# Exploring the Dynamics of Dietary Self-Monitoring Adherence Among Participants in a Digital Behavioral Weight Loss Program: Model Development Study

## TL;DR

This study uses a cognitive model to explore how people stick to tracking their diet in a digital weight loss program and finds that tailored feedback and support improve adherence.

## Contribution

The study introduces a novel use of the ACT-R cognitive architecture to model and analyze adherence dynamics in dietary self-monitoring.

## Key findings

- ACT-R modeling effectively captured adherence trends with low RMSE values across intervention groups.
- Tailored feedback and intensive support were linked to stronger goal pursuit and sustained behavior.
- Habit formation mechanisms declined over time, while goal pursuit remained dominant.

## Abstract

Self-monitoring of dietary behaviors is typically a central component of behavioral weight loss programs, and it is widely recognized for its effectiveness in promoting healthy behavior changes and improving health outcomes. However, understanding the adherence dynamics of self-monitoring of dietary behaviors remains a challenge.

We aimed to develop a prognostic model for adherence to self-monitoring of dietary behaviors using the Adaptive Control of Thought-Rational (ACT-R) cognitive architecture and to qualitatively investigate adherence dynamics and the impact of various interventions through model-based analyses.

The modeling data were derived from a digital behavioral weight loss program targeting adults who expressed a willingness to improve their lifestyle. Participants were assigned to 1 of 3 intervention groups: self-management, tailored feedback, and intensive support. ACT-R, a cognitive architecture simulating human cognitive processes, was used to model adherence to self-monitoring of dietary behaviors over 21 days, focusing on the mechanisms of goal pursuit and habit formation. Predictor and outcome variables were defined as adjacent elements in the sequence of self-monitoring of dietary behaviors. Model performance was evaluated using mean square error, root mean square error (RMSE), and goodness of fit. Mechanistic contributions were visualized to analyze adherence patterns and the impacts of different interventions.

The total sample size for modeling was 97, with 49 in the self-management group, 23 in the tailored feedback group, and 25 in the intensive support group. The ACT-R model effectively captured the adherence trends of self-monitoring of dietary behaviors, with RMSE values of 0.099 for the self-management group, 0.084 for the tailored feedback group, and 0.091 for the intensive support group. The visualized results revealed that, across all groups, the goal pursuit mechanism remained dominant throughout the intervention, whereas the influence of the habit formation mechanism diminished in the later stages. Notably, the presence of tailored feedback and the higher levels of social support were associated with greater goal pursuit and more sustained behavioral practice.

This study highlights the potential of ACT-R modeling for dynamic analysis of self-monitoring behaviors in digital interventions. The findings indicate that tailored feedback combined with intensive support may significantly improve adherence. Future studies should (1) extend the intervention duration to explore sustained adherence mechanisms, (2) integrate social cognitive factors to capture behavioral compliance insights, and (3) adapt dynamic models to inform just-in-time adaptive interventions for broader applications.