (data through 7 Oct 2026). Builds on 35 of its 80 references that are in the map. First posted on arXiv 14 Jun 2022.

Builds on

35 of 80 references.

References · 35

  1. 2014Microsoft COCO: Common Objects in Contextcited by 17,498
  2. 2016Visual Genome: Connecting Language and Vision Using Crowdsourced Dense Image Annotationscited by 2,646
  3. 2014CIDEr: Consensus-based Image Description Evaluationcited by 2,027
  4. 2015Microsoft COCO Captions: Data Collection and Evaluation Servercited by 1,447
  5. 2017Bottom-Up and Top-Down Attention for Image Captioning and Visual Question Answeringcited by 1,310
  6. 2020Oscar: Object-Semantics Aligned Pre-training for Vision-Language Taskscited by 1,213
  7. 2020Language (Technology) is Power: A Critical Survey of "Bias" in NLPcited by 951
  8. 2014Show and Tell: A Neural Image Caption Generatorcited by 877
  9. 2016SPICE: Semantic Propositional Image Caption Evaluationcited by 820
  10. 2021Documenting Large Webtext Corpora: A Case Study on the Colossal Clean Crawled Corpuscited by 481
  11. 2021Multimodal datasets: misogyny, pornography, and malignant stereotypescited by 326
  12. 2019Measurement and Fairnesscited by 299
23 more
  1. 2016Image Captioning with Semantic Attentioncited by 260
  2. 2020Long-Tailed Classification by Keeping the Good and Removing the Bad Momentum Causal Effectcited by 256
  3. 2018Women also Snowboard: Overcoming Bias in Captioning Modelscited by 236
  4. 2016Knowing When to Look: Adaptive Attention via A Visual Sentinel for Image Captioningcited by 231
  5. 2019Towards Fairer Datasets: Filtering and Balancing the Distribution of the People Subtree in the ImageNet Hierarchycited by 156
  6. 2019Does Object Recognition Work for Everyone?cited by 149
  7. 2020REVISE: A Tool for Measuring and Mitigating Bias in Visual Datasetscited by 136
  8. 2019Lessons from Archives: Strategies for Collecting Sociocultural Data in Machine Learningcited by 116
  9. 2018Turning a Blind Eye: Explicit Removal of Biases and Variation from Deep Neural Network Embeddingscited by 114
  10. 2021Understanding and Evaluating Racial Biases in Image Captioningcited by 108
  11. 2017Towards a Visual Privacy Advisor: Understanding and Predicting Privacy Risks in Imagescited by 101
  12. 2019Predictive Inequity in Object Detectioncited by 77
  13. 2018Discriminability objective for training descriptive captionscited by 67
  14. 2017Towards Diverse and Natural Image Descriptions via a Conditional GANcited by 59
  15. 2018Learning to Evaluate Image Captioningcited by 53
  16. 2017Convolutional Image Captioningcited by 52
  17. 2020Mitigating Gender Bias in Captioning Systemscited by 50
  18. 2019Exposing and Correcting the Gender Bias in Image Captioning Datasets and Modelscited by 36
  19. 2016Attention Correctness in Neural Image Captioningcited by 35
  20. 2019Fairness in representation: quantifying stereotyping as a representational harmcited by 26
  21. 2016Stereotyping and Bias in the Flickr30K Datasetcited by 25
  22. 2021Image Scene Graph Generation (SGG) Benchmarkcited by 15
  23. 2015Describing Common Human Visual Actions in Imagescited by 12

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2022: 2; 2023: 11; 2024: 10; 2025: 7; 2026: 6.

36 papers.

Sort:

By area:

  1. 2023AI's Regimes of Representation: A Community-centered Study of Text-to-Image Models in South Asiacited by 54
  2. 2023Uncurated Image-Text Datasets: Shedding Light on Demographic Biascited by 51
  3. 2023Overwriting Pretrained Bias with Finetuning Datacited by 32
  4. 2024Beyond Perplexity: Multi-dimensional Safety Evaluation of LLM Compressioncited by 32
  5. 2022FairGRAPE: Fairness-aware GRAdient Pruning mEthod for Face Attribute Classificationcited by 30
  6. 2023ToViLaG: Your Visual-Language Generative Model is Also An Evildoercited by 27
  7. 2024Identifying Implicit Social Biases in Vision-Language Modelscited by 27
  8. 2024SCoFT: Self-Contrastive Fine-Tuning for Equitable Image Generationcited by 19
  9. 2024Beyond Behaviorist Representational Harms: A Plan for Measurement and Mitigationcited by 17
  10. 2024BendVLM: Test-Time Debiasing of Vision-Language Embeddingscited by 17
  11. 2023Safety and Fairness for Content Moderation in Generative Modelscited by 15
  12. 2023Investigating Practices and Opportunities for Cross-functional Collaboration around AI Fairness in Industry Practicecited by 15
24 more
  1. 2024Harm Amplification in Text-to-Image Modelscited by 14
  2. 2023Nichelle and Nancy: The Influence of Demographic Attributes and Tokenization Length on First Name Biasescited by 12
  3. 2024FairDeDup: Detecting and Mitigating Vision-Language Fairness Disparities in Semantic Dataset Deduplicationcited by 11
  4. 2023Fairness in Visual Clustering: A Novel Transformer Clustering Approachcited by 10
  5. 2025AI Mismatches: Identifying Potential Algorithmic Harms Before AI Developmentcited by 10
  6. 2023VLSlice: Interactive Vision-and-Language Slice Discoverycited by 7
  7. 2023Toward Operationalizing Pipeline-aware ML Fairness: A Research Agenda for Developing Practical Guidelines and Toolscited by 7
  8. 2023Explaining CLIP's performance disparities on data from blind/low vision userscited by 7
  9. 2025Building Trustworthy Multimodal AI: A Review of Fairness, Transparency, and Ethics in Vision-Language Taskscited by 7
  10. 2024Unveiling and Mitigating Bias in Large Language Model Recommendations: A Path to Fairnesscited by 6
  11. 2022A survey of Identification and mitigation of Machine Learning algorithmic biases in Image Analysiscited by 3
  12. 2024Evaluating Model Bias Requires Characterizing its Mistakescited by 3
  13. 2025The Backfiring Effect of Weak AI Safety Regulationcited by 3
  14. 2025Investigating Intersectional Bias in Large Language Models using Confidence Disparities in Coreference Resolutioncited by 3
  15. 2025VIGNETTE: Socially Grounded Bias Evaluation for Vision-Language Modelscited by 2
  16. 2025Are you sure? Measuring models bias in content moderation through uncertaintycited by 2
  17. 2024Common-Sense Bias Modeling for Classification Taskscited by 1
  18. 2025What Is the Point of Equality in Machine Learning Fairness? Beyond Equality of Opportunitycited by 1
  19. 2026Do Image-Text Metrics Respect Semantic Invariances?cited by 1
  20. 2026Prompt Revision as a Source of Cultural Bias in Text-to-Image Systemscited by 1
  21. 2026Your Multimodal Speech Model Says I Have a Face for Radiocited by 0
  22. 2026Beyond wheelchairs and blindfolds: Investigating disability stereotypes in T2I models with INCLUDE-BENCHcited by 0
  23. 2026Guardrail-Agnostic Societal Bias Evaluation in Large Vision-Language Modelscited by 0
  24. 2026Algorithmic Harms Associated with Generative Model-Augmented Recommendation Systemscited by 0