In this research article, Katherine Barron, Ryan B. Collis, Aaron J. Richmond, and Ellouise Van Berkel explore disability representation in flash fiction generated by artificial intelligence (AI). The meteoric adoption of generative AI in K–12 education raises concerns about how tools like OpenAI's ChatGPT might perpetuate and amplify biases present in their AI training data. In the educational sphere, there is a vital need for both teachers and students to develop critical literacy skills in order to resist discriminatory narratives about historically marginalized groups. The article aims to identify specific expressions of disability-related discrimination in AI-generated short stories. Using critical content analysis and critical disability theory, the authors analyze forty stories about disabled and neurodivergent children generated by ChatGPT-4. The analysis is guided by Connor's (2017) lists of positive and negative disability representations and Landrum's (2001) criteria for evaluating story elements. The authors first identify how the stories reflect and reinforce societal biases in the context of disability, ableism, and disableism. They then offer the term disability evasiveness to describe a process where nondisabled people claim to not “notice disability.” The article concludes with suggestions and resources for critical literacy instruction in K–12 settings. This research contributes to disability studies in education scholarship and to ongoing discussions of the use of AI in K–12 classrooms.
Barron et al. (Mon,) studied this question.