Warning: This is an author submission that is not peer-reviewed or edited. Preprints - unless they show as "accepted" - should not be relied on to guide clinical practice or health-related behavior and should not be reported in news media as established information.
Effects of Response Length, Interaction Mode, and Age on Interaction Experience with a Large Language Model–Powered AI Assistant in Home Care: Mixed Methods Study
ABSTRACT
Background:
Population aging is placing unprecedented demands on home-based health care. Voice assistants and smart speakers are increasingly powered by large language models (LLM) and are becoming candidate solutions for home health care. However, how the response length of an LLM-powered assistant and its interaction mode shape user experience, and how these effects vary with user age, remains rarely understood.
Objective:
This study aimed to examine the effects of response length, interaction mode, and user age on user interaction experience with an AI (artificial intelligence) assistant in a home health care context.
Methods:
We conducted a 3 (response length: short, 10-20 Chinese characters; medium, 30-50 characters; detailed, 50-100 characters) × 2 (interaction mode: voice + text vs. text) mixed design, with response length as a between-subjects factor, interaction mode as a within-subjects factor, and age as a covariate. A total of 162 participants (mean age 42.95, SD 14.25 years; range 21-74 years) interacted remotely with a smart home AI assistant (built on the Baidu Qianfan platform and powered by the DeepSeek-V3.2 LLM). Each participant completed 12 scripted dialogue turns across 6 home health care scenarios and 1 free conversation, completing questionnaires after each interaction mode. Outcomes were comprehension performance, trust, user experience, emotional needs, and cognitive load. Fifteen participants additionally took part in interviews, which were analyzed using content analysis.
Results:
Significant interaction effects between interaction mode and response length were found for user experience (F 2, 158 = 3.7, P = 0.03, η2 = 0.044) and cognitive load (F 2, 158 = 5.26, P = 0.006, η2 = 0.062). Under voice + text interaction, cognitive load was significantly lower for medium responses than for detailed responses (P = 0.03); under text interaction, response length made no significant difference. Response length and interaction mode had no significant influence on comprehension performance, trust, user experience, or emotional needs. Age as a covariate had significant positive effects on comprehension performance (F 1, 158 = 10.4, P = 0.002, η2 = 0.062), trust (F 1, 158 = 12.1, P < .001, η2 = 0.073), and emotional needs (F 1, 158 =11.7, P < .001, η2 = 0.069). Interviews surfaced additional needs concerning response quality, interface design, system response speed, speech recognition accuracy, and ease of operation.
Conclusions:
In home health care contexts, a response length of 30-50 Chinese characters best balances informational completeness and cognitive load when combined with voice + text interaction. Older users reported better comprehension, greater trust, and stronger emotional needs than younger users, indicating that design for different age is required. Design of LLM-powered home health care assistants should jointly consider response length, interaction mode, and user age.
Citation
Request queued. Please wait while the file is being generated. It may take some time.
Copyright
© The authors. All rights reserved. This is a privileged document currently under peer-review/community review (or an accepted/rejected manuscript). Authors have provided JMIR Publications with an exclusive license to publish this preprint on it's website for review and ahead-of-print citation purposes only. While the final peer-reviewed paper may be licensed under a cc-by license on publication, at this stage authors and publisher expressively prohibit redistribution of this draft paper other than for review purposes.