Personalized Feedback, Guidance, and Motivation for Older Adults Exercising

Exercise is not one-size-fits-all, and that becomes even more clear over time. What motivates someone on day one may not work a few weeks later. This project explores how a robotic exercise coach can adapt not just to different people, but to the same person as their preferences, performance, and needs evolve.

I develop and deploy a conversational robotic exercise coach that interacts with users during real exercise sessions. The system provides real-time feedback on performance, guidance on movement, and motivational support, all while engaging users in natural conversation. Rather than treating interaction style as fixed, the robot explores different ways of engaging with users.

A core part of this work is understanding how these interaction styles influence both objective outcomes, such as exercise performance, and subjective outcomes, such as motivation, comfort, and engagement. In a longitudinal study, we examine how users respond to different robot personalities over repeated sessions and how their preferences shift with experience.

Building on this, my work moves toward adaptive systems that personalize behavior over time. The goal is to develop a robot that can learn how each individual prefers to be coached, how much feedback they want, and how social or structured the interaction should feel, then adjust accordingly across sessions.

This research is grounded in real-world deployment. By working directly with older adults, we aim to design systems that are not only effective, but also enjoyable, supportive, and sustainable in everyday use.

This work is conducted under the supervision of Dr. Aaron Steinfeld, in collaboration with Roshni Kaushik and Dr. Reid Simmons, at Carnegie Mellon University.

All work is funded through the AI-CARING project.