Smartphone apps are becoming more intelligent and capable with each passing day and now a new app capable of predicting and intervening in users’ overeating habits has been developed.

Researchers at Worcester Polytechnic Institute (WPI) and the University of Connecticut (UConn) are working on an app aimed at helping users manage their overeating challenges in three key ways: by tracking eating patterns, providing interventions, and helping change behavior. The app — and its distinctive methodology — showed tremendous promise in an early pilot study.

To that end researchers have also presented a paper at the annual symposium for the American Medical Informatics Association. In the paper scientists noted that a user study of 16 participants, all over the age of 18, overweight, and not necessarily trying to lose weight, showed positive results. At the end of the month-long pilot, nine of the participants lost an average of five pounds, three weighed the same, and four gained an average of two pounds.

The research team’s three-pronged approach is an important differentiator in a crowded marketplace. According to a 2017 report in the International Journal of Obesity, there are nearly 29,000 weight-related apps on the market, with most focused only on tracking physical activity, calories, and body weight. The drawback with most weight-loss apps is that they are burdensome to use and don’t focus on actually trying to change the user’s behavior.

The WPI/UConn app, dubbed SlipBuddy, is built to do just that. In addition to tracking stress and eating, it’s a personalized intervention system built through the integration of behavioral strategies and technologies like mobile devices, machine learning, and text mining. The work is aimed at helping users identify what triggers them to overeat and inserting new stimuli that, instead, trigger healthy behaviors.

SlipBuddy is designed to keep user interaction simple. Users are asked to check in three times a day to note their stress levels, fatigue, and hours of sleep, and whenever they feel they’ve overaten. Over time, as the app collects information (which is stored as anonymized data on a WPI server), the system uses machine learning to find patterns that can be used to predict when the user is likely to overeat. The app uncovers user-specific triggers, like late-night eating, watching TV, or sleepless nights. When the app predicts that the likelihood of overeating is high, it suggests interventions–for example, taking a walk, turning off the TV, or engaging in other activities that can help users reduce stress without resorting to eating.