Artificial intelligence conversational agents for high school wellbeing interventions: A scoping review

Student wellbeing has been steadily declining over time, surpassing typical developmental fluctuations. In high school settings, educators are tasked with supporting student wellbeing alongside various other aspects of student learning and development. As educators face expanding curriculum and resource constraints, they are increasingly turning to Artificial Intelligence (AI) technologies. While the adoption of AI technologies is increasing rapidly, primarily in the form of AI conversational agents (CAs), commonly referred to as chatbots, their efficacy in promoting student wellbeing remains largely unexplored. Given the current state of the literature, it is challenging for practitioners to devise policies and regulations that guide the use of AI in supporting student wellbeing. To address this issue, we conducted a scoping review that incorporated both peer-reviewed academic literature and grey literature to assess the current landscape of AI-driven wellbeing interventions in educational contexts. A total of 12 studies met the eligibility criteria, encompassing a diverse range of designs and methodological qualities. Findings suggest that AI can have a positive influence on student wellbeing, particularly in reducing depressive symptoms and stress levels. However, due to limited studies with high-quality evaluation design, inconsistent outcome measures and generally low sample sizes, the strength of the evidence behind conversational agents having a positive impact on wellbeing is weak. Further, the significant variation in AI conversational agent design made comparison across the studies difficult. While AI holds the potential to support student wellbeing, most papers reviewed lacked clear student perspectives (n = 2/12) in conversational agent design and exhibited methodological limitations (n = 6/12). Future research should prioritise transparent design processes, employ robust evaluation methods, and follow a clear wellbeing framework to allow for validation of findings.