AI for All: Bridging Accessibility and Usability Through User-Centered AI Design

Abstract

Artificial Intelligence (AI) technologies are promised to improve digital services and automate tasks. However, there are still significant barriers to ensuring that AI technologies are accessible and usable by a broad range of users. As AI solutions proliferate across mainstream systems and applications, design-based approaches that explicitly bring in inclusive and human-centric values have become critical. This paper provides a concerted look at user-centered design at the intersection of AI, accessibility, and usability, proposing a framework that cuts across technological, social, and regulatory challenges. Contributions include identifying existing work and current literature gaps, key research questions, and a methodology to explore how to optimize AI systems for the widest possible range of users. We anchor our recommendations with a use-inspired case of an AI-driven public transportation assistant for individuals with diverse physical and cognitive abilities to demonstrate how our framework could benefit real-world applications. On the basis of existing standards and theoretical insights, this paper argues that the design process should be proactive, iterative, and implemented with the participation of multiple stakeholders. In their design of AI systems, this is meant to make the systems adaptive to users, rather than users being adaptive to the AI systems, thus revealing that “AI for all” can indeed be a realistic and realizable paradigm.