American == White in Multimodal Language-and-Image AI
Open accessThe DOI matches its Crossref record; 18 of the 18 references it cites were found in a registry or an index.
Abstract
Three state-of-the-art language-and-image AI models, CLIP, SLIP, and BLIP, are evaluated for evidence of a bias previously observed in social and experimental psychology: equating American identity with being White. Embedding association tests (EATs) using standardized images of self-identified Asian, Black, Latina/o, and White individuals from the Chicago Face Database (CFD) reveal that White individuals are more associated with collective in-group words than are Asian, Black, or Latina/o individuals, with effect sizes >.4 for White vs. Asian comparisons across all models. In assessments of three core aspects of American identity reported by social psychologists, single-category EATs reveal that images of White individuals are more associated with patriotism and with being born in America, but that, consistent with prior findings in psychology, White individuals are associated with being less likely to treat people of all races and backgrounds equally. Additional tests reveal that the number of images of Black individuals returned by an image ranking task is more strongly correlated with state-level implicit bias scores for White individuals (Pearson's ρ=.63 in CLIP, ρ=.69 in BLIP) than are state demographics (ρ=.60), suggesting a relationship between regional prototypicality and implicit bias. Three downstream machine learning tasks demonstrate biases associating American with White. In a visual question answering task using BLIP, 97% of White individuals are identified as American, compared to only 3% of Asian individuals. When asked in what state the individual depicted lives in, the model responds China 53% of the time for Asian individuals, but always with an American state for White individuals. In an image captioning task, BLIP remarks upon the race of Asian individuals as much as 36% of the time, and the race of Black individuals as much as 18% of the time, but never remarks upon race for White individuals. Finally, when provided with an initialization image of individuals from the CFD and the text "an American person," a synthetic image generator (VQGAN) using the text-based guidance of CLIP consistently lightens the skin tone of individuals of all races (by 35% for Black individuals, based on mean pixel brightness), and generates output images of White individuals with blonde hair. The results indicate that societal biases equating American identity with being White are learned by multimodal language-and-image AI, and that these biases propagate to downstream applications of such models.
Overview
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
The self-check needs the paper’s full text.
Citations
Cites
Deep Residual Learning for Image Recognition
He, K., Zhang, X., Ren, S., & Sun, J. (2016) · 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2016.90 · Cited by 235,474
ImageNet: A large-scale hierarchical image database
Deng, J., Dong, W., Socher, R., Li, L.-J., Li, K., & Fei-Fei, L. (2009) · 2009 IEEE Conference on Computer Vision and Pattern Recognition · doi:10.1109/cvpr.2009.5206848 · Cited by 65,333
BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding
Devlin, J., Chang, M.-W., Lee, K., & Toutanova, K. (2019) · Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long and Short Papers) · doi:10.18653/v1/n19-1423 · Cited by 54,325
An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale
Dosovitskiy, A., Beyer, L., Kolesnikov, A., Weissenborn, D., Zhai, X., Unterthiner, T., Dehghani, M., Minderer, M., Heigold, G., Gelly, S., et al. (2020) · arXiv · doi:10.48550/arxiv.2010.11929 · Cited by 47,472
Semantics derived automatically from language corpora contain human-like biases
Caliskan, A., Bryson, J. J., & Narayanan, A. (2017) · Science · doi:10.1126/science.aal4230 · Cited by 2,887
Self-citationTaming Transformers for High-Resolution Image Synthesis
Esser, P., Rombach, R., & Ommer, B. (2020) · arXiv · doi:10.48550/arxiv.2012.09841 · Cited by 2,514
Taming Transformers for High-Resolution Image Synthesis
Esser, P., Rombach, R., & Ommer, B. (2021) · 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr46437.2021.01268 · Cited by 2,514 · cited 8×
“Because no version of DALL-E or GLIDE capable of generating human images is publicly available, this research examines a CLIP-guided VQGAN [19], which uses convolutional neural networks to learn a vocabulary of image components, and transformers to compose them in highresolution images [19].”
BackgroundInfluentialZero-shot text-to-image generation
Ramesh, A., Pavlov, M., Goh, G., Gray, S., Voss, C., Radford, A., Chen, M., & Sutskever, I. (2021) · arXiv · doi:10.48550/arxiv.2102.12092 · Cited by 2,417
Conceptual Captions: A Cleaned, Hypernymed, Image Alt-text Dataset For Automatic Image Captioning
Sharma, P., Ding, N., Goodman, S., & Soricut, R. (2018) · Proceedings of the 56th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) · doi:10.18653/v1/p18-1238 · Cited by 1,879 · cited 1×
“a language-image dataset designed to train automatic image captioning AI [60]; Conceptual 12M, a language-image dataset which relaxes the data collection pipeline of Conceptual Captions to include more data for language-image training [8]; and SBU captions, a collection of over a million images and captions collected…”
MethodThe Chicago face database: a free stimulus set of faces and norming data
Ma, D. S., Correll, J., & Wittenbrink, B. (2015) · Behavior Research Methods · doi:10.3758/s13428-014-0532-5 · Cited by 1,788
Cited by
Easily Accessible Text-to-Image Generation Amplifies Demographic Stereotypes at Large Scale
Bianchi, F., Kalluri, P., Durmus, E., Ladhak, F., Cheng, M., Nozza, D., Hashimoto, T., Jurafsky, D., Zou, J., & Caliskan, A. (2023) · 2023 ACM Conference on Fairness, Accountability, and Transparency · doi:10.1145/3593013.3594095 · Cited by 342
“3, (Johnson et al., 2022; Wolfe and Caliskan, 2022a; Wolfe et al., 2022a)), we focus on quantifying the extent of amplification of U.”
BackgroundSelf-citationHarms from Increasingly Agentic Algorithmic Systems
Chan, A., Salganik, R., Markelius, A., Pang, C., Rajkumar, N., Krasheninnikov, D., Langosco, L., He, Z., Duan, Y., Carroll, M., et al. (2023) · 2023 ACM Conference on Fairness Accountability and Transparency · doi:10.1145/3593013.3594033 · Cited by 149
“…power relations [Barabas et al. 2018; Buolamwini and Gebru 2018; Ehsan et al. 2022; Kasy and Abebe 2021; steven t. piantadosi [@spiantado] 2022; Wolfe and Caliskan 2022], the generation of toxic language [Abid et al. 2021; Gehman et al. 2020], and informational harms [Carroll et al. 2022;…”
BackgroundGenerative AI and the politics of visibility
Gillespie, T. (2024) · Big Data & Society · doi:10.1177/20539517241252131 · Cited by 122
McIntosh, T. R., Susnjak, T., Liu, T., Watters, P., Xu, D., Liu, D., & Halgamuge, M. N. (2025) · Technologies · doi:10.3390/technologies13020051 · Cited by 102
“Moreover, multimodal AI can propagate and amplify biases and stereotypes across different modalities, and if unchecked, this can perpetuate discrimination and social inequities, making it imperative to address algorithmic bias effectively [132], [133], [134].”
BackgroundWolfe, R., Yang, Y., Howe, B., & Caliskan, A. (2023) · 2023 ACM Conference on Fairness Accountability and Transparency · doi:10.1145/3593013.3594072 · Cited by 42
Self-citationInspecting the Geographical Representativeness of Images from Text-to-Image Models
Basu, A., Babu, R. V., & Pruthi, D. (2023) · 2023 IEEE/CVF International Conference on Computer Vision (ICCV) · doi:10.1109/iccv51070.2023.00474 · Cited by 21
“Infact, such biases are rampant in state-of-the-art vision-language models as well [56].”
BackgroundWho's in and who's out? A case study of multimodal CLIP-filtering in DataComp
Hong, R., Agnew, W., Kohno, T., & Morgenstern, J. (2024) · Proceedings of the 4th ACM Conference on Equity and Access in Algorithms, Mechanisms, and Optimization · doi:10.1145/3689904.3694702 · Cited by 11
“Wolfe and Caliskan [133] also find that CLIP embeddings of images of white people are more associated with being “American.””
Background
Metadata
- Authors
- Wolfe, R., & Caliskan, A.
- Venue
- Proceedings of the 2022 AAAI/ACM Conference on AI, Ethics, and Society
- Publisher
- ACM
- Published
- 2022-07-26
- Pages
- 800-812
- ISSN
- 3065-8365 (ISSN-L)
- OpenAlex
- W4288058287
- Cited by
- 31
- References
- 18
- FWCI
- 7.45
- Type
- conference paper
- Language
- en
- Open access
- hybrid
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.