AI for mental health: clinician expectations and priorities in computational psychiatry

Mental disorders represent a major global health challenge, with an estimated lifetime prevalence approaching 30%. Despite the availability of effective treatments, access to mental health care remains inadequate. Computational psychiatry, leveraging advancements in artificial intelligence (AI) and machine learning (ML), has shown potential for transforming mental health care by improving diagnosis, prognosis, and the personalization of treatment. However, integrating these technologies into routine clinical practice remains limited due to technical and infrastructure challenges. While ongoing computational developments will enhance AI's precision, many studies focus on its broad potential without providing specific, clinician-informed guidance for immediate application. To address this gap and the urgent need for clinically actionable AI tools, we surveyed 53 psychiatrists and clinical psychologists to identify their priorities for AI in mental health care. Our findings reveal a strong preference for tools enabling continuous monitoring and predictive modeling, particularly in outpatient settings. Clinicians prioritize accurate predictions of symptom trajectories and proactive patient monitoring over interpretability and explicit treatment recommendations. Self-reports, third-party observations, and sleep quality and duration emerged as key data inputs for effective models. Together, this study provides a clinician-driven roadmap for AI integration, emphasizing predictive models based on ecological momentary assessment (EMA) data to forecast disorder trajectories and support real-world practice.