Using Large Language Models to Assess Psychopathology

Comprehensive psychiatric assessments require considerable time and human judgement. But do they really? Whitney Ringwald, faculty, conducted a study, “Scalable, Context-Sensitive Psychiatric Assessment with Large Language Models and Brief Diaries,” examining an efficient, non-human alternative to classic clinical interviews.

Typically, psychopathology is assessed through either clinical interviews, which are very time- and resource-intensive, or patient survey reports, which cannot provide the thorough context required for a robust examination. Ringwald and her collaborators sought to determine whether Large Language Models (LLMs), a specific type of AI, can ameliorate this issue by measuring psychopathology from how people describe their day-to-day lives.

The researchers found that LLMs can accurately assess psychopathology from mere minutes of open-ended daily diary recordings. These results suggest that LLMs can be a valuable tool for translating people's unique, contextualized experiences into usable diagnostic information. If bolstered by future replications and research testing generalizability and potential for bias,  the study findings support the potential for LLMs to revolutionize psychiatric assessment, particularly in clinical care settings and with studies that currently come with high administrative burden.