From Barriers to Tactics: A Behavioral Science-Informed Agentic Workflow for Personalized Nutrition Coaching

Eric Yang,Tomas Garcia,Hannah Williams,Bhawesh Kumar,Martin Ramé,Eileen Rivera,Yiran Ma,Jonathan Amar,Caricia Catalani,Yugang Jia

2024-10-18

Abstract:Effective management of cardiometabolic conditions requires sustained positive nutrition habits, often hindered by complex and individualized barriers. Direct human management is simply not scalable, while previous attempts aimed at automating nutrition coaching lack the personalization needed to address these diverse challenges. This paper introduces a novel LLM-powered agentic workflow designed to provide personalized nutrition coaching by directly targeting and mitigating patient-specific barriers. Grounded in behavioral science principles, the workflow leverages a comprehensive mapping of nutrition-related barriers to corresponding evidence-based strategies. A specialized LLM agent intentionally probes for and identifies the root cause of a patient's dietary struggles. Subsequently, a separate LLM agent delivers tailored tactics designed to overcome those specific barriers with patient context. We designed and validated our approach through a user study with individuals with cardiometabolic conditions, demonstrating the system's ability to accurately identify barriers and provide personalized guidance. Furthermore, we conducted a large-scale simulation study, grounding on real patient vignettes and expert-validated metrics, to evaluate the system's performance across a wide range of scenarios. Our findings demonstrate the potential of this LLM-powered agentic workflow to improve nutrition coaching by providing personalized, scalable, and behaviorally-informed interventions.

Machine Learning,Computation and Language

What problem does this paper attempt to address?

The main problem that this paper attempts to solve is: **How to effectively manage cardiometabolic diseases (such as diabetes and hypertension) through personalized nutrition guidance and overcome individualization barriers to achieve continuous healthy eating habits**.

Specifically, the research aims to:

1. **Identify and address personalized barriers**: Traditional nutrition guidance usually fails to provide personalized solutions for each patient's unique barriers. This paper proposes an intelligent workflow based on large - language models (LLM) that can identify and alleviate patient - specific dietary barriers.

2. **Provide personalized interventions supported by behavioral science**: By integrating the research results of behavioral science, especially frameworks such as the Capability - Opportunity - Motivation - Behavior (COM - B) model and the Behavior Change Wheel (BCW), this workflow can provide patients with targeted behavior - change strategies.

3. **Achieve a balance between scalability and personalization**: Direct human management is difficult to promote on a large scale, and previous automated nutrition guidance lacks sufficient personalization. This research achieves the combination of personalization and scalability through a multi - agent dialogue system.

### Main contributions

1. **Comprehensive barrier identification and strategy mapping**: Through user research and literature review, common nutrition - related barriers are identified and mapped to a series of effective behavioral science strategies and tactics, thus providing more comprehensive and detailed solutions.

2. **Multi - agent workflow design**: A multi - agent workflow composed of two core agents is designed. One is used to identify barriers, and the other is used to execute strategies, which improves the effectiveness of personalized nutrition guidance.

3. **Real - world verification**: The practicality and effectiveness of the system are verified by interacting with people actually suffering from cardiometabolic diseases.

4. **Benchmark testing and expert evaluation**: Detailed evaluation criteria are proposed, including barrier identification, strategy execution, and overall dialogue quality, and strict evaluation is carried out through expert annotation.

5. **Large - scale simulation evaluation**: By generating a large number of simulated dialogues, the performance of the system in different scenarios is evaluated to ensure its wide applicability and performance stability.

Through these methods, this research shows the potential of the LLM - based intelligent workflow in personalized nutrition guidance, especially in the management of cardiometabolic diseases, and can provide more effective and sustainable behavior - change support.