ATTITUDES TOWARD ARTIFICIAL INTELLIGENCE CLINICAL DECISION SUPPORT IN PSYCHIATRY TRAINING: CURRENT EVIDENCE AND FUTURE DIRECTIONS

Background

Artificial intelligence (AI) is increasingly integrated into healthcare, with psychiatry adopting predictive modeling, suicide risk assessment, and documentation support. AI Clinical Decision Support (CDS) tools hold promise for improving diagnostic accuracy, personalizing treatment, and streamlining workflows. Yet ethical concerns, autonomy, and limited training temper enthusiasm. The primary objective of this review is to examine evidence on clinician attitudes and identify gaps in psychiatry training.

Methods

A literature review was conducted across PubMed, PsycINFO, and Scopus using keywords including artificial intelligence, psychiatry, clinical decision support, attitudes, residents, education, and mental health professionals. Studies were included if they examined AI perceptions, attitudes, training, or adoption among psychiatrists, trainees, or mental health professionals. No date or language limits were applied. The search yielded 210 records; after duplicate removal, 40 full texts were reviewed, and 9 met inclusion criteria.

Results

Across regions, clinicians showed cautious optimism toward AI through various applications. Surveys in Nigeria and Saudi Arabia evaluated CDS for diagnosis and treatment planning in hospital psychiatry, recognizing its potential but limited familiarity. Concerns included empathy, confidentiality, and data security. A mixed-methods study found nearly half of clinicians lacked AI training, yet many valued predictive algorithms for suicide risk, symptom monitoring, and relapse detection, especially outpatient. Broader physician surveys reinforced ambivalence: clinicians doubted AI could replace relational aspects but endorsed its use for documentation, efficiency, and diagnostic support. Educational studies highlighted the potential of adaptive AI learning platforms in psychiatry and psychology training. Systems-level analyses emphasized big data analytics for hospital efficiency and forecasting, while noting barriers with infrastructure, workforce, and readiness.

Conclusions

This review synthesizes evidence, underscoring both optimism and skepticism toward AI in psychiatry. While professionals recognize AI’s potential for diagnosis, prediction, and education, adoption remains constrained by training gaps and ethical concerns. No studies examined psychiatry residents’ perspectives, leaving a gap in understanding how trainees view AI in relation to trust, autonomy, and curricular integration. Another gap is cultural context: most studies did not assess how regional and cultural differences influence attitudes, trust, or ethical concerns. Training AI models on linguistically and ethnically cognizant datasets is essential to reduce bias and improve cultural relevance across populations. By identifying these limitations, this review clarifies the evidence base and highlights directions for future research, including trainee-focused investigations and culturally informed approaches to AI in psychiatric education.