Artificial intelligence in forensic psychiatry: potential applications and key considerations

Artificial intelligence (AI) has enormous potential to transform society. Forensic psychiatrists must consider its utility specific to the forensic population, and requirements for its safe and ethical implementation. In this narrative review, we consider how AI has the potential to assist clinical forensic psychiatry with diagnosis and risk assessment, and the practical, ethical and legal challenges to overcome. Early research shows potential for the use of AI in screening and identification of mental disorders, but with significant limitations and little consideration of forensic populations. More research has been done in the area of offending risk assessment, but clinical utility is limited by external validation and differentiation of offence type predicted. Practical and ethical challenges identified include obtaining the large volumes of representative training data required from various sources, heterogeneity of diagnostic constructs, lack of transparency as to how models develop outputs and the reliability of them, and the propagation of bias. We conclude by recommending increased data collection and sharing across agencies, external validation of AI models, enhanced clinician training and involvement in development of AI tools, and the development of ethical and legal safeguards and regulations on the use of AI in this setting.