The experts in this study largely found consensus around measured assessments of the capabilities and harms of AI. As a tool for increasing computing power and expanding the capability of numerous technologies, AI presents both opportunities and risks for people with disabilities. As with image description software and wayfinding, AI may power tools that allow people with disabilities to live, work, and travel more independently. At the same time, AI models, such as those used in employment screening or healthcare decision-making, also run the risk of standardizing individual biases against people with disabilities in a way that is not always transparent to users of the technologies. As a result, this research finds that both developers and deployers of AI technologies should thoughtfully and proactively validate AI models for accurate and fair representations of people with disabilities. In addition, humans will continue to play an important role in setting parameters for, reviewing outputs of, and authorizing final decisions informed by AI tools. AI users will need skills development opportunities that prepare them to use AI in a way that does not discriminate against people with disabilities. At the same time, people with disabilities should be actively trained for and recruited into roles that develop and deploy AI in a wide variety of contexts. Finally, government regulations must protect people with disabilities from harm and balance access to these tools with protections for individual privacy.
AI may power tools that allow people with disabilities to live, work, and travel more independently.
However, this research identifies only the areas in which the participating experts agree. Additional research is needed to understand, with greater clarity, many of the questions that were raised by some of the experts who participated in this study but that were not affirmed by the consensus panel. For example, was the lack of consensus about the benefits of autonomous vehicles due to opinions about the capability of the automated driving system or about other barriers that people with disabilities face in using autonomous vehicles, such as cost, physical vehicle design, or geographic dispersion? In addition, while this research addresses what experts in academia, industry, and disability organizations currently know about the impact of AI on people with disabilities, it does not directly present a more comprehensive understanding of the lived experience of AI users with disabilities in a variety of domains. Further research should elucidate how individuals with disabilities are using AI tools for greater access as well as where they have experienced direct and indirect barriers or harms.
