Emerging Trends in Artificial Intelligence and Blindness: A Delphi Study

Abstract

Introduction: This research article describes anticipated directions of future artificial intelligence (AI) research and development and potential concerns and barriers associated with development as it relates to individuals with disabilities, particularly individuals who are visually impaired (i.e., those who are blind or have low vision). Methodology: A Delphi method was used to identify experts’ opinions and predictions around future AI developments focused around individuals who are blind or visually impaired. A combination of semistructured interviews and web-based surveys were utilized to identify consensus statements regarding the future of AI. Results: In the first round of semistructured interviews, 32 AI experts participated and shared their views on the future of AI and how it will affect individuals with disabilities. From the interviews, 72 common themes and insights were identified. Through 2 web-based surveys, the 72 common themes were narrowed down to 32 statements that reached consensus amongst the expert participants. The statement with the strongest support by the experts was: “A human in the loop is necessary for candidate screening”; which had a rating of 6.76 and a standard deviation of 0.44. The statement with the weakest support was: “AI-driven prenatal screening tools will create actions of eugenics”; which had a rating of 4.69 and a standard deviation of 0.95. Discussion: Experts discussed the careful balance between pushing AI development to support individuals’ lives and the potential harms of development without careful consideration. These findings point to a need for caution in the implementation of AI technologies and tools without proper dedication to fair representation and involvement of individuals with disabilities. Implications for Practitioners: Practitioners should acknowledge and explore the potential benefits that AI can provide for individuals with disabilities but also be aware of the bias and discrimination that is inherent in these tools.

Ähnliche Arbeiten