Human-centered AI to promote youth mental health: a serendipitous natural experiment enabled by a digital health platform

Abstract

Background: Health systems are struggling to deliver timely preventive care, particularly for marginalized populations, necessitating integration across health, education, and social services. For Indigenous youth in rural communities, fragmented services, isolation, and limited culturally safe options worsen mental health disparities. Interactive technologies, particularly human-centered artificial intelligence (AI)-enabled digital health platforms grounded in human-computer interaction (HCI), can enable remote interaction with citizens and decision-makers. This study investigated a serendipitous natural experiment to assess varying levels of platform nudging on Indigenous youth compliance in a longitudinal intervention. Method: This study emerged from the final year of a 5-year initiative embedding a culturally appropriate digital health intervention into school curricula in rural Indigenous communities. While the broader aim was to assess long-term mental health outcomes, an unexpected system disruption assessment of digital nudging on compliance. The platform featured two interfaces: a citizen-facing mobile app for ecological assessments and nudges, and a scientist dashboard for monitoring engagement and triggering nudges. Youth received three nudges: (1) daily system-triggered reminders to complete assessments, (2) weekly non-personalized messages (e.g., land-based activity reminders), and (3) weekly personalized “Best Picture” messages showcasing youth-submitted images. The disruption created four phases: Phase 1 included all nudges; Phase 2 removed non-personalized and personalized nudges; Phase 3 reintroduced them; Phase 4 removed only personalized nudges. Data were analyzed using one-way analysis of variance (ANOVA) with Tukey post hoc tests in R 4.4.2. Results: Compliance, measured by completed mobile ecological prospective assessments (mEPAs), varied significantly across most phases. Comprehensive nudging (Phase 1) yielded the highest completion rates and fastest response times, which declined following the removal of personalized scientist-triggered nudges. Loss of personalized scientist-triggered nudges had the most substantial impact on compliance. Conclusions: Consistent system-triggered reminders and personalized “Best Picture” nudges were most effective in sustaining compliance. Findings highlight the importance of integrating personalized, two-way communication features into digital health platforms to strengthen engagement in rural Indigenous communities. By enabling real-time interaction between youth and scientists, the platform supported integration across health, education, and research sectors. Its human-controlled backend and customizable citizen-facing interface reflect principles of human-centered AI, emphasizing trust and autonomy. This approach offers a scalable model for ethical, effective digital interventions that balance technological precision and participant agency.