Gender Biases and Generative AI Models
Approaches Based on Qualitative Analysis of Image Generation
DOI:
https://doi.org/10.35362/issn.1850-0013-1258Keywords:
generative AI, gender biases, language models, stereotypes, photorealistic imagesAbstract
This article examines the extent of gender bias present in large generative AI language models (GenAI) that incorporate image generation, as well as the challenges involved in studying this issue from a Latin American perspective. It presents a review of the specialized literature from the past decade, primarily using quantitative methodological approaches, which highlights the relationship between these biases and training data, as well as the development of biased algorithmic patterns. Based on this, a qualitative experimental exercise is proposed to identify gender stereotypes in the generation of visual images of women scientists, women scientists in the social sciences, and women scientists in the Latin American social sciences. The qualitative experimental methodology utilizes two recently updated models, ChatGPT 5.5 and Gemini Flash 3.5, in their image generation functions. Finally, the results obtained are discussed, analyzed from a gender perspective, and a series of challenges are outlined for further research on visual gender stereotypes and biases from a feminist and Latin American perspective.
Downloads
References
Atkinson-Abutridy, John (2023). Grandes Modelos de Lenguaje: conceptos, técnicas y aplicaciones. Barcelona: Editorial Marcombo.
Balmaceda, Tomás (2024). IA generativa y sus disrupciones en López, Consuelo et al. OK Pandora. Seis ensayos sobre Inteligencia Artificial. Buenos Aires: Editorial La Caja y el Gato.
Bender, Emily, Gebru, Timnit, McMillan-Major, Angelina & Shmitchell, Shmargaret (2021). On the dangers of stochastic parrots: Can language models be too big? Proceedings of the 2021 ACM Conference on Fairness, Accountability, and Transparency (610-623). Recuperado de: https://dl.acm.org/doi/abs/10.1145/3442188.3445922.
Buie, Hannah & Croft, Alyssa (2023). The Social Media Sexist Content (SMSC) database: A database of content and comments for research use. Collabra: Psychology, 9(1), 7134. DOI: https://doi.org/10.1525/collabra.71341.
Buolamwini, Joy & Gebru, Timnit (2018). Gender shades: Intersectional accuracy disparities in commercial gender classification. Proceedings of Machine Learning Research, (81), 1-15. Recuperado de: https://proceedings.mlr.press/v81/buolamwini18a/buolamwini18a.pdf.
Busker, Tony, Choenni, Sunil & Bargh, Mortaza S. (2023). Stereotypes in ChatGPT: An empirical study. 16th International Conference on Theory and Practice of Electronic Governance (ICEGOV 2023), 1-9. DOI: https://doi.org/10.1145/3614321.3614325.
Doria, Vanesa, Korzeniewski, María, Flores, Carola & del Prado, Ana M. (2023). Herramientas y tips para generar prompts con Inteligencia Artificial en Repositorio Institucional de Acceso Abierto. Catamarca: Universidad Nacional de Catamarca. Recuperado de: https://riaa-tecno.unca.edu.ar/handle/123456789/923.
Eichler, Magrit (2001). Moving forward: Measuring gender bias and more en Gender Based Analysis in Public Health Research Policy and Practice. Documentation of the International Workshop in Berlin.
Guilbeault, Douglas, Delecourt, Solène, Hull, Tasker, Desikan, Bhargav, Chu, Mark & Nadler, Ethan (2024). Online images amplify gender bias. Nature, 626(8001), 1049-1055. DOI: https://doi.org/10.1038/s41586-024-07068-x.
Hernández-Sampieri, Roberto, Fernández-Collado, Carlo & Baptista-Lucio, Pilar (2014). Metodología de la investigación, Sexta edición. México: McGraw Hill / Interamericana de Ediciones.
Kotek, Hadas, Dockum, Rikker & Sun, David (2023). Gender bias and stereotypes in Large Language Models. Proceedings of the 2023 ACM Conference on Fairness, Accountability, and Transparency (12-24). DOI: https://doi.org/10.1145/3582269.3615599.
Liu, Yian, Elekes, Ákos, Lu, Junjie, Dorantes-Gilardi, Rodrigo & Barabási, Albert-László (2025). Unequal scientific recognition in the age of LLMs. Findings of the 2025 Conference on Empirical Methods in Natural Language Processing, 1. Recuperado de: https://aclanthology.org/2025.findings-emnlp.1279.pdf.
Holmes, Walter & Miao, Fengchun (2024). Guía para el uso de IA generativa en educación e investigación. UNESCO Publishing. Recuperado de: https://unesdoc.unesco.org/ark:/48223/pf0000389227.
Mohammadi, Ehsan, Thelwall, Mike, Cai, Yizhou, Collier, Taylor, Tahamtan Iman & Eftekhar, Azar (2026). Is generative AI reshaping academic practices worldwide? A survey of adoption, benefits, and concerns. Information Processing & Management, 63(1), 104350. DOI: https://doi.org/10.1016/j.ipm.2025.104350.
Perdomo Reyes, Inmaculada (2024). Injusticia epistémica y reproducción de sesgos de género en la inteligencia artificial. Revista Iberoamericana de Ciencia, Tecnología y Sociedad, 19(56), 89-100. DOI: https://doi.org/10.52712/issn.1850-0013-555.
Pérez-Ugena Coromina, María (2024). Sesgo de género (en IA). Eunomía. Revista en Cultura de la Legalidad, (26), 311-330. DOI: https://doi.org/10.20318/eunomia.2024.8515.
Robles, Melissa, Bernal, Catalina, Raigoso, Denniss & Dulce Rubio, Mateo (2025), SESGO: Spanish Evaluation of Stereotypical Generative Outputs. DOI: https://doi.org/10.48550/arXiv.2509.03329.
Ricaurte, Paola & Zasso, Mariel (2022). Inteligencia Artificial Feminista: hacia una agenda de investigación en América Latina y el Caribe. Costa Rica: Editorial Tecnológica de Costa Rica. Recuperado de: https://feministai.pubpub.org/lachub-libro.
Rodríguez-Sánchez, Francisco, Carrillo de Albornoz, Jorge, Plaza, Laura, Gonzalo, Julio., Rosso, Paolo, Comet, Miriam & Donoso, Trinidad (2021). Overview de EXIST 2021: Identificación de sexismo en redes sociales. Procesamiento del Lenguaje Natural, (67), 195-202. DOI: https://doi.org/10.26342/2021-67-17.
Rombach, Robin, Blatmann, Andreas, Lorenz, Dominic, Esser, Patrick & Ommer, Björn (2022). [PrePrint]. DOI: https://doi.org/10.48550/arXiv.2112.10752.
Sánchez-Prieto, José, Izquierdo-Álvarez, Vanessa, del Moral-Marcos, María Teresa & Martínez-Abad, Fernando (2025). Inteligencia artificial generativa para autoaprendizaje en educación superior: Diseño y validación de una máquina de ejemplos. RIED - Revista Iberoamericana de Educación a Distancia, 28(1), 59-81. DOI: https://doi.org/10.5944/ried.28.1.41548.
Umoja Noble, Safiya (2018). Algorithms of oppression: How search engines reinforce racism. NYU Press. DOI: https://doi.org/10.2307/j.ctt1pwt9w5.
Unesco & IRCAI (2024). Challenging systematic prejudices: An investigation into gender bias in Large Language Models. UNESCO Publishing. Recuperado de: https://unesdoc.unesco.org/ark:/48223/pf0000388971.
Van Blerck, Irene, Soares de Lima, Edirlei, Neggers, Margot M.E. & Calders, Toon (2025). Unveiling gender bias in LLM-generated hero and heroine narratives. Entertainment Computing, 55, 100972. DOI: https://doi.org/10.1016/j.entcom.2025.100972.
Vázquez Recio, Rosa (2014). Investigación, género y ética: una triada necesaria para el cambio. Forum: Qualitative Social Research, 15(2). Recuperado de: https://www.researchgate.net/publication/265013227_Investigacion_genero_y_etica_una_triada_necesaria_para_el_cambio.
Wasielewski, Amanda (2024). Unnatural images: On AI-generated photographs. Critical Inquiry, 51(1), 1-29. DOI: https://doi.org/10.1086/731729.
Downloads
Published
How to Cite
Issue
Section
License
Copyright (c) 2026 CC Attribution 4.0
This work is licensed under a Creative Commons Attribution 4.0 International License.
All CTS's issues and academic articles are under a CC-BY license.
Since 2007, CTS has provided open and free access to all its contents, including the complete archive of its quarterly edition and the different products presented in its electronic platform. This decision is based on the belief that offering free access to published materials helps to build a greater and better exchange of knowledge.
In turn, for the quarterly edition, CTS allows institutional and thematic repositories, as well as personal web pages, to self-archive articles in their post-print or editorial version, immediately after the publication of the final version of each issue and under the condition that a link to the original source will be incorporated into the self-archive.
