Natural Language Processing for Behavior Change and Health Improvement

Maladaptive health behaviors contribute to lifestyle-related diseases such as obesity and type 2 diabetes. One key factor driving these behaviors is delay discounting---the tendency to prioritize immediate rewards over delayed ones. Episodic Future Thinking (EFT) is an intervention designed to reduce delay discounting by engaging individuals in vivid mental simulations of future events, thereby influencing decision-making and emotional well-being. Participants generate descriptions of personally significant future events. Research studies have shown EFT's effectiveness in promoting healthier behaviors, including improved exercise and medication adherence. However, the mechanisms underlying EFT and the factors influencing its efficacy remain unclear. With advancements in machine learning (ML) and natural language processing (NLP), new techniques can enhance EFT analysis and user experience. This study investigates EFT cue texts to identify characteristics that make them effective, and explores how a chatbot can assist users in generating impactful cues. We use language models and fine-tune pre-trained models to classify EFT cue texts, facilitating further analysis of their features. Additionally, we leverage instruction-tuned Large Language Models (LLMs) to classify cue texts to address annotation variability. We explore zero-shot and few-shot prompt tuning, demonstrating that a small number of high-quality labeled samples significantly improves classification performance. We present the design, implementation, and evaluation of an AI-powered chatbot using the GPT-4 LLM that generates EFT cue texts for individuals with lifestyle-related conditions. We evaluated the chatbot using both quantitative and qualitative approaches. This included automated assessment by prompting language models as well as user studies incorporating usability assessments and qualitative evaluations. These methods demonstrated the chatbot's effectiveness in generating personalized EFT cues.

Name thesis:degree_name

Doctor of Philosophy

Level thesis:degree_level

doctoral

Discipline thesis:degree_discipline

Computer Science & Applications

Department dc:contributor.department

Computer Science and#38; Applications

Grantor dc:publisher

Virginia Tech

Year dc:date.issued

2025

Author dc:creator

• Ahmadi, Sareh

Chair dc:contributor.committeechair

• Fox, Edward A.

Committee members dc:contributor.committeemember

• Bhattacharya, Debswapna
• Wang, Xuan
• Ramakrishnan, Narendran
• Manikonda, Lydia

• Natural Language Processing
• Large Language Models
• AI ChatBot
• Episodic Future Thinking (EFT)

Statement dc:rights

• In Copyright

Licence dc:rights.uri

• http://rightsstatements.org/vocab/InC/1.0/

Language dc:language.iso

en

Dc Identifier Other

vt_gsexam:44516

OAI identifier oai:identifier

oai:vtechworks.lib.vt.edu:10919/137741