Explainable Artificial Intelligence in Mental Health Assessment: A Systematic Review of Interpretability, Psychometric Reporting and Clinical Utility

The introduction of artificial intelligence (AI) into qualitative data analysis has significantly sped up the coding, clustering, and pattern identification process at scale more than ever before. The surveys of the industry show that by 2024, it is expected that about 77 percent of academic research teams will have utilized some type of AI-assisted analytical tooling, an increase of nearly seven times compared to 11 percent in 2018. Nevertheless, despite this high level of diffusion, technologies that focus on automation endanger the interpretive, reflexive and context-sensitive epistemological principles of qualitative inquiry. The current paradigms offer a paradigm methodological dilemma that has been addressed through manual techniques that offer interpretive profundity and theoretical sensitivity, but are limited in scalability and time-efficiency, and AI-based systems that offer computational efficiency but epistemic overreach, lack of analytic transparency, and diminished construct validity. To fill this gap, this paper suggests the Human-Governed Qualitative Intelligence Model (HGQIM), which was based on the architectural concepts of the Qualivers platform. The HGQIM combines the computational processing and systematic human control, allowing researchers to confirm, alter or dismiss AI-generated interpretations in a transparent and reflexive system. The model is based on the principles of interpretivist epistemology, construct-validity theory, and human-in-the-loop (HITL) AI principles and redefines AI as an enhancement tool and not an autonomous interpreter. This paper presents a new methodological synthesis by introducing the notion of qualitative intelligence that balances the rigour of analytic methods with the scalability of computational methods, improves accountability, and maintains researcher agency. The research has a methodological innovation contribution to the field of qualitative research and sets a conceptual background to future empirical verification of AI systems run by humans.