Large Language Models at AI-M: Language-Based AI for Psychiatry

Large language models (LLMs) are artificial intelligence systems designed to interpret and generate human language, creating new opportunities to support psychiatry across clinical care, education, and research. When informed by neuroscience and psychiatric knowledge – such as cognitive processes, affective dynamics, and clinical reasoning – LLMs can be developed to better reflect real-world mental health contexts and communication needs.

At AI-M, LLMs are studied and applied within a bidirectional learning framework in which clinical expertise shapes model design, evaluation, and safeguards, while LLM outputs support clinicians and researchers by synthesizing information, highlighting patterns, and assisting structured decision-making. By embedding LLMs within responsible development and deployment processes, our program ensures that language-based AI remains aligned with human cognition, clinical judgment, and patient safety.