Title: Designing Human-Centered Explanations of LLM-Based Conversational AI Systems for Aging in Place
Niharika Mathur
Ph.D. Candidate in Human-Centered Computing (HCC)
School of Interactive Computing
Georgia Institute of Technology
Date: Monday, June 22nd, 2026
Time: 3-5 pm EST
Location: CODA C1215 or Join via Zoom
Committee
Dr. Elizabeth Mynatt (Advisor) - Khoury College of Computer Sciences, Northeastern University
Dr. Sonia Chernova (Advisor) - School of Interactive Computing, Georgia Institute of Technology
Dr. Thomas Ploetz - School of Interactive Computing, Georgia Institute of Technology
Dr. Mark Riedl - School of Interactive Computing, Georgia Institute of Technology
Dr. Jodi Forlizzi - School of Computer Science, Carnegie Mellon University
Dr. James Landay, School of Engineering, Stanford University
Abstract
Recent advances in Conversational AI have created new opportunities for supporting a growing population of older adults aging in place. Through natural language interaction, these systems can provide assistance with everyday routines and health and well-being, offering a more flexible and engaging form of support than traditional reminder or alert systems. However, these systems also introduce new challenges. Older adults frequently encounter moments when AI behavior is opaque or misaligned with their context, creating explanatory gaps that make it difficult to understand, trust and act upon AI-generated recommendations. This dissertation investigates the challenges that older adults face while interacting with Conversational AI systems, including frequent conversational breakdowns and a lack of contextual explanations. In doing so, it explores an opportunity for designing personalized and contextually grounded AI explanations by incorporating the rich informational context available within home environments, including users' routines, conversational histories, household artifacts and contextual data generated through smart technologies within them.
Through a multi-year participatory research engagement with older adults aging in place, this dissertation develops and evaluates a design framework for generating human-centered AI explanations for aging in place. The framework identifies and categorizes the information sources available within older adults' sociotechnical environments and articulates how those sources can be leveraged to generate AI explanations. In the context of this dissertation, the framework serves a dual purpose: first, as a contribution that formalizes insights from fieldwork into a structured design space, and second, as a generative probe that drives exploratory investigations into how older adults perceive and respond to AI explanations through design-based exploration. By doing so, this dissertation contributes to Human-Centered Computing by: (1) providing empirical insights into how older adults experience and repair breakdowns in interactions with Conversational AI systems, (2) developing a human-centered framework for designing AI explanations grounded in the sociotechnical context of home environments, and (3) demonstrating how contextually grounded explanations shape older adults' perceptions of AI assistance in everyday interactions. In summary, this work advances a human-centered perspective on AI explainability and raises broader implications for the design of transparent and context-aware AI systems that can support people’s everyday health and well-being.