Comparing Demographic Representation in AI-Generated and Stock Images of Occupations
DOI:
https://doi.org/10.47611/jsr.v13i3.2628Keywords:
AI, Occupations, DALL·E 3, Getty Images, Race, Gender, Age, Bias, Training DataAbstract
This paper analyzes how AI-generated images from DALL·E 3 and stock images from Getty Images compare in their representation of race, gender, and age across various occupations. The researcher randomly selected ten occupations from the O*NET Database, generating 30 images from DALL·E 3 and selecting 30 images from Getty Images for each occupation. Following this, the researcher categorized each image based on its race, gender, and age. The results showed significant difference in the portrayal of race and gender between the image sources, with DALL·E 3 depicting more white male individuals than Getty Images. However, the researcher found little difference in the representation of age between DALL·E 3 and Getty Images. This study highlights the extent to which AI models like DALL·E 3 might reflect or amplify societal biases present in their training data–and what steps are needed to mitigate this issue.
Downloads
Metrics
References or Bibliography
Bianchi, F., Kalluri, P., Durmus, E., Ladhak, F., Cheng, M., Nozza, D., Hashimoto, T., Jurafsky, D., Zou, J., & Caliskan, A. (2022). Easily Accessible Text-to-Image Generation Amplifies Demographic Stereotypes at Large Scale. Arxiv.org. https://doi.org/10.48550/arXiv.2211.03759
Birhane, A., & Prabhu, V. U. (2021). Large image datasets: A pyrrhic win for computer vision? 2021 IEEE Winter Conference on Applications of Computer Vision (WACV). https://doi.org/10.1109/wacv48630.2021.00158
Braga, A. (2022, March 22). The importance of children’s representation in literature and media. Humanium. https://www.humanium.org/en/the-importance-of-childrens-representation-in-literature-and-media/
Carrera, F. (2020). A raça e o gênero da estética e dos afetos: algoritmização do racismo e do sexismo em bancos contemporâneos de imagens digitais. MATRIZes, 14(2), 217–240. https://doi.org/10.11606/issn.1982-8160.v14i2p217-240
Crawford, K., & Paglen, T. (2019, September 19). Excavating AI: The Politics of Training Sets for Machine Learning. The AI Now Institute, NYU. https://excavating.ai/?itid=lk_inline_enhanced-template
DALL·E now available without waitlist. (2022, September 28). OpenAI. https://openai.com/index/dall-e-now-available-without-waitlist
Garcia, N., Hirota, Y., Wu, Y., & Nakashima, Y. (2023). Uncurated Image-Text Datasets: Shedding Light on
Demographic Bias. https://doi.org/10.1109/cvpr52729.2023.00672
Górska, A. M., & Dariusz Jemielniak. (2023). The invisible women: uncovering gender bias in AI-generated images of professionals. Feminist Media Studies, 1–6. https://doi.org/10.1080/14680777.2023.2263659
Hughes, B., Srivastava, S., Leszko, M., & Condon, D. M. (2022). Occupational Prestige: The Status Component of Socioeconomic Status. PsyArXiv (OSF Preprints). https://doi.org/10.31234/osf.io/6qgxv
Is generative AI a game changer? (2024, February 14). J.P. Morgan. https://www.jpmorgan.com/insights/global-research/artificial-intelligence/generative-ai
Kay, M., Matuszek, C., & Munson, S. A. (2015). Unequal Representation and Gender Stereotypes in Image Search Results for Occupations. Proceedings of the 33rd Annual ACM Conference on Human Factors in Computing Systems - CHI ’15. https://doi.org/10.1145/2702123.2702520
McDuff, D., Cheng, R., & Kapoor, A. (2018, September 30). Identifying Bias in AI using Simulation. ArXiv.org. https://doi.org/10.48550/arXiv.1810.00471
Mortensen, O. (2024, April 24). How Many Users Does ChatGPT Have? Statistics & Facts (2024). Seo.ai. https://seo.ai/blog/how-many-users-does-chatgpt-have
What are AI Models? (2021). Hewlett Packard Enterprise Development LP. https://www.hpe.com/emea_europe/en/what-is/ai-models.html
What is AI? (2023, April 24). McKinsey & Company. https://www.mckinsey.com/featured-insights/mckinsey-explainers/what-is-ai
Zhou, M., Abhishek, V., Derdenger, T., Kim, J., & Srinivasan, K. (2024). Bias in Generative AI. ArXiv (Cornell University). https://doi.org/10.48550/arxiv.2403.02726
Published
How to Cite
Issue
Section
Copyright (c) 2025 Kabir Goklani; Tom Flanagan
This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.
Copyright holder(s) granted JSR a perpetual, non-exclusive license to distriute & display this article.
