The protocol targets a core challenge in digital health: most physical-activity interventions are not genuinely individualized over time. YourMove's question is whether a control-systems-engineering approach — building a personalized dynamic model of each person and then driving an adaptive controller from it — can outperform a sensible non-adaptive program. The study is structured as a control optimization trial nested inside a randomized controlled trial, so it serves both an optimization goal (learning each person's dynamics) and a confirmatory goal (a head-to-head comparison against an active control).
The design is a 12-month, two-arm randomized controlled trial enrolling 386 inactive adults aged 25 to 80, randomized 1:1 and stratified by sex. The intervention arm proceeds through three sequential phases: a 10-day baseline establishing each participant's median daily steps; a 22-day-cycle open-loop system-identification phase in which daily step goals and point rewards are varied pseudorandomly to fit an individualized computational dynamic model; and a closed-loop controller phase in which a control-systems decision algorithm continually adapts goals and rewards using that personalized model, paired with a self-experimentation tool. The active control arm receives a fixed daily goal of 10,000 steps and a static daily point allocation, holding the structure constant while removing the adaptive, model-driven element.
Both arms share common scaffolding so the contrast isolates the controller: a Fitbit Versa smartwatch, weekly SMS messages, and modest incentives, with objective activity measured by an ActiGraph GT3X+ accelerometer at baseline, 6 months, and 12 months. The primary outcome is minutes per week of moderate-to-vigorous physical activity at 12 months. The recruitment and data-collection windows are specified (recruitment October 2022 to August 2024; data collection concluding August 2025), and the author list places the team spanning the Hekler and Rivera groups, with Park among the authors, continuing the control-engineering line of this research program.
Honest context: this reports the published study design only and deliberately contains no participant data or outcomes — appropriate both to the protocol's status and to the fact that the confirmatory results lie outside the protocol paper. YourMove is best read as the next rung after JustWalk: where the dissertation trial established that individualized just-in-time states can be identified in an open-loop experiment, YourMove operationalizes those insights into a closed-loop controller and subjects it to a randomized comparison at substantially larger scale.