Can a Semi-Automated Coach Keep Dieters on Track? A $3.4M NIH Grant Will Find Out

Here is a reliable pattern in weight-loss programs: tracking what you eat, how much you move, and what the scale says is one of the strongest predictors of success. Here is an equally reliable pattern: most people stop tracking within weeks.

The missing ingredient, according to Rebecca Krukowski, PhD, at the University of Virginia School of Medicine, is feedback. Personalized messages from a trained professional - responding to what someone actually logged, not generic encouragement - can double the amount of weight lost in behavioral programs. But writing those messages takes about 26 minutes per person per week. Most clinical and community programs simply do not have the staff hours.

A new five-year, $3.4 million grant from the National Institutes of Health will fund Krukowski's attempt to solve this bottleneck by building a semi-automated feedback system that pairs human expertise with algorithmic support.

The homework analogy

Krukowski frames the problem simply. Self-monitoring in a weight-loss program is like homework. If no teacher ever grades it, most students stop doing it. The tracking apps and activity monitors exist. What is missing is the human response that makes the effort feel worthwhile - the praise, the accountability, the personalized suggestions for reaching individual goals.

The study, co-led by Kathryn M. Ross, PhD, MPH, at Wake Forest University School of Medicine, will unfold in two phases. In the first phase, 300 participants across the country will enroll in a 16-week weight-loss program. Each will receive an e-scale, an activity monitor, and a diet tracking app. Trained professionals will write personalized weekly feedback messages responding to participants' self-monitoring data.

Over two years, the research team will analyze how different types and lengths of feedback affect tracking behaviors and weight loss. They will also examine whether variables such as age, sex, and rate of weight loss should change how feedback is personalized - a precision-medicine approach to behavioral support.

Building the hybrid system

Those findings will inform the second phase: developing and refining a semi-automated feedback system. The goal is not to replace human coaches entirely but to find what Krukowski calls the sweet spot - the combination of human input and automated generation that maintains effectiveness while dramatically reducing the time burden on professionals.

The computer science expertise comes from University of Florida researchers Jaime Ruiz, PhD, and Lisa Anthony, PhD, along with biostatistics expert Peihua Qiu, PhD. Once the system is built, it will be tested with 50 additional participants.

If it works, the implications extend beyond the study itself. Effective weight-management programs could become easier to offer at scale, particularly in underserved and rural communities where trained professionals are scarce. The approach could potentially apply to programs supporting people after bariatric surgery or those taking obesity medications.

What we do not yet know

The study is still in its earliest stages - no participants have been enrolled and no automated system has been built yet. The central question of whether a machine can replicate the motivational effect of a human coach remains genuinely open. Past research shows the human element matters, but it is unclear exactly which aspects of personalized feedback drive the effect. Is it the specific dietary advice? The tone? Simply knowing someone is paying attention?

There is also a selection issue. People who enroll in weight-management studies tend to be more motivated than the general population. Whether a semi-automated system would work as well for less-engaged participants - arguably the people who need it most - is an open question.

The 26-minute-per-person time estimate also comes from programs with specific protocols. How much that number can realistically be reduced while maintaining quality feedback is exactly what the study aims to discover. It is possible the sweet spot requires more human involvement than the economics of scalability would prefer.

Still, the underlying logic is sound. If personalized feedback doubles weight-loss outcomes - and the evidence consistently shows it does - then finding ways to deliver that feedback at lower cost is a problem worth the investment.