Artificial intelligence and machine learning : disruptors and enablers of technology enhancing accessibility

Abstract

Artificial intelligence (AI) and machine learning (ML) are topics of growing interest, especially with the increased popularity of large language models (LLMs) such as ChatGPT. Accessibility is another important area, as an estimated 16% of the world’s population has some form of disability. The intersection of AI, ML and accessibility could create opportunities that might be of benefit to people with disabilities. Therefore, this dissertation investigates accessibility tools that use AI and ML methods, evaluates how these tools may support inclusion and independence, and identifies the main barriers and enablers that affect their adoption. The study followed a mixed methods approach, combining quantitative data with qualitative insights. An online questionnaire was distributed to collect information from the general population about their use and perceptions of AI technology, accessibility features, usefulness, ease of use, trust, cost, privacy and ethics. In total, 118 questionnaire responses were collected. Additionally, 10 interviews were conducted with individuals familiar with AI, including AI professionals, tutors and others with relevant experience. The findings suggest that the majority of respondents had a positive perception of AI and ML technologies, as they viewed these technologies as useful for improving accessibility, usability and independence. The tools and features discussed included chatbots for writing and summarising tasks, speech-to-text, text-to-speech, image descriptions, text extraction from images, live captioning, live translation, voice control, simplified content, smart glasses and wearables. The most common barriers to adoption were cost, followed by privacy, trust, accuracy and reliability, unclear regulation, ethical concerns and limited awareness. The main enablers were affordability, clear usefulness, simplicity, higher awareness, user education and higher levels of trust through transparency. Although the study did not directly include participants with disabilities, it still provides useful insights into how AI accessibility tools are viewed and adopted. Overall, the findings support the research questions and hypotheses, while suggesting that AI accessibility tools should be accurate, affordable, transparent and designed with the involvement of people with disabilities.