Person: An AI Diary: An Experiment at the Intersection of Art, Technology and Embedded Bias in Generative AI Imaging



Author Information

Amy Webster, Independent Scholar, United States

Abstract

Despite aesthetic progress, questions pertaining to the ethics of AI image creation remain open. This project “Person: An AI Diary” is a multi-year experiment, the genesis of which was a single text to image prompt: woman gets a mammogram. This produced surreal stereotypes of women in pink dresses smiling delightedly at always white male doctors. It produced men peeking out of mismatched machinery at their female patients. Results were not of the same quality as other subjects. Compared to contemporaneous prompts of polo players in outer space, heroine in the style of Blade Runner and others the difference was stark. Routine prompts at the time also produced a disproportionate number of white characters and sexualized representations of women. Initiated in 2023, this project measures bias within one top tier text to image AI tool. It educates users by providing examples of safe use guardrails and advocates to reduce biased results by powerfully illustrating the need for diverse engineering teams, diverse users and continuous feedback loops. The experiment hypothesized that results could improve over time through responsible use, curation and feedback. While Year 1 results showed improvement in diverse character creation, there was also an increase in reportable safety issues (nudity, sexual victimization, racial stereotypes, violence) and results remain variable as versions of the tool are updated. This paper summarizes measurable results through May 2025, provides visual examples of both passing & failing results and provides guidance to the user community to consider when using text to image AI tools.


Paper Information

Conference: ECAH2025
Stream: Cyberspace

This paper is part of the ECAH2025 Conference Proceedings (View)
Full Paper
View / Download the full paper in a new tab/window


To cite this article:
Webster A. (2025) Person: An AI Diary: An Experiment at the Intersection of Art, Technology and Embedded Bias in Generative AI Imaging ISSN: 2188-1111 – The European Conference on Arts & Humanities 2025: Official Conference Proceedings (pp. 67-82) https://doi.org/10.22492/issn.2188-1111.2025.8
To link to this article: https://doi.org/10.22492/issn.2188-1111.2025.8


Comments & Feedback

Place a comment using your LinkedIn profile

Comments

Share on activity feed

Powered by WP LinkPress

Share this Research

Posted by James Alexander Gordon