AI-powered counseling chatbots for adolescent mental health in schools

Thesis Abstract

This study addresses the growing challenge of adolescent mental health in school settings by examining the effectiveness, acceptability, and ethical implications of AI-powered counseling chatbots as a scalable, stigma-free support mechanism. The aim is to assess whether chatbot-delivered psychosocial support can reduce symptoms of anxiety and depression, improve help-seeking behavior, and augment traditional school counseling services. Specific objectives include (1) evaluating changes in standardized measures of anxiety (GAD-7) and depressive symptoms (PHQ-9) among 600 students aged 13–18 over a 12-week intervention; (2) assessing user engagement, therapeutic alliance, and perceived usefulness via the System Usability Scale (SUS) and the Working Alliance Inventory for Technology-Based Interventions; (3) exploring perceived barriers, ethical concerns, and privacy considerations through semi-structured interviews with students, school counselors, and parents; (4) comparing outcomes across gender, grade level, and prior mental health service use; and (5) developing a contextualized implementation framework for integrating AI chatbots within school mental health systems. The research adopts a mixed-methods, quasi-experimental design, underpinned by the Unified Theory of Acceptance and Use of Technology (UTAUT) and the model of Therapeutic Alliance in Computer-Mediated Counseling. The quantitative component employs a cluster-randomized trial with 20 schools assigned to intervention (n ? 400 students) or control (n ? 200 students) conditions, using pre-intervention, midline (6 weeks), and post-intervention (12 weeks) assessments. Primary outcomes include changes in GAD-7 and PHQ-9 scores analyzed through linear mixed-effects modeling to account for nested data (students within schools). Secondary analyses comprise repeated-measures ANOVA for engagement metrics, and logistic regression to identify predictors of help-seeking behavior. The qualitative strand involves thematic analysis of 40–60 semi-structured interviews with students, counselors, and parents, triangulated with chatbot interaction transcripts to extract themes related to user experience, trust, and ethical considerations. Data collection instruments include validated scales (GAD-7, PHQ-9, SUS, Working Alliance Inventory for Technology-Based Interventions) and a bespoke interview guide addressing privacy, cultural relevance, and perceived safety. Validity and reliability procedures encompass pilot testing of the chatbot interface, intercoder reliability checks (Cohen’s kappa ? 0.80) for qualitative coding, and instrument reliability assessments (Cronbach’s alpha ? 0.70). Anticipated findings indicate a modest but statistically significant reduction in anxiety and depressive symptoms in the intervention group compared to controls, with higher engagement correlating positively with symptom improvement. The study is expected to reveal nuanced variability by gender and prior service use, and to identify ethical tensions surrounding data privacy, consent, and boundary conditions for AI-mediated conversations. The contribution to knowledge includes empirical evidence on the feasibility, effectiveness, and ethical governance of AI-driven mental health support in schools, a theoretical refinement of therapeutic alliance constructs in technology-enabled counseling, and a practical implementation framework for integrating AI chatbots with existing school-based services. The findings are anticipated to inform policy discussions on resource allocation, data governance, and adolescent welfare, as well as to guide developers in refining chatbot behavior, cultural adaptability, and safeguarding features. The research concludes with recommendations for scalable integration of AI-powered chatbots into school mental health ecosystems, including guidelines for consent procedures, inclusivity across diverse student populations, ongoing monitoring of risk indicators, and collaboration protocols between educators, counselors, and technologists to sustain ethical, effective, and adolescent-centered practice.