ResearchPod Summary
Artificial intelligence offers significant potential to improve augmentative and alternative communication (AAC) systems, yet the field currently relies heavily on technical performance metrics like speed (words per minute) and accuracy (word error rate). The authors argue that this approach is insufficient and potentially harmful. By focusing solely on optimizing for a hypothetical 'standard' user, developers risk ignoring the diverse, intersectional needs of disabled people. This narrow focus can lead to 'technoableism,' where systems are designed to 'fix' a disability rather than support the user's agency, identity, and social participation.
The paper proposes that evaluation must be pluralistic, combining technical system metrics with human-centered research methods. The authors analyze six critical design domains—speed and accuracy, physical and mental effort, identity presentation, context adaptation, turn-taking, and fluctuating physical ability—to show how AI can be better integrated and assessed. For example, rather than just measuring how fast a user can type, researchers should evaluate whether an AI-assisted interface allows the user to maintain their desired social presence, manage fatigue, and effectively navigate different communication environments.
Communication is a social act, not just a data entry task. When AAC systems are evaluated only through technical benchmarks, they may fail to support the nuanced ways people express themselves, such as through tone, regional accents, or code-switching. By adopting the authors' proposed framework, researchers and designers can create systems that respect the user's autonomy and better facilitate meaningful interaction with communication partners. This shift is essential for ensuring that AI-powered AAC serves as a tool for empowerment rather than a source of further marginalization.
AI-generated third-party summary by ResearchPod. Not official content or an endorsement by the paper authors or affiliated organizations.