Autoethnographic Study Evaluates the Effectiveness of AI Tools for Accessibility

by

A recent autoethnographic study conducted by researchers at the University of Washington has examined the potential utility of AI tools for improving accessibility for individuals with disabilities. The study focused on two specific AI tools: ChatGPT, an AI-powered language tool, and Midjourney, an AI-powered image generator. These tools have the potential to assist people with disabilities by summarizing content, composing messages, and describing images.

However, the researchers uncovered significant issues and limitations with these AI tools in most use cases. The tools were found to regularly produce inaccuracies, fail at basic reasoning, and perpetuate ableist biases. Despite some cases where the tools proved helpful, such as generating references for a paper or assisting with writing Slack messages, overall, the researchers found that the AI tools fell short in providing reliable and inclusive accessibility solutions.

The study presented its findings at the ASSETS 2023 conference in New York. The researchers utilized their own experiences as individuals both with and without disabilities to examine the effectiveness of the AI tools. They presented their research through seven vignettes, combining multiple experiences into anonymized accounts. Each vignette highlighted specific challenges and issues faced when using the AI tools.

For example, one vignette featured Mia, who experiences intermittent brain fog. Mia used a tool called ChatPDF.com to summarize PDFs for work. While this tool occasionally provided accurate information, it often produced completely incorrect answers. In one instance, the tool not only provided an incorrect summary but also displayed ableist biases by suggesting that researchers should consult caregivers instead of chronically ill people. Mia, however, was able to recognize the error due to her familiarity with the paper. This highlights one of the insidious problems with using AI, as subtle errors can easily go unnoticed.

On the other hand, Mia found the chatbots helpful in creating and formatting references while experiencing brain fog. Although these AI models still made mistakes, they lightened the cognitive load and improved accessibility.

The study also examined the experiences of an autistic author who used AI tools to write Slack messages at work. While their colleagues found the messages to be robotic, the author felt more confident and efficient in their interactions.

Additionally, three authors attempted to use AI tools to increase the accessibility of content, such as tables for a research paper or a slideshow for a class. While the AI programs could state accessibility rules, they struggled to consistently apply those rules when generating content.

Moreover, an author with aphantasia, an inability to visualize, found image-generating AI tools useful in interpreting imagery from books. However, when attempting to create an illustration of people with disabilities, the AI tool only generated images that included ableist incongruities, such as a disembodied hand resting on a prosthetic leg.

The researchers concluded that further work is required to address the problems and limitations revealed by the study. One particularly complex aspect is developing new ways for people with disabilities to validate the outputs generated by AI tools, as the tools themselves can often create inaccessible content. Additionally, the researchers plan to investigate issues related to ableism and inaccessibility present in AI-generated content, as well as explore problems in other areas like AI-written code.

The results of this study highlight the importance of considering accessibility and inclusive design when developing and deploying AI tools. By addressing the limitations and biases of these tools, developers can ensure that they truly benefit individuals with disabilities and promote inclusivity in technological advancements.

*Note:

  1. Source: Coherent Market Insights, Public sources, Desk research
  2. We have leveraged AI tools to mine information and compile it
Ravina
+ posts

Ravina Pandya,  Content Writer, has a strong foothold in the market research industry. She specializes in writing well-researched articles from different industries, including food and beverages, information and technology, healthcare, chemical and materials, etc. With an MBA in E-commerce, she has an expertise in SEO-optimized content that resonates with industry professionals.