"It feels like hard work trying to talk to it": Understanding Older Adults' Experiences of Encountering and Repairing Conversational Breakdowns with AI Systems

TL;DR

This longitudinal study investigates how older adults experience and repair conversational breakdowns with a home-based AI system (MATCHA) for medication management. By analyzing 844 interactions and 324 breakdowns across seven older adult dyads over 20 weeks, the authors identify four breakdown types (language/semantics, flow, historical understanding, explanations) and map ad hoc repair strategies, highlighting the substantial invisible labor and reliance on environmental resources. The work links these findings to a seamful, sociotechnical perspective, arguing that AI should leverage distributed knowledge from users’ homes and provide context-aware explanations to support more human-centered, collaborative interactions. The results have practical implications for designing AI that better integrates memory, routines, and external information sources, informing both present conversational systems and future generative AI approaches with attention to intent recognition and explainability in real-world aging-in-place contexts.

Abstract

Designing Conversational AI systems to support older adults requires more than usability and reliability, it also necessitates robustness in handling conversational breakdowns. In this study, we investigate how older adults navigate and repair such breakdowns while interacting with a voice-based AI system deployed in their homes for medication management. Through a 20-week in-home deployment with 7 older adult participant dyads, we analyzed 844 recoded interactions to identify conversational breakdowns and user-initiated repair strategies. Through findings gleaned from post-deployment interviews, we reflect on the nature of these breakdowns and older adults' experiences of mitigating them. We identify four types of conversational breakdowns and demonstrate how older adults draw on their situated knowledge and environment to make sense of and recover from these disruptions, highlighting the cognitive effort required in doing so. Our findings emphasize the collaborative nature of interactions in human-AI contexts, and point to the need for AI systems to better align with users' expectations for memory, their routines, and external resources in their environment. We conclude by discussing opportunities for AI systems to integrate contextual knowledge from older adults' sociotechnical environment and to facilitate more meaningful and user-centered interactions.

Figures (1)

• Figure 1: Two-staged structure of the study. Stage 1 (in purple) consisted of user research, system design, field deployment and preliminary evaluation (published in depth in mathur2022collaborative and briefly summarized in this paper) . Stage 2 (in pink) consisted of additional data collection, interviews and analysis related to conversational breakdowns (reported in this paper).