Applying the Universal Design Approach to the Design of Artificial Intelligence Systems

As Artificial Intelligence (AI) emerges as a transformative force, opportunities for, and risks to, disabled people emerge alongside it. Important moves taken by the European Union to regulate AI have laid the groundwork to support inclusive and accessible AI, free from bias and respectful of human rights. The Artificial Intelligence (AI) Act requires that 'high risk AI systems' comply with the requirements of the European Accessibility Act (EAA) [1]. The Act also calls for AI systems to be designed universally [2]. In doing so, the AI Act opens the opportunity for organizations designing AI systems to use the upcoming European standard prEN:17161 to meet their obligations and to optimize for accessibility in the design and development of AI systems [3]. However, the application of Universal Design to AI is a newly developing area. Given the novelty and complexity of some AI systems, the application of Universal Design will need to be tailored to the new and emerging approaches. The objective is to ensure that, like other domains, AI-based systems can be used by the widest range of people, irrespective of age, size, ability or disability. This paper will explore the application of the prEN:17161 to the design and development of AI systems, with particular attention to the way the requirements on the inclusion of 'non-users' in the design, testing, and feedback of the product or service are applied.