“What Do Librarians Look Like? Stereotyping of a Profession by Generative AI”

“This study aims to investigate the presence of bias in the visual representation of librarians generated by ChatGPT across three different library settings: school, public, and academic. It focuses on analysing biases related to gender, ethnicity, age, attire, hairstyles and library design in the generated images. The research employed a zero-shot prompting technique to instruct ChatGPT to create visualisations of librarians in the specified settings, either interacting with another librarian or advising a library user. The generated images were then evaluated based on criteria such as positioning, posture, visual cues indicating age and gender and the characteristics of the library environment. The analysis revealed significant biases in the generated images, with a predominant depiction of librarians as Caucasian. Gender representation overstated the presence of men in all libraries, most notably in academic libraries with only 6% of academic librarians depicted as female. Additionally, there was a noticeable trend towards older librarians in public and academic settings, and the size of library buildings increased from school to academic environments. These findings highlight the reinforcement of stereotypes and the misrepresentation of authority dynamics, particularly the portrayal of men in positions of power relative to female colleagues. This study contributes to the growing body of research on biases in generative AI outputs, emphasising the potential dangers of relying on such tools for image generation. It underscores the importance of critically examining AI-generated content to avoid perpetuating discrimination and inequality within the profession of librarianship. The findings serve as a cautionary note regarding the implications of using generative AI for visual representation in professional contexts.”