Using Tangible Voice Interfaces to improve accessibility of Generative AI for Older Adults

Sammanfattning

This thesis investigates how to improve the accessibility of generative artificial intelligence for older adults, specifically those aged 75 and above. Current large language models predominantly rely on screen-based graphical user interfaces, which often exclude older adults due to high cognitive load and assumptions of digital literacy. To address this, a Research through Design approach was applied to develop a tangible voice interface. A vintage telephone was retrofitted with a microcomputer and conversational artificial intelligence personas, replacing abstract text prompting with physically grounded spoken interaction. Through a comparative qualitative study using a Wizard of Oz setup, four older adults interacted with both a standard text-based chatbot and the tangible telephone handset prototype across four distinct conversational personas. The findings demonstrate that the tangible voice interface substantially reduced cognitive load and visual information overload. By relying on familiar embodied routines and muscle memory, the physical telephone handset created a more intuitive and emotionally approachable usability experience. However, the research also reveals that trust and conversational flow in voice interfaces remain highly dependent on accurately calibrated vocal tone, pacing, and emotional appropriateness. Furthermore, the study examines critical ethical challenges surrounding conversational artificial intelligence for older adults, specifically highlighting the risks of emotional dependency, anthropomorphic attachment, and the complexities of trust calibration in human like interactions. Ultimately, this project highlights that making generative artificial intelligence accessible is not solely a technical problem, but an embodied, emotional, and ethical design challenge.