A young adult in a wheel chair using a tablet and listening through headphones. In the spring and summer of 2024, researchers at the American Foundation for the Blind conducted a Delphi study to gather expert opinions about the current and future impact of AI on people with disabilities. A Delphi study is a rigorous research technique for achieving consensus among a group of experts on a specific topic through multiple rounds of questions, using both qualitative and quantitative methods (Dalkey & Helmer, 1963; Falzarano & Zipp, 2013). This study expanded upon the existing literature in at least three ways. First, the expert panel included the voices of professionals working in private industry at companies that develop AI technologies to integrate industry perspectives with those from academia and policy analysis. Second, this study explored intersections between AI and disability across five application domains: transportation, education, employment, healthcare and benefits decision making, and assistive AI. Third, in addition to capturing experts’ views on how AI might both benefit and harm people with disabilities, the study captured recommendations from the experts on specific ways to optimize opportunities and mitigate risks. These recommendations were incorporated into a set of principles for AI developers and policymakers.

A total of 32 experts participated in the study, including 13 people employed in private industry, 9 employed in the nonprofit sector, 7 employed in academia, and 3 employed in the federal government or by a federal contractor. Each expert provided feedback through an interview and two follow-up questionnaires. Following best practices for Delphi studies, the experts participated anonymously, which minimizes bias and promotes honest feedback. The interviews were coded to derive opinion statements, which were shared with the experts in two rounds of questionnaires to identify which opinions were consensually adopted by the panel. Appendix A contains a detailed description of the research study methods and Appendix B contains details on the demographics of participating experts.

In addition to capturing experts’ views on how AI might both benefit and harm people with disabilities, the study captured recommendations from the experts on specific ways to optimize opportunities and mitigate risks.

In the next section, a summary of the expert panel’s consensus opinions and recommendations is presented, along with some opinions upon which the expert panel generated pertinent insights but did not reach consensus. First, specific findings are presented related to AI use in the five focus domains. Then, more general findings are presented regarding concerns about AI’s implications for privacy and discrimination. Finally, a series of proposed solutions and mitigation measures are presented. After presenting findings for the full expert sample, findings will also be summarized from the 13 experts from private industry. Appendix C includes all opinion statements with which the experts agreed, while Appendix D lists the statements with which the experts varied in their level of agreement.

Next Narrative and Findings