Aligning the design of voice assistants with older adults’ learning preferences and needs in health-related contexts
Abstract
Older adults often face challenges when seeking health information online due to limited digital literacy and low self-efficacy. While voice assistants have been proposed as supportive tools, little is known about how older adults’ self-efficacy and performance are affected by the use of a voice assistant (vs. a human agent) and by the type of voice assistant in this domain. We conducted a 2x2 between-participants mixed-method study (N = 40) to investigate the effects of assistant type (voice assistant vs. human assistant) and instruction type (mastery experience-based, or do what I say, vs. vicarious experience-based instruction, or listen to what I do) on older adults’ self-efficacy in online health information seeking, general health practice self-efficacy, and the usability and effectiveness of the assistants. The voice assistant-vicarious guidance condition enabled greater perceived ease of use and effectiveness. However, there was no significant effect of the experimental conditions (in terms of main effects or the interaction) on either general or domain-specific self-efficacy. Qualitative interview findings provided nuanced insights, highlighting the perceived clarity and cognitive scaffolding offered by voice assistant-based, vicarious experience-guidance. We discuss implications for designing voice assistants that align with older adults’ learning preferences and needs in health-related contexts.
Authors: Noah Zijie Qu, Jamy J. Li, Mark Chignell
Published in: Computers in Human Behavior Artificial Humans (2026)