JMIR Publications

ABSTRACT

Background:

Artificial intelligence (AI) has rapidly transformed psychological research and mental health practice through advances in machine learning, deep learning, natural language processing, and large-scale data analytics. AI-based systems are increasingly employed to support psychological assessment, diagnosis, intervention, monitoring, and clinical decision-making. This rapid expansion has resulted in a substantial and growing body of empirical and review literature. However, despite the accelerated development of AI applications in psychology, discussions surrounding ethics, legal frameworks, and governance have not progressed at a comparable pace.

Objective:

Concerns related to privacy, transparency, data security, informed consent, algorithmic bias, and emotional safety remain particularly critical in psychological contexts, where AI systems may influence highly sensitive aspects of human experience. Given the rapidly evolving and heterogeneous nature of the literature, this study aimed to conduct a scoping umbrella review to map the breadth of existing evidence, identify key thematic domains, and highlight gaps in the application of AI in mental health.

Methods:

Abstracts of 1,827 records retrieved from Web of Science (n = 50), PubMed (n = 677), and Scopus (n = 1,100) were screened. Following a full-text assessment of 218 potentially eligible studies, a total of 182 review articles were included in the final synthesis.

Results:

The findings indicate that research on AI in psychology is primarily organized around five thematic domains: intervention, diagnosis, prediction, theoretical framework, and ethical issues. The intervention domain represents a substantial proportion of the literature, suggesting that AI is most frequently examined in relation to applied psychological functions. In contrast, ethical issues are comparatively underrepresented. This pattern reflects a broader imbalance in which the field is progressing through application-driven innovation, while ethical reflection remains relatively limited and often theoretical. Although AI-based interventions and assessment tools are expanding rapidly, only a small number of reviews have systematically examined how these systems address core ethical concerns, including informed consent, data privacy, accountability, cultural bias, and emotional safety. Furthermore, the increasing reliance on cloud-based infrastructures introduces additional challenges related to confidentiality, cross-border data transfers, third-party access, and system reliability in sensitive clinical settings.

Conclusions:

Taken together, these findings underscore the risk of integrating AI technologies into psychological practice without sufficient ethical, clinical, and infrastructural safeguards. Future research should prioritize the development of evidence-based and context-sensitive ethical frameworks, alongside the exploration of alternative implementation models—such as local or hybrid infrastructures—that can better balance scalability, privacy, and institutional control.