Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

"Sometimes You Need Facts, and Sometimes a Hug": Understanding Older Adults' Preferences for Explanations in LLM-Based Conversational AI Systems

Niharika Mathur, Tamara Zubatiy, Agata Rozga, Jodi Forlizzi, Elizabeth Mynatt

TL;DR

The paper investigates how older adults in aging-in-place contexts prefer explanations from LLM-based conversational AI. Using Speed Dating with storyboard-driven scenarios, it reveals that explanatory needs are highly context-dependent and evolve with emotional state, task risk, and information source. The study shows that environment-derived data and prior conversational history are valued, while numeric confidence scores are often distrusted without observable justification. It argues for HCXAI approaches that support interactive, layered explanations shared with caregivers, and calls for privacy-aware, context-sensitive design to improve trust, actionability, and autonomy for older adults.

Abstract

Designing Conversational AI systems to support older adults requires these systems to explain their behavior in ways that align with older adults' preferences and context. While prior work has emphasized the importance of AI explainability in building user trust, relatively little is known about older adults' requirements and perceptions of AI-generated explanations. To address this gap, we conducted an exploratory Speed Dating study with 23 older adults to understand their responses to contextually grounded AI explanations. Our findings reveal the highly context-dependent nature of explanations, shaped by conversational cues such as the content, tone, and framing of explanation. We also found that explanations are often interpreted as interactive, multi-turn conversational exchanges with the AI, and can be helpful in calibrating urgency, guiding actionability, and providing insights into older adults' daily lives for their family members. We conclude by discussing implications for designing context-sensitive and personalized explanations in Conversational AI systems.

"Sometimes You Need Facts, and Sometimes a Hug": Understanding Older Adults' Preferences for Explanations in LLM-Based Conversational AI Systems

TL;DR

The paper investigates how older adults in aging-in-place contexts prefer explanations from LLM-based conversational AI. Using Speed Dating with storyboard-driven scenarios, it reveals that explanatory needs are highly context-dependent and evolve with emotional state, task risk, and information source. The study shows that environment-derived data and prior conversational history are valued, while numeric confidence scores are often distrusted without observable justification. It argues for HCXAI approaches that support interactive, layered explanations shared with caregivers, and calls for privacy-aware, context-sensitive design to improve trust, actionability, and autonomy for older adults.

Abstract

Designing Conversational AI systems to support older adults requires these systems to explain their behavior in ways that align with older adults' preferences and context. While prior work has emphasized the importance of AI explainability in building user trust, relatively little is known about older adults' requirements and perceptions of AI-generated explanations. To address this gap, we conducted an exploratory Speed Dating study with 23 older adults to understand their responses to contextually grounded AI explanations. Our findings reveal the highly context-dependent nature of explanations, shaped by conversational cues such as the content, tone, and framing of explanation. We also found that explanations are often interpreted as interactive, multi-turn conversational exchanges with the AI, and can be helpful in calibrating urgency, guiding actionability, and providing insights into older adults' daily lives for their family members. We conclude by discussing implications for designing context-sensitive and personalized explanations in Conversational AI systems.

Paper Structure

This paper contains 39 sections, 4 figures, 1 table.

Table of Contents

  1. Introduction
  2. Contributions
  3. Related Work and Background
  4. Older adults’ Use of Conversational AI Systems
  5. Opportunities and Challenges with LLM-generated Explanations
  6. Towards Human-Centered AI Explanations
  7. AI Explanations for Aging-in-Place
  8. Method
  9. Ideation Process
  10. Contexts of Use
  11. Scenario Ideation and Storyboard Development
  12. Participants
  13. Study Procedure
  14. Data Analysis
  15. Results
  16. ...and 24 more sections

Figures (4)

  • Figure 1: The two assistive contexts in the study. Context 1 represents Routine Reminders that are low-risk, recurring, and non-urgent. Context 2 represents Emergency Alerts that are high-risk, one-off, and time-critical.
  • Figure 2: Two examples from our final set of storyboards designed to evoke discussions during the Speed Dating sessions. The two scenarios depict an AI reminding an older adult to stretch during the day in (a), and an alert to check their kitchen for smoke in (b). The two scenarios in (a) and (b) represent the two distinct assistive contexts in the study, i.e., routine reminders and emergency alerts, respectively. In each of the scenarios, the AI, called Rosey, interacts with the older adult and is asked to provide an explanation. The four vignettes after the explanation request represent the four different explanation designs explored in this study, structured using the categorization in mathur2024categorizing.
  • Figure 3: Study Session Flow.
  • Figure 4: Distribution of explanation ratings across the two assistive contexts. These percents (rounded up to the nearest decimal) represent the proportion of positive ratings that each explanation type received. Positive ratings are the total number of instances when an explanation was rated either “strongly agree” or “agree” on the Likert-scale, indicating an overall positive-leaning sentiment.