An explainable AI framework for early mental health screening and clinical decision support using reflective listening and generative AI

Counts differ because each database indexes a different set of publications. We treat OpenAlex as the canonical count; Google Scholar is not shown (no API, and crawling it violates its ToS).

Mental health disorders represent a significant global public health challenge, and many individuals remain unidentified during the early stages due to limited awareness, social stigma, and restricted access to mental healthcare services. Recent advances in Artificial Intelligence (AI) and Generative AI provide new opportunities to support early mental health screening and assist healthcare professionals during the initial assessment process. This paper proposes an Explainable AI-based Mental Health Screening and Clinical Decision Support Framework that integrates reflective listening, Large Language Models (LLMs), Knowledge Graph reasoning, and DSM-5-TR-informed clinical guidelines to facilitate empathetic conversational screening and structured mental health risk estimation. The proposed framework employs a two-stage screening workflow. In the first stage, a reflective-listening conversational agent conducts empathetic dialogue, extracts clinically relevant symptoms, and performs an initial mental health risk screening using contextual reasoning and Knowledge Graph representations. In the second stage, standardized psychometric instruments, including the Penn State Worry Questionnaire (PSWQ), Beck Depression Inventory-II (BDI-II), and McLean Screening Instrument for Borderline Personality Disorder (MSI-BPD), are incorporated to provide quantitative assessment and support disorder-specific risk estimation. To facilitate model development, we constructed a structured dataset comprising AI-generated reflective conversations, DSM-5-TR-informed screening labels, extracted symptom features, Knowledge Graph representations, and clinician-reviewed annotations, which was subsequently used to fine-tune the proposed language model. Experimental results demonstrate encouraging agreement between the framework’s AI-assisted mental health screening outcomes and clinician-reviewed assessments, while highlighting the effectiveness of integrating reflective listening, psychometric evaluation, Knowledge Graph reasoning, and Generative AI to improve screening performance, interpretability, and user engagement. The proposed framework is intended as an AI-assisted mental health screening and clinical decision-support tool to support early identification of individuals.