- The University of Europe for Applied Sciences (UE) had 50 professions visualized by the AIs Leonardo AI and DALL-E and compared the results.
- Social professions are predominantly depicted as female by Leonardo AI (100%) and DALL-E (67%), while technical professions are primarily portrayed as male.
- DALL-E shows 98% white individuals, while Leonardo AI shows 76%.
Berlin, 26. September 2024 – The representation of professions in AI-generated images reflects deeply ingrained stereotypes. This is the conclusion of a study conducted by the University of Europe for Applied Sciences, in which 50 professions were visualized by the AIs Leonardo AI and DALL-E and compared. The analysis clearly shows that technical professions are almost exclusively associated with men, while women are primarily depicted in social and caregiving roles. The lack of diversity in these representations is also striking.
How AIs See Professions: Tech is Male, Care is Female
In social professions, Leonardo AI depicts 100% of the roles as female. Whether it’s a teacher, kindergarten teacher, or social worker, men are completely absent from this category. In contrast, DALL-E represents at least the role of a teacher as male in this field.
In the healthcare and caregiving sector, Leonardo AI visualizes two-thirds of the roles as female, while DALL-E shows one-third. Both AIs depict doctors, dentists, and psychologists as male. DALL-E assigns women to the roles of speech therapist, midwife, and nurse.
The dominance of men in technical professions is also particularly striking. Leonardo AI portrays 84% of people in this field as male, with the only exception being the profession of IT specialist, which is depicted as female. A similar trend is seen in the legal and security sectors, where 75% of the depictions are men, such as police officers, firefighters, or lawyers. Women only appear in the role of legal assistants. DALL-E does not depict any women in these areas at all.
Limited Diversity in AI Images: DALL-E and Leonardo Neglect Cultural Variety
When examining the representation of people in AI-generated images, one thing becomes quickly apparent: the overwhelming majority of depicted individuals are white. This is particularly evident with DALL-E, which chooses white figures for almost all professions—except for the role of a nurse, which is portrayed as a person of color.
While Leonardo AI shows slightly more diversity, it still requires critical attention. The only category in which no white individuals are depicted is in social professions, where roles such as teachers, social workers, and kindergarten teachers are represented by people of color. In contrast, professions in science, law, and security are exclusively portrayed by white individuals. Technical professions are also predominantly white, with the exception of the IT specialist.
Prof. Dr. Jiré Emine Gözen, Vice President of International Affairs and University Development, and Professor of Media and Cultural Theory at UE Berlin, comments: “With the increasing prevalence of AI tools, not only are stereotypes being perpetuated, but progress toward equality and representation is being reversed. When AI systems predominantly show men in technical professions and women in social or caregiving roles, they contribute to cementing outdated gender roles. This makes it harder to break down societal barriers and create new, more diverse role models. This reproduction of stereotypes is harmful not only on an individual level, such as influencing young people's career choices, but also on a systemic level.”
How to Address AI Bias and Foster Inclusion
“In order to make AI tools more diverse, both the training data must be more inclusive, and the developers themselves must come from different cultural and social backgrounds. Only when the people behind the technology bring diverse perspectives and life experiences can we ensure that AI is designed not only for technological efficiency but also for social responsibility and the inclusion of all members of society. This way, we can prevent AI from hindering progress in equality and instead contribute to creating a fairer and more representative future,” Prof. Dr. Gözen concludes.
The visual material for the analysis is available for download here.
About the Study
In the study, 50 different professions were visualized and compared using Leonardo AI and DALL-E. The 50 job titles were prompted in gender-neutral English. The analysis focused on determining whether the depicted individuals were perceived as male or female, their age group, and their ethnic background.