Related Experiment Video
Updated: Sep 20, 2025
Concept Development and Use of an Automated Food Intake and Eating Behavior Assessment Method
Published on: February 19, 2021
Evaluation of Dietary Management Using Artificial Intelligence and Human Interventions: Nonrandomized Controlled
Fusae Okaniwa1, Hiroshi Yoshida2
1Department of Theoretical Social Security Research, National Institute of Population and Social Security Research, Tokyo, Japan.
This study compared how different types of feedback—automated AI messages versus a mix of AI and human video messages—helped people stick to a diet plan. Researchers found that combining AI with human support led to better program retention and greater reductions in body fat compared to using AI alone.
Area of Science:
- Digital health informatics within dietary management research
- Behavioral medicine and health promotion sciences
Background:
Prior research has shown that personal health records are expanding due to the widespread adoption of wearable technology and mobile applications. Traditional health promotion strategies managed by professionals often face significant constraints regarding financial resources and scalability. This gap motivated the medical community to explore automated guidance systems powered by machine learning. That uncertainty drove interest in whether software can successfully replicate complex human tasks. No prior work had resolved whether technology alone can maintain long-term lifestyle modifications. Current evidence suggests that automated systems struggle to keep individuals engaged over extended periods. This study addresses the necessity of determining if human interaction remains a requirement for successful health outcomes. The field currently lacks clarity on how to optimize digital tools for sustained behavioral change.
Purpose Of The Study:
This study aims to investigate whether automated systems can effectively encourage healthy behaviors or if human interaction remains necessary. The researchers sought to determine the optimal conditions for maximizing health improvements through digital guidance. They addressed the challenge of sustaining behavioral change using technology alone in the modern medical landscape. The team hypothesized that integrating human support with automated feedback would produce superior results. They focused on evaluating dietary management as a primary outcome for health improvement. The study intended to clarify the limitations of current software-based health promotion programs. By comparing different feedback models, the authors aimed to identify the most effective strategy for long-term engagement. This work provides insight into the potential for human-AI collaboration in future health interventions.
Main Methods:
The investigators performed a three-month nonrandomized controlled trial to assess behavioral adherence. They recruited one hundred and two individuals who actively used a mobile nutrition tracking application. These subjects were partitioned into three distinct cohorts to test varying levels of feedback. The first cohort received automated text messages generated by the software platform. The second cohort obtained both automated texts and personalized video messages from a human companion. The third cohort functioned as the control group, recording their intake without receiving any external feedback. The team evaluated program continuity by tracking participant dropout rates throughout the study duration. They also analyzed changes in physical markers to determine the efficacy of each feedback strategy.
Main Results:
The combined approach of automated texts and human video messages yielded the strongest improvements in participant retention. The hazard ratio for program dropout in the combined group reached 0.078, which was statistically significant. Participants receiving human video messages experienced a greater reduction in body fat percentage compared to the control group. The rate of body fat loss was particularly pronounced among those receiving highly individualized human support. Automated text messages alone successfully influenced body mass index but showed no significant impact on body fat percentage. The control group exhibited higher dropout rates than those receiving combined human and digital feedback. These findings indicate that human-led video messaging promotes voluntary health behaviors more effectively than automated systems. The data demonstrate that the source of information significantly alters the success of health-related behavioral changes.
Conclusions:
The authors propose that relying solely on automated systems is insufficient for maintaining healthy lifestyle habits. Their synthesis suggests that human-delivered information enhances the impact of digital health guidance. The findings imply that personal connection remains a key driver for voluntary behavioral adherence. The researchers highlight that nonexpert companions can effectively support health goals through video communication. This evidence indicates that combining automated messages with human video feedback improves program retention rates. The study suggests that individualized human interaction maximizes the efficacy of dietary management programs. These results demonstrate that the source of health information influences the success of behavioral interventions. The authors conclude that integrating human elements into digital platforms improves physical health markers like body fat percentage.
Frequently Asked Questions
The researchers propose that combining automated text messages with human video feedback maximizes program retention and body fat reduction. While AI-only messages influenced body mass index, they failed to significantly lower body fat percentage, demonstrating that human interaction is necessary for sustained behavioral change.
The study utilized a smartphone diet management application to deliver interventions. Participants in the combined group received video messages from a companion, whereas the AI-only group received standard automated text messages, and the control group received no feedback while maintaining dietary records.
The authors suggest that human video messaging is necessary to sustain participant engagement. The Cox proportional-hazards model revealed a hazard ratio of 0.078 for the combined group, indicating that human-led video support significantly reduces dropout rates compared to groups receiving no feedback.
The researchers employed a nonrandomized controlled trial design to analyze participant data. This data type allowed for the comparison of dropout rates and physical indicators, such as body fat percentage and body mass index, across three distinct intervention groups over a three-month period.
The study measured physical indicators including body mass index and body fat percentage. The researchers observed that the combined intervention group achieved a greater reduction in body fat percentage compared to the control group, particularly when the intervention was highly individualized.
The authors imply that person-to-person communication is vital for health interventions. They propose that even nonexpert companions can effectively promote voluntary health behaviors, suggesting that the human element of communication is more effective than automated information delivery alone.
