theaicatalog
Google ยท Research

Google Research builds wearable health-age tool for habit coaching

Google Research describes an investigational machine learning framework that turns passive wearable data into a coachable health age and uses a world model to simulate 12-week "what-if" habit changes. The sensor world model, SensorWM, was pre-trained on 2.28 million person-weeks of data from roughly 100,000 consenting Fitbit users and simulates 125 candidate habit combinations for each person. In a blinded study with five certified lifestyle coaches and five board-certified physicians across 49 participant profiles, coaches chose the counterfactual-grounded plan as the best overall plan in 71.4% of evaluations, and physicians rated 99.2% of plans within safe exercise intensity limits. The authors say these are research prototypes tested offline on retrospective cohorts, not clinical diagnostic tools.