Learning to Be Proficient? A Structural Model of User Dynamic Engagement in eHealth Behavioral Interventions

Digital health interventions can be beneficial, yet users may disengage before realizing their benefits because noisy feedback makes it difficult to tell whether an intervention is actually working. Using data from an online weight-loss platform, we examine how users learn about intervention effectiveness through repeated experience and how this learning shapes their subsequent engagement. We find that learning conditions differ across intervention types: behavior-specific interventions provide relatively clearer feedback, whereas outcome-oriented and general self-regulation interventions involve greater uncertainty. Our counterfactual analyses suggest that reducing noise can improve learning and help prevent premature disengagement. For digital health platforms, this means designing not only effective interventions but also better environments for users to evaluate them. Platforms can aggregate outcomes over longer periods, provide relevant peer experiences, and use realistic historical benchmarks to calibrate expectations. They can also tailor support to the source of uncertainty, for example, contextualizing short-term outcomes for outcome-oriented interventions and providing more structured guidance for self-regulation activities. Helping users distinguish temporary fluctuations from meaningful evidence of effectiveness can support more informed decisions and sustained engagement.

• 저널Information Systems Research
• 토픽Digital Mental Health Interventions · Applied Psychology
• DOI10.1287/isre.2022.0377
• 저자Tongxin Zhou, Yingfei Wang, Lu Yan, Yong Tan