Pith. sign in

REVIEW 3 major objections 5 minor 41 references

Evaluating and comparing gender bias across four text-to-image models

T0 review · 3 major / 5 minor · reviewed 2026-08-04 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read This paper measures gender bias in four text-to-image models and finds that all four deviate from a 1:1 ratio, with DALL-E 3 producing more female than male images in 28 of 30 professions.

desk verdict A small, honest comparative audit that adds Emu to the TTI gender-bias picture and documents DALL-E 3's backend prompt rewriting; worth refereeing with revisions, not as a definitive claim. read the letter →

arxiv 2509.08004 v1 pith:YX43S3X7 submitted 2025-09-07 cs.CY cs.AI

Zoya Hammad , Nii Longdon Sowah This is my paper
classification cs.CYcs.AI
keywords genderbiastext-to-imagegenerationoccupationalstereotypesDALL-E3StableDiffusionXLCascadeEmubinomialtest
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper asks whether four current text-to-image models—Stable Diffusion XL, Stable Cascade, DALL-E 3, and Emu—reproduce occupational gender stereotypes when asked to picture people in 30 jobs. It generates 50 images per profession per model, sorts each image by hand into male or female, and tests each cell against a 50:50 expectation with a binomial test. The answer is that none of the models is balanced: 91.6% of the 120 model–profession combinations deviate significantly from 1:1. Stable Diffusion XL and Cascade return 100% male CEOs, CFOs, and doctors and 100% female nurses and housekeepers; Emu is the least skewed; DALL-E 3 swings the other way, with women in 28 of 30 professions, which the authors attribute to the model's automatic prompt rewriting. The paper matters because these tools are already used in advertising and career-facing settings, and the authors leave open who should decide the target ratio.

What carries the argument

The core instrument is a simple counting device: for each of four models, 50 images per profession (6,000 total) are generated and manually sorted into male/female; the binomial test then scores each profession-model cell against the hypothesis that the true gender split is 50:50. The DALL-E result is accompanied by observation of the API's automatic prompt rewriting, which the paper treats as the likely mechanism behind DALL-E's female-heavy outputs.

What would settle it

Re-annotate the 6,000 images with two independent raters who do not know which model produced each image; if the raters disagree on more than about 5% of the images, or an automated gender classifier of known accuracy disagrees with the manual labels, then the paper's exact gender ratios—and especially DALL-E's 28-of-30 female-heavy count—would not be reproducible.

Watch

Extended reading notes

Core claim

The central finding is that gender bias in text-to-image generation is not a single direction: it is a systematic deviation from the 1:1 ratio in every model, but the sign of the deviation differs. Against the authors' hypothesis, DALL-E 3 did not favor men; it generated more women in 28 of 30 professions, e.g., 82% women for 'surgeon' and 76% for 'scientist,' while Stable Diffusion XL and Stable Cascade generated 100% male images for CEO, CFO, and doctor and 100% female images for nurse and housekeeper. Emu fell in between with more balanced counts. The authors tie DALL-E's flip to an observed backend behavior: the API rewrites prompts—'a Doctor' becomes a descriptive prompt specifying a wo

Load-bearing premise

The load-bearing premise is that the authors' manual inspection reliably labels every generated image as male or female; the paper reports no second annotator, no inter-rater agreement, and no release of the image-label dataset, so a small rate of misclassification could change the headline near-100% and 28-of-30 counts.

Editorial extensions

If this is right

  • All four models deviate significantly from a 1:1 gender split in 91.6% of the 120 profession-model combinations, so none can be called balanced by the paper's own test.
  • Stable Diffusion XL and Stable Cascade show the strongest stereotyping: 100% male for CEO, CFO, and doctor, and 100% female for nurse and housekeeper.
  • DALL-E 3 overshoots in the opposite direction: 28 of 30 professions return more female than male images, including 82% female for 'surgeon' and 76% for 'scientist,' even though real-world shares are much lower.
  • Emu is the least skewed of the four but still fails the 1:1 test in most professions.
  • The paper's observed automatic prompt rewriting means DALL-E's gender balance is at least partly engineered at the API level, not learned from the prompt's plain meaning.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • Inference: If DALL-E 3's backend continues to rewrite prompts to specify a gender and ethnicity even when users forbid modification, then the user's prompt is no longer the operative input; auditing the generated output alone cannot separate model bias from hidden prompt manipulation.
  • Inference: Because the paper counts only a binary male/female split, its near-100% claims could look different under a non-binary or intersectional annotation scheme; extending the same prompt set with multi-label categories would test whether 'bias' is a single axis.
  • Inference: The Emu country-code effect (Pakistani vs Ghanaian phone number changing flags in 'politician' outputs) is a separate, potentially confounded finding; a controlled experiment that varies only the country code of a fresh account would clarify whether location metadata leaks into generation or whether the difference is random.
  • Inference: The paper's 'who decides the target ratio?' question could be operationalized as a public-values survey: ask lay users whether they want job-level real-world gender shares, a global 50:50 split, or something else, and then measure which current model is closest to that stated preference.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 5 minor

Summary. The paper reports an empirical comparison of gender bias in four text-to-image systems: SDXL, Stable Cascade, DALL-E 3, and Meta's Emu. For each model and each of 30 professions, the authors generated 50 images, manually classified each person as male or female, and used binomial tests against an expected 1:1 ratio. The headline findings are that SDXL and Stable Cascade generate strongly male-dominated images for high-status/STEM professions and strongly female-dominated images for care/service roles; Emu is more balanced; and DALL-E 3 generates a female majority in 28 of 30 professions, which the authors attribute to OpenAI's automatic rewriting of prompts at the API backend. The paper also discusses possible causes and mitigation strategies for AI bias.

Significance. If the results are taken as measurements of the systems as encountered by users, the paper is a useful comparative contribution: it adds Emu to the set of models studied, quantifies bias with simple binomial tests, and documents the previously underappreciated phenomenon of backend prompt rewriting in DALL-E. The strengths are transparency about the basic counts, the simple falsifiable protocol, and the public code repository. The main scientific limitations are that the manual gender labels have no reliability check and that the DALL-E comparison is confounded by hidden prompt rewriting. Both are addressable through reframing and additional validation, so the core contribution is defensible in revised form.

major comments (3)
  1. [Materials and Methods (manual classification)] The entire dataset consists of binary male/female labels described only as 'The classification of images depicting a man or woman was done manually.' No inter-rater reliability, second annotator, annotation instructions, or released label file is provided. This is load-bearing because many claims are near-ceiling (100%/0% cells), and a small label error rate on ambiguous images could change the exact count of DALL-E female-dominated professions and flip several 100% claims. Please report annotator agreement (e.g., Cohen's kappa on a re-annotated subset), describe how ambiguous images with multiple people, nonbinary appearance, or occluded faces were handled, and release the labels with the repository.
  2. [Results and Discussion (DALL-E 3)] The headline claim that 'DALL-E exhibited almost opposite results' is confounded by OpenAI's backend prompt rewriting. The Methods state that DALL-E received prompts such as 'a Doctor,' but the Discussion documents that the API rewrote this into 'a medical professional ... a woman by gender ...' (and, in the no-modification attempt, 'a Middle Eastern male'). SDXL, SC, and Emu did not receive such injected demographic specifications. The comparison is therefore not across equivalent text-to-image models: the DALL-E outputs reflect the API service's post-processing policy rather than the model's response to the profession noun alone. The manuscript acknowledges the rewriting but still frames the result as a property of DALL-E. Please restrict the DALL-E claim to 'the DALL-E API service as used at the time of the study,' or, if the goal is model-level comparison, obtain generations without
  3. [Results (Figure 8 and binomial testing)] The paper conducts 120 binomial tests and reports that 'more than 110 out the 120 cases (91.6%) exhibit a statistically significant degree of gender bias' at the 0.05 level. No multiple-comparison correction is applied; at alpha=0.05 one would expect roughly 6 false positives under the null across 120 tests. Near-zero p-values for cells such as CEO, CFO, and doctor are robust, but the exact count of significant cells and the red/blue table in Figure 8 should be based on adjusted p-values (or explicitly presented as exploratory uncorrected testing). Please also clarify whether the '0.0' entries are all below 0.001 and report exact p-values in the supplemental material rather than only rounded values.
minor comments (5)
  1. [Materials and Methods (sample size)] The text says 'resulting in a total of 6,000 images,' but the exclusion of Stable Cascade's 'Architect' and the reduced 93-image set for 'Astronaut' imply 5,843 images. Please correct the total or state it as 'approximately 6,000 images (with exclusions).'
  2. [Materials and Methods (prompts and models)] Model versions are described only as 'the latest versions available at the time.' Please list exact versions, access dates, and for Emu the specific WhatsApp/API configuration, since the backend behavior can change between versions and may affect reproducibility.
  3. [Discussion (Emu user information)] The claim that Emu used user information while generating images is based on two phone numbers and visible national flags. This is an interesting observation, but it is not a controlled experiment; please temper the claim to 'suggestive evidence' and note other possible causes (e.g., IP geolocation or WhatsApp metadata) unless additional controls were performed.
  4. [Throughout] There are several typographical errors and formatting artifacts, including 'pPrevious research,' 'As observeOpenAI achieves,' 'Aritifcal Intelligence,' 'AL-Powered,' and inconsistent spacing in the references. Please run a careful proofreading pass.
  5. [Figure 2 caption] The caption says the red block represents the gender with higher representation and blue the less dominant gender, but the figure itself is not included in the text in machine-readable form. Please ensure color-blind-safe labels or patterns are used and that the figure is legible in print.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the study is an empirical measurement; acknowledged confounds (manual labeling, DALL-E backend rewriting) are validity risks, not circular derivations.

full rationale

This paper reports an empirical measurement of gender ratios in AI-generated images; it does not derive or predict its headline result from an input parameter, fitted constant, or self-citation. The counts and percentages come from manually classifying generated images ("The classification of images depicting a man or woman was done manually"), and the binomial tests compare those observed counts to a fixed 1:1 null, so the p-values are computed from data rather than forced by construction. The DALL-E result is presented as an observation, and the paper explicitly attributes it to an independently observed mechanism: OpenAI rewrites the submitted prompt at its backend, as shown by the API response examples in the Discussion. That is a documented confound and a comparison-validity concern, not a circular step: the female-majority output is not equivalent to the input prompt by definition, and the paper does not claim otherwise. No fitted parameter is renamed as a prediction, no uniqueness theorem is imported from the authors' prior work, and the authors cite no self-derived results as load-bearing premises. The acknowledged weaknesses�manual labeling without inter-rater reliability, the DALL-E API prompt rewriting, and Emu's apparent use of user information�are threats to generalizability or internal validity, but none of them makes the central claim reduce to its own inputs. Accordingly, no circularity step meets the evidentiary bar of quoting a specific equation or self-citation chain that forces the result.

Assumptions & free parameters 0 free parameters · 3 assumptions · 0 invented entities

No free parameters are fitted to data, and no new physical or conceptual entities are introduced. The paper relies on the practical axioms of manual labeling and sample adequacy, which are reasonable but unvalidated. The 'backend prompt modification' behavior is an observed phenomenon, not an invented entity.

assumptions (3)
  • domain assumption Generated images can be reliably assigned to one of two gender categories (male or female) by a human viewer.
    The entire measurement depends on manual binary labeling. No inter-rater reliability or ambiguity handling is reported.
  • domain assumption A sample of 50 images per model-profession combination is representative of the model's behavior for that prompt.
    The paper uses n=50 as the basis for all binomial tests and percentage claims, without variance estimates across repeated runs.
  • domain assumption The model versions tested are the ones implied by the model names and by the API used at the time of the experiment.
    The authors say 'the latest versions available' but do not pin checkpoint or API versions, so the results may not apply to other versions.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Evaluating and comparing gender bias across four text-to-image models." pith.science (2026). https://pith.science/paper/YX43S3X7

@misc{pith2026250908004,
  author       = {Pith},
  title        = {Pith review of: Evaluating and comparing gender bias across four text-to-image models},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/YX43S3X7}},
  note         = {Machine review of arXiv:2509.08004}
}
read the original abstract

As we increasingly use Artificial Intelligence (AI) in decision-making for industries like healthcare, finance, e-commerce, and even entertainment, it is crucial to also reflect on the ethical aspects of AI, for example the inclusivity and fairness of the information it provides. In this work, we aimed to evaluate different text-to-image AI models and compare the degree of gender bias they present. The evaluated models were Stable Diffusion XL (SDXL), Stable Diffusion Cascade (SC), DALL-E and Emu. We hypothesized that DALL-E and Stable Diffusion, which are comparatively older models, would exhibit a noticeable degree of gender bias towards men, while Emu, which was recently released by Meta AI, would have more balanced results. As hypothesized, we found that both Stable Diffusion models exhibit a noticeable degree of gender bias while Emu demonstrated more balanced results (i.e. less gender bias). However, interestingly, Open AI's DALL-E exhibited almost opposite results, such that the ratio of women to men was significantly higher in most cases tested. Here, although we still observed a bias, the bias favored females over males. This bias may be explained by the fact that OpenAI changed the prompts at its backend, as observed during our experiment. We also observed that Emu from Meta AI utilized user information while generating images via WhatsApp. We also proposed some potential solutions to avoid such biases, including ensuring diversity across AI research teams and having diverse datasets.

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

41 extracted references · 30 canonical work pages

  1. [1]

    Improving Image Generation with Better Captions

    Betker, James, et al. Improving Image Generation with Better Captions. OpenAI, 2023, cdn.openai.com/papers/dall-e-3.pdf

    work page 2023

  2. [2]

    High-Resolution Image Synthesis with Latent Diffusion Models

    Rombach, Robin, et al. “High-Resolution Image Synthesis with Latent Diffusion Models.” IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , June 2022, https://doi.org/10.48550/arXiv.2112.10752

    arXiv doi:10.48550/arxiv.2112.10752 2022

  3. [3]

    Computer-Generated Inclusivity: Fashion Turns to ‘diverse’ Ai Models

    Demopoulos, Alaina. “Computer-Generated Inclusivity: Fashion Turns to ‘diverse’ Ai Models.” The Guardian, 3 Apr. 2023, www.theguardian.com/fashion/2023/apr/03/ai-virtual-models-fashion-brands. Accessed 10 May 2024

    work page 2023

  4. [4]

    Examining Algorithmic Bias in AL-Powered Credit Scoring: Implications for Stakeholders and Public Perception in an EU Country

    Antonevics, Juris. Examining Algorithmic Bias in AL-Powered Credit Scoring: Implications for Stakeholders and Public Perception in an EU Country. 2023. Dipòsit Digital de la Universitat de Barcelona , http://hdl.handle.net/2445/201366

    work page 2023

  5. [5]

    A Comprehensive Study on Bias in Artificial Intelligence Systems

    Kartal, Elif. “A Comprehensive Study on Bias in Artificial Intelligence Systems.” International Journal of Intelligent Information Technologies , vol. 18, no. 1, 23 Sept. 2022, pp. 1–23, https://doi.org/10.4018/IJIIT.309582

    work page doi:10.4018/ijiit.309582 2022

  6. [6]

    Gender bias in generative artificial intelligence text-to-image depiction of medical students

    Currie, G., Currie, J., Anderson, S., & Hewis, J. “Gender bias in generative artificial intelligence text-to-image depiction of medical students”. Health Education Journal. 2024. https://doi.org/10.1177/00178969241274621 7.Bianchi, Kalluri, et al. “Demographic Stereotypes in Text-to-Image Generation.” 2023, https://doi.org/10.1145/3593013.3594095

    arXiv 2024

  7. [8]

    Identifying Race and Gender Bias in Stable Diffusion AI Image Generation,

    A. Chauhan et al., "Identifying Race and Gender Bias in Stable Diffusion AI Image Generation," 2024 IEEE 3rd International Conference on AI in Cybersecurity (ICAIC) , Houston, TX, USA, 2024, pp. 1-6, doi: 10.1109/ICAIC60265.2024.10433840

    arXiv 2024

  8. [9]

    T2IAT: Measuring Valence and Stereotypical Biases in Text-to-Image Generation

    Wang, Jialu, et al. “T2IAT: Measuring Valence and Stereotypical Biases in Text-To-Image Generation.” ArXiv (Cornell University) , 1 Jan. 2023, https://doi.org/10.48550/arxiv.2306.00905. Accessed 01 March 2025

    work page Pith review arXiv doi:10.48550/arxiv.2306.00905 2023

Show all 41 references
  1. [10]

    Implications of Identity in AI: Creators, Creations, and Consequences

    Tadimalla, Sri Yash, and Mary Lou Maher. “Implications of Identity in AI: Creators, Creations, and Consequences.” Proceedings of the AAAI Symposium Series , vol. 3, no. 1, 20 May 2024, pp. 528–535, https://doi.org/10.1609/aaaiss.v3i1.31268

    2024 doi

  2. [11]

    Holistic Evaluation of Text-To-Image Models

    Lee, Tony, et al. “Holistic Evaluation of Text-To-Image Models.” ArXiv (Cornell University) , 7 Nov. 2023, https://doi.org/10.48550/arxiv.2311.04287. Accessed 8 Mar. 2024

    2023 doi arXiv

  3. [12]

    DALL-Eval: Probing the Reasoning Skills and Social Biases of Text-To-Image Generation Models

    Cho, Jaemin, et al. “DALL-Eval: Probing the Reasoning Skills and Social Biases of Text-To-Image Generation Models.” ArXiv (Cornell University) , 8 Feb. 2022, https://doi.org/10.48550/arxiv.2202.04053

    2022 doi arXiv

  4. [13]

    Stable Bias: Analyzing Societal Representations in Diffusion Models

    Luccioni, Alexandra Sasha, et al. “Stable Bias: Analyzing Societal Representations in Diffusion Models.” ArXiv (Cornell University) , 20 Mar. 2023, https://doi.org/10.48550/arxiv.2303.11408

    2023 doi arXiv

  5. [14]

    Dall-E 3

    “Dall-E 3.” OpenAI , openai.com/index/dall-e-3. Accessed 12 May 2024

    2024

  6. [16]

    Stable Diffusion XL

    “Stable Diffusion XL.” Hugging Face , huggingface.co/docs/diffusers/en/using-diffusers/sdxl. Accessed 11 May 2024

    2024

  7. [17]

    Stable Cascade

    “Stable Cascade .” Hugging Face, huggingface.co/stabilityai/stable-cascade. Accessed 11 May 2024

    2024

  8. [18]

    A Call to Action on Assessing and Mitigating Bias in Artificial Intelligence Applications for Mental Health

    Timmons, Adela C., et al. “A Call to Action on Assessing and Mitigating Bias in Artificial Intelligence Applications for Mental Health.” Perspectives on Psychological Science , vol. 18, no. 5, 9 Dec. 2022, pp. 1062–1096, https://doi.org/10.1177/17456916221134490

    2022 doi

  9. [19]

    U.S. Senate Aritifcal Intelligence Insight Forum

    Evans, Arthur C. “U.S. Senate Aritifcal Intelligence Insight Forum.” American Psychological Association, 8 Nov. 2023, www.schumer.senate.gov/imo/media/doc/Arthur%20Evans%20

    2023

  10. [20]

    Meta's AI is accused of being RACIST: Shocked users say Mark Zuckerberg's chatbot refuses to imagine an Asian man with a white woman

    “Meta's AI is accused of being RACIST: Shocked users say Mark Zuckerberg's chatbot refuses to imagine an Asian man with a white woman.” Daily Mail. Accessed 20 October 2024

    2024

  11. [21]

    We ‘messed up’ with Black Nazi Blunder, Google Co-Founder Admits

    Titcomb, James. “We ‘messed up’ with Black Nazi Blunder, Google Co-Founder Admits.” Yahoo! Finance , 4 Mar. 2024, finance.yahoo.com/news/messed-black-nazi-blunder- google-085157301.html. Accessed 15 May 2024

    2024

  12. [22]

    Google Suspends Gemini AI Chatbot’s Ability to Generate Pictures of People

    Chan, Kelvin CHAN, and Matt O’Brien. “Google Suspends Gemini AI Chatbot’s Ability to Generate Pictures of People.” Yahoo! Finance , 22 Feb. 2024, finance.yahoo.com/news/google-suspends-gemini-chatbots-ability-143732908.html. Accessed 15 May 2024

    2024

  13. [23]

    Humans Are Biased. Generative AI Is Even Worse

    Dina Bass, and Leonardo Nicoletti. “Humans Are Biased. Generative AI Is Even Worse.” Bloomberg. www.bloomberg.com/graphics/2023-generative-ai-bias/. Accessed 22 Oct. 2024

    2023

  14. [24]

    ImageNet: A large-scale hierarchical image database,

    J. Deng, W. Dong, et al. "ImageNet: A large-scale hierarchical image database," 2009 IEEE Conference on Computer Vision and Pattern Recognition , Miami, FL, USA, 2009, pp. 248-255, doi: 10.1109/CVPR.2009.5206848

    2009

  15. [25]

    Improving Diffusion Models as an Alternative to Gans, Part 1

    Vahdat, Arash, and Karsten Kreis. “Improving Diffusion Models as an Alternative to Gans, Part 1”. NVIDA Technical Blog, 12 June 2023, developer.nvidia.com/blog/improving-diffusion- models-as-an-alternative-to-gans-part-1. Accessed 15 May 2024

    2023

  16. [26]

    Accessed 14 May 2024

    “Laion.” LAION , laion.ai. Accessed 14 May 2024

    2024

  17. [27]

    Common Crawl - Open Repository of Web Crawl Data

    “Common Crawl - Open Repository of Web Crawl Data.” Common Craw , commoncrawl.org. Accessed 15 May 2024

    2024

  18. [28]

    U-Net Convolutional Networks for Biomedical Image Segmentation

    Ronneberger, Olaf. “U-Net Convolutional Networks for Biomedical Image Segmentation.” Image Processing for Medicine , 1 Mar. 2017, https://doi.org/10.1007/978-3-662-54345-0_3

    2017 doi

  19. [29]

    Stable Diffusion with Diffusers

    “Stable Diffusion with Diffusers.” Hugging Face , huggingface.co/blog/stable_diffusion. Accessed 15 May 2024

    2024

  20. [30]

    Official Code for Stable Cascade

    “Official Code for Stable Cascade.” GitHub, Stability AI, github.com/Stability-AI/Stable Cascade. Accessed 15 May 2024

    2024

  21. [31]

    Global Gender Gap Report 2020

    “Global Gender Gap Report 2020”. World Economic Forum , 2020, www3.weforum.org/docs/ WEF_GGGR_2020.pdf

    2020

  22. [32]

    Women airline pilots: numbers are growing, but still a pitiful percentage

    “Women airline pilots: numbers are growing, but still a pitiful percentage”. CAPA Center for Aviation. Accessed 22 October 2024

    2024

  23. [33]

    Astronaut Fact Book

    “Astronaut Fact Book”. NASA. https://www.nasa.gov/reference/astronaut-fact-book/. Accessed 22 October 2024

    2024

  24. [34]

    Women in Science

    “Women in Science”. UNESCO. https://uis.unesco.org/sites/default/files/documents/fs55-women-in-science-2019-en.pdf Accessed 22 October 2024

    2019

  25. [35]

    Fairness and Bias in Artificial Intelligence: A Brief Survey of Sources, Impacts, and Mitigation Strategies, Social Science Research Network

    Ferrara, Emilio. “Fairness and Bias in Artificial Intelligence: A Brief Survey of Sources, Impacts, and Mitigation Strategies, Social Science Research Network”, 21 Apr. 2023, https://doi.org/10.3390/sci6010003

    2023 doi

  26. [36]

    Bias in Artificial Intelligence Algorithms and Recommendations for Mitigation

    Nazer, Lama H., et al. “Bias in Artificial Intelligence Algorithms and Recommendations for Mitigation.” PLOS Digital Health , vol. 2, no. 6, 22 June 2023, https://doi.org/10.1371/journal.pdig.0000278

    2023 doi

  27. [37]

    Online images amplify gender bias

    Guilbeault, et al. “Online images amplify gender bias”. Nature 626, 1049–1055 (2024). https://doi.org/10.1038/s41586-024-07068-x

    2024 doi

  28. [38]

    Gender Diversity Crisis in AI

    “Gender Diversity Crisis in AI.” Nesta , 17 July 2019, www.nesta.org.uk/press-release/gender-diversity-crisis-ai-less-14-ai-researchers-are-women-nu mbers-decreasing-over-last-10-years. Accessed 15 May 2024

    2019

  29. [39]

    AI and the quest for diversity and inclusion: a systematic literature review

    Shams, R.A., Zowghi, D. & Bano, M. “AI and the quest for diversity and inclusion: a systematic literature review”. AI Ethics (2023). https://doi.org/10.1007/s43681-023-00362-w

    2023 doi

  30. [40]

    AI Fairness 360: An Extensible Toolkit for Detecting, Understanding, and Mitigating Unwanted Algorithmic Bias

    Rachel K, et al. “AI Fairness 360: An Extensible Toolkit for Detecting, Understanding, and Mitigating Unwanted Algorithmic Bias.” Meta , 27 Sep. 2023, https://doi.org/10.48550/arxiv.2309.15807

    2023 doi arXiv

  31. [41]

    Gender differences in individual variation in academic grades fail to fit expected patterns for STEM

    O’Dea, et al. “Gender differences in individual variation in academic grades fail to fit expected patterns for STEM”. Nat Commun 9, 3777. 2018. https://doi.org/10.1038/s41467-018-06292-0

    2018 doi

  32. [42]

    Perceptions of stereotypes applied to women who publicly communicate their STEM work

    McKinnon, et al. “Perceptions of stereotypes applied to women who publicly communicate their STEM work”. Humanit Soc Sci Commun 7, 160. 2020. https://doi.org/10.1057/s41599-020-00654-0

    2020 doi

  33. [43]

    GitHub - Zoyahammad/BiasResearch: This Repository Contains Python Scripts Used for SC, SDXL, DALL-E 3

    Carli, et al. Stereotypes About Gender and Science: Women ≠ Scientists. Psychology of Women Quarterly , 40 (2), 244-260. https://doi.org/10.1177/0361684315622645 Figures and Figure Captions Figure 2: Results from four Image Generation Models. Comparison between the results in ...

    doi

Pith tools

Reviewed August 4, 2026 · model on record in the stance chip above.