Promoting a healthy and comprehensive diet through theory-driven large language models-based agents

Bak, Michelle Chaewon

Permalink

https://hdl.handle.net/2142/132472 Copy

Description

Title

Promoting a healthy and comprehensive diet through theory-driven large language models-based agents

Author(s)

Bak, Michelle Chaewon

Issue Date

2025-10-23

Director of Research (if dissertation) or Advisor (if thesis)

Chin, Jessie

Doctoral Committee Chair(s)

Chin, Jessie

Committee Member(s)

• Wang, Dong
• Brooks, Ian
• Diesner, Jana
• Bhat, Suma

Department of Study

Information Sciences

Discipline

Information Sciences

Degree Granting Institution

University of Illinois Urbana-Champaign

Degree Name

Ph.D.

Degree Level

Dissertation

Keyword(s)

• digital health
• health behavior promotion
• Large Language Models

Abstract

This dissertation investigates the use of theoretical frameworks in health promotion counseling to improve Large Language Models (LLMs) for recognizing and responding to diverse motivational states. The study identifies the limited capabilities of Large Language Models (LLMs) in providing information tailored to the motivational readiness for change among individuals who are resistant or ambivalent to behavior change. Such information gap can be particularly critical given that individuals in the earlier stages of change require different types of support than those in later stages of behavior change, such as information to encourage self-assessment of one’s cognitive and affective state in relation to health behavior. To address this gap, this study integrates behavior change theories, specifically the Transtheoretical Model (TTM) and Motivational Interviewing (MI), into prompt engineering strategies to address the information needs of these individuals in adopting a healthy and comprehensive diet. The improved LLM demonstrate potential in encouraging cognitive and affective self-image assessment in relations to the health behaviors to reduce ambivalence and strengthen commitment to behavior change through targeted, psychologically grounded interactions. The improved LLM also significantly increased behavioral intention without altering knowledge or risk perception, aligned with the characteristics of individuals in the contemplation stage of the TTM. Through MI-based strategies like reflective listening and affirmations, the improved LLM helped participants reduce ambivalence and considering actionable steps to dietary change. The research lays a foundation for LLM-based digital health solutions that support personalized interventions and support the long-term maintenance of health behaviors.

Graduation Semester

2025-12

Type of Resource

Thesis

Handle URL

https://hdl.handle.net/2142/132472

Copyright and License Information

Copyright 2025 Michelle Bak

Owning Collections