Disability Representation in AI-Generated Short Stories: Updating Frameworks and Introducing Disability Evasiveness

• In Event: Toward Authentic and Affirming Representation: Disabled and Neurodivergent Perspectives in Life and Literature

Abstract

With the rapid adoption of generative AI in education, concerns have emerged about how Artificial Intelligence (AI) can perpetuate and even amplify biases in their training data. This research aims to identify specific manifestations of disability-related discrimination in AI-generated short stories. Using critical content analysis and critical disability theory, we analyze 40 short stories about disabled and neurodivergent children generated by ChatGPT4. We identify biases in the context of disability, ableism, and disablism. Next, we build on two existing frameworks to address evolving manifestations of disability discrimination. Finally, we introduce the concept of disability-evasiveness to describe a process where non-disabled people claim to not “see disability." This research contributes to ongoing discussions of disability discrimination and ethical use of AI.

Authors

• Ryan Collis, York University

• Katherine Barron, York University

• Ellouise VanBerkel, York University

• Aaron Richmond, York University