Skip to main content

REHABDATA Detailed Record

An autoethnographic case study of generative artificial intelligence's utility for accessibility. In Proceedings of the 25th International ACM SIGACCESS Conference on Computers and Accessibility.

NARIC Accession Number: O23462. What's this? Download article in Full Text
ISSN:.
Author(s): Glazko, Kate S., Yamagami, Momona, Desai, Aashaka, Mack, Kelly A., Potluri, Venkatesh, Xu, Xuhai, Mankoff, Jennifer.
Project Number: 90ARCP0005.
Publication Year: 2023.
Number of Pages: 8.
Abstract: The authors, a team of seven individuals with and without disabilities, conducted a three-month autoethnography of their use of generative artificial intelligence (GAI) to meet personal and professional needs. During the data collection period, the authors actively looked for opportunities to use GAI tools to either address an access need for a disability or to make their research and teaching materials more accessible. The tools used during this time included Github Copilot, MidJourney, DALL-E 2, ChatGPT, automatic image descriptions, and automated captioning. The authors categorized their experiences into two different types of needs: (1) using GAI to help meet their own access needs and (2) using GAI to help make things more accessible for others. GAI was used to meet various access needs in domains including summarization, communication, image generation, graphical user interface and visualization design, and making documents and visualizations accessible. Currently available tools were used for these tasks, such as ChatGPT and Midjourney, and found that their utility varied significantly and almost always required human verification. The findings demonstrate a wide variety of potential accessibility-related uses for GAI while also highlighting concerns around verifiability, training data, ableism, and false promises. The authors discuss concerns about verifiability and success metrics, the relevance of training data for some tasks, false promises, and the subtle nature of ableism in some of the GAI-generated results.
Descriptor Terms: ACCESSIBILITY, ARTIFICIAL INTELLIGENCE, CASE STUDIES, COMPUTER APPLICATIONS, COMPUTERS, DISABILITIES, WORK PERFORMANCE.


Can this document be ordered through NARIC's document delivery service*?: Y.
Get this Document: https://dl.acm.org/doi/pdf/10.1145/3597638.3614548.

Citation: Glazko, Kate S., Yamagami, Momona, Desai, Aashaka, Mack, Kelly A., Potluri, Venkatesh, Xu, Xuhai, Mankoff, Jennifer. (2023.) An autoethnographic case study of generative artificial intelligence's utility for accessibility. In Proceedings of the 25th International ACM SIGACCESS Conference on Computers and Accessibility Retrieved 5/10/2026, from REHABDATA database.


* The majority of journal articles, books, and reports in our collection are only available by regular mail, rather than downloadable electronic format. Learn more about our digital collection and our document delivery service.