REACT is a five-week micro-randomized trial in University of Florida students age 18 and older. Participants complete a baseline battery, wear a Garmin Venu 3, and respond to five to six smartphone check-ins per day. Week one is a no-prompt run-in that establishes each participant's own baseline distribution of momentary affective variability, indexed by the mean square of successive differences on consecutive affect ratings. From week two through week five, the app evaluates up to six decision points per day. A decision point is flagged when the participant's current variability exceeds their own 80th percentile threshold, computed within person rather than against a group average. At each available decision point the app randomizes with probability 0.5 whether to deliver a brief coping message, and if so selects a message type from those eligible in that context. Availability rules exclude sleep, stale device data, a 60-minute cooldown, a daily cap, and any safety override. Proximal outcomes are drawn from the first check-in within two hours after the decision point. Analysis uses weighted and centered least squares to estimate causal excursion effects, with pre-specified moderators.
