Ableism in AI generated images
Fliscu, Ana Delia (2026)
Kandidaatintyö
2026
School of Engineering Science, Tietotekniikka
Kaikki oikeudet pidätetään.
Julkaisun pysyvä osoite on
https://urn.fi/URN:NBN:fi-fe20260904122805
https://urn.fi/URN:NBN:fi-fe20260904122805
Tiivistelmä
Since generative AI tools have become more popular in recent years, many users have noticed that generated images lack diversity. Despite the topic of accurate representation has been discussed in terms of age, gender, and race, there is a limited amount of research regarding disability.
This paper showcases how older and newer versions of generative AI platforms portray disabled people and how ableism is present even in generated images.
One of the main objectives was to present the complexity of mental health disorders and neurodivergence, and that AI tools are unable to depict those in a thoughtful manner without reinforcing stereotypes. Another important aspect discussed in this thesis was how gen AI represents better physical disabilities compared to autism, ADHD, and OCD.
Overall, new AI chatbots are improved in terms of diversity, however the issue of ableism remains. Even if the people depicted in the generated images appear to be from various demographics, the representation of disabilities needs development. The proposal of this paper is to ensure that disabled people are heard and professionals in disability studies are part of the evaluation process of AI tools, to create a more accurate illustration of disability.
This paper showcases how older and newer versions of generative AI platforms portray disabled people and how ableism is present even in generated images.
One of the main objectives was to present the complexity of mental health disorders and neurodivergence, and that AI tools are unable to depict those in a thoughtful manner without reinforcing stereotypes. Another important aspect discussed in this thesis was how gen AI represents better physical disabilities compared to autism, ADHD, and OCD.
Overall, new AI chatbots are improved in terms of diversity, however the issue of ableism remains. Even if the people depicted in the generated images appear to be from various demographics, the representation of disabilities needs development. The proposal of this paper is to ensure that disabled people are heard and professionals in disability studies are part of the evaluation process of AI tools, to create a more accurate illustration of disability.