Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

PlanFitting: Personalized Exercise Planning with Large Language Model-driven Conversational Agent

Donghoon Shin, Gary Hsieh, Young-Ho Kim

TL;DR

This work tackles the challenge of creating personalized, actionable exercise plans without ongoing expert involvement. It introduces PlanFitting, an LLM-driven conversational agent that collects user goals, availability, and obstacles through natural dialogue and generates weekly plans grounded in established guidelines via implementation intentions. The system combines retrieval-augmented generation from a curated exercise dataset with a dual-LMM pipeline (dialogue generation and plan synthesis) and a dashboard to track constraints and progress. Evaluation across a formative study, a user study (N=18), and expert assessments shows plans that adhere to guidelines and are perceived as useful and actionable, while highlighting areas for improving balance across exercise types and intensity. The findings offer actionable insights for designing scalable, domain-grounded conversational planning assistants and point to broader applicability in other planning domains and long-term use scenarios.

Abstract

Creating personalized and actionable exercise plans often requires iteration with experts, which can be costly and inaccessible to many individuals. This work explores the capabilities of Large Language Models (LLMs) in addressing these challenges. We present PlanFitting, an LLM-driven conversational agent that assists users in creating and refining personalized weekly exercise plans. By engaging users in free-form conversations, PlanFitting helps elicit users' goals, availabilities, and potential obstacles, and enables individuals to generate personalized exercise plans aligned with established exercise guidelines. Our study -- involving a user study, intrinsic evaluation, and expert evaluation -- demonstrated PlanFitting's ability to guide users to create tailored, actionable, and evidence-based plans. We discuss future design opportunities for LLM-driven conversational agents to create plans that better comply with exercise principles and accommodate personal constraints.

PlanFitting: Personalized Exercise Planning with Large Language Model-driven Conversational Agent

TL;DR

This work tackles the challenge of creating personalized, actionable exercise plans without ongoing expert involvement. It introduces PlanFitting, an LLM-driven conversational agent that collects user goals, availability, and obstacles through natural dialogue and generates weekly plans grounded in established guidelines via implementation intentions. The system combines retrieval-augmented generation from a curated exercise dataset with a dual-LMM pipeline (dialogue generation and plan synthesis) and a dashboard to track constraints and progress. Evaluation across a formative study, a user study (N=18), and expert assessments shows plans that adhere to guidelines and are perceived as useful and actionable, while highlighting areas for improving balance across exercise types and intensity. The findings offer actionable insights for designing scalable, domain-grounded conversational planning assistants and point to broader applicability in other planning domains and long-term use scenarios.

Abstract

Creating personalized and actionable exercise plans often requires iteration with experts, which can be costly and inaccessible to many individuals. This work explores the capabilities of Large Language Models (LLMs) in addressing these challenges. We present PlanFitting, an LLM-driven conversational agent that assists users in creating and refining personalized weekly exercise plans. By engaging users in free-form conversations, PlanFitting helps elicit users' goals, availabilities, and potential obstacles, and enables individuals to generate personalized exercise plans aligned with established exercise guidelines. Our study -- involving a user study, intrinsic evaluation, and expert evaluation -- demonstrated PlanFitting's ability to guide users to create tailored, actionable, and evidence-based plans. We discuss future design opportunities for LLM-driven conversational agents to create plans that better comply with exercise principles and accommodate personal constraints.
Paper Structure (42 sections, 3 figures, 5 tables)

This paper contains 42 sections, 3 figures, 5 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Crafting Personalized and Actionable Exercise Plans
  4. Technology-mediated Exercise Support and Planning
  5. Leveraging LLM-driven Conversation Agents for Personalized Exercise Planning
  6. Formative Study
  7. Practice of Personalized Exercise Planning
  8. Challenges of Personalized Exercise Planning
  9. Difficulty of contextualizing the exercise within their own schedule
  10. Limited availability of planners affecting the iteration process and adaptation to fluctuating schedules
  11. Limited adaptability of planners in engaging with and incorporating client feedback
  12. PlanFitting
  13. Interaction and Conversational UI Design
  14. Collecting exercise-related user constraints
  15. Exercise type recommendation
  16. ...and 27 more sections

Figures (3)

  • Figure 1: Key screen and interaction flow of PlanFitting. ① Once the user describes the goal of the exercise and their own constraints in a natural language on the chat panel, they are parsed and synchronized with the dashboard. ② Based on the collected information, PlanFitting recommends exercises and ③ the user can provide the exercise type(s) they want to include. Once the user finalizes exercise types, ④ the agent returns a weekly exercise plan, where the user can ⑤ continuously iterate on the plan through natural language.
  • Figure 2: Illustration of how the PlanFitting computes and returns the next dialogue of the conversational agent and updates the dashboard based on the current dialogues
  • Figure 3: Sequence of how the participants interacted with the conversational agent to tailor their exercise plan