AI and the clinical blind spot: why psychiatry should be part of the conversation about artificial intelligence and human behaviour

Abstract

Artificial intelligence (AI) systems that model human behaviour are built on data humans generate about themselves, yet the assumption that this data is reliable remains largely unexamined. Psychiatry's evidence base directly challenges it. Self-report is shaped by cognitive distortions, the gap between clinical presentation and underlying experience, and conditions that make accurate self-report impossible. The populations most affected by AI-driven decisions can be the least represented in training data. Clinical expertise in the reliability of human data, which psychiatry holds and practises daily, is largely absent from AI development. This omission is correctable, but psychiatry must enter the conversation.