Problem
Research questions and friction points this paper is trying to address.
Comparing AI and human coaching messages
Evaluating LLM AI's weight loss coaching effectiveness
Assessing feasibility of AI-generated personalized messages
Comparing Large Language Model AI and Human-Generated Coaching Messages for Behavioral Weight Loss
🤖 AI Summary
Prior studies lack empirical evidence on the clinical viability of large language models (LLMs) for behavioral weight-loss interventions in real-world settings.
Method: This study conducted the first randomized controlled trial comparing LLM-generated (GPT-series-based) behavioral guidance with human coach–delivered guidance across efficacy, credibility, and user acceptability. Multimodal evaluation employed validated questionnaires, mixed-effects modeling, qualitative thematic analysis, and a novel interpretable assessment framework.
Contribution/Results: LLM-generated guidance achieved near-human performance in information quality and perceived empathy, while significantly outperforming human guidance in conciseness and actionability. Overall user acceptability reached 82%. These findings demonstrate the clinical feasibility and high acceptability of LLMs in behavioral weight-loss support, providing empirical validation and a methodological paradigm for AI-driven, personalized health coaching.
Innovation
Methods, ideas, or system contributions that make the work stand out.
LLM AI for personalized coaching
ChatGPT mimics human-generated content
AI enhances message helpfulness iteratively
🔎 Similar Papers
No similar papers found.
