Getting personalized feedback on meal photos may help people keep using a healthy-eating app, but it does not necessarily change what they eat, according to a randomized field experiment examining how social features shape diet-tracking behavior.
The study, published in the International Journal of Information Management, randomly assigned 337 U.S. adults to different versions of a photo-based healthy-eating app for four months. All versions included a photographic food diary, while some added personalized feedback from a registered dietitian, peer meal-sharing features or both. The data were collected in 2014, so the findings do not reflect today’s more advanced apps, but the experiment offers evidence about how expert guidance and social comparison may affect eating-related behavior.
Of the 337 people randomized, 197 recorded at least one meal and were included in the app-use analyses. A smaller group of 180 participants had enough meal information for researchers to evaluate food choices.
Personalized dietitian feedback appeared to help people stay engaged. Users who received weekly feedback based on their meal photos recorded meals more often, and their decline in app use over time was slower than among users without the expert feature.
That greater engagement did not translate into a significant improvement in meal balance.
Researchers measured food choices by having a dietitian rate how closely users’ meals matched the USDA MyPlate recommendation that fruits and vegetables make up about half the plate. Neither expert feedback nor peer observation significantly improved that measure over the course of the study.
“Diet-related apps reach a wide range of people, but we know little about their effects on users’ eating behaviors,” said study co-author Rema Padman of Carnegie Mellon University. “Having a deeper understanding of these features is essential for both research and practice.”
The peer feature produced a very different pattern.
Participants who could see other users’ meal photos were more likely to disengage from the app, and their use declined faster over time. Peer observation also did not significantly improve meal balance.
Further analysis suggested that the negative effect was concentrated among users whose eating patterns were less healthy than those of the people they were observing. The farther those users’ meals were from the group norm, the more likely they were to reduce their app use.
The researchers interpreted that pattern through social comparison and what they call defensive avoidance. Seeing other people consistently post healthier-looking meals may motivate some users, but for people who feel far behind the group, the comparison may instead become discouraging enough that they disengage.
One participant assigned to the peer group even reported dropping out because watching others eat healthfully made them feel dispirited.
That finding complicates a common assumption built into many health apps: that seeing other people model healthier behavior will automatically motivate users to do the same.
It also highlights the difference between app engagement and actual dietary change. Logging more meals, opening an app more often or sticking with it longer may be useful signs of self-monitoring, but those behaviors do not necessarily mean someone is eating differently.
The experiment has several limitations.
Participants were primarily younger U.S. Android users who already had an interest in healthy eating, which limits how broadly the findings can be applied. The app was also intentionally simplified so researchers could isolate the effects of expert and peer features, meaning it did not include many of the gamification, competition and personalization tools found in commercial diet apps.
The age of the data is another important consideration. The study was conducted in 2014, well before today’s widespread use of AI-generated feedback, automated image recognition, wearable integration and more sophisticated recommendation systems.
The researchers argue, however, that the underlying behavior may still matter because the study tested responses to expert guidance and peer comparison rather than the performance of a particular app.
The food-choice measure was also relatively narrow. It focused on the share of fruits and vegetables in a meal rather than other aspects of diet quality such as macronutrients, micronutrients, cooking methods or overall calorie intake. The ratings were based on user-submitted meal photos and descriptions, which introduces some subjectivity even though the dietitian making those ratings was blinded to participants’ study groups.
For consumers, the findings suggest that more social interaction inside a diet app is not automatically better. Personalized expert guidance may help people keep tracking their meals, while comparison with other users may discourage some of the very people an app is intended to help.
But neither feature, by itself, produced a clear improvement in what participants put on their plates.
The experiment and data collection were supported in part by funds provided by Carnegie Mellon University faculty members S. Roehrig and R. Krishnan, as well as the Berkman and GuSH funds at Carnegie Mellon. Additional support for analysis and manuscript preparation came from a Summer Research Grant from Hofstra University’s Frank G. Zarb School of Business. The researchers also acknowledged PHRQL Inc., the company that provided access to and technical support for the smartphone app used in the study