Document Type

Honors Thesis

Abstract

High levels of sedentary lifestyles can cause adverse effects in individuals’ health. This has prompted researchers to analyze ways to increase physical activity, including the use of Large Language Models (LLMs) to generate motivational messages. While research has found LLMs to be feasible for this task, the findings are limited in availability and scope given that the research focuses on a conversational, chatbot setting—which is not ideal in the real world. This research assesses OpenAI’s GPT-4o mini’s (one of several models powering ChatGPT) ability to tailor messages towards a user. This is done by passing user health data to the LLM and having the LLM generate motivational messages tailored to this data. To ensure the messages are motivational, two behavioral science models are used: the Behavior Change Techniques Taxonomy and Motivational Interviewing. To mitigate hallucination when using these models, the behavioral science models are encapsulated into a knowledge graph using PathRAG. The results indicate that using knowledge graphs and behavioral science models can help LLMs generate motivational messages. However, there are still improvements that need to be made to ensure that the messages are diverse and that the LLM can make use of the whole knowledge graph.

Disciplines

Artificial Intelligence and Robotics

Publication Date

5-2025

Language

English

Faculty Mentor of Honors Project

Chengkai Li

Share

COinS
 
 

To view the content in your browser, please download Adobe Reader or, alternately,
you may Download the file to your hard drive.

NOTE: The latest versions of Adobe Reader do not support viewing PDF files within Firefox on Mac OS and if you are using a modern (Intel) Mac, there is no official plugin for viewing PDF files within the browser window.