Training AI to perform semi-structured psychopathology assessments of common mental health symptoms reliably and to scale
Project Description
Accurate measurement of psychopathology is fundamental to public mental health information, policy and planning, and in clinical practice assessment of intervention need and outcome. The persistent growth in rates of common mental health conditions (confirmed in APMS 2025) depends on fully structured measurement in surveys in which the investigator has limited control over measurement accuracy. Semi-structured assessments such as the Present State Examination (or SCAN) are expert rated (akin to medical pathology microscopy diagnostics) but prohibitively costly. AI trained interviewing may be a solution. Previous research using expert systems failed to do so reliably (Brugha et al, 1996).
Following a literature review the student, who will also be SCAN trained, will assemble a PPIE group to codesign the study. S/he will begin by training an AI application to measure first one and then more core psychopathology symptoms (e.g. worry, anhedonia, hallucinations). Only data internal to the study will be used. S/he will then evaluate measurement accuracy jointly with PPIE co-designing members of the public and if successful, in consenting informed NHS patients.
AI training will use existing WHO affiliated SCAN trainer expertise and a large database of already captured examples of patient responses to SCAN interviews, responses and expert ratings. Co supervision and collaboration will be by members of a WHO affiliated worldwide expert group including AI expertise (keeping abreast of technology progress).
Reference
Brugha, T.S., Teather D, et al. (1996) Present State Examination by microcomputer: Objectives and experience of preliminary steps. Int. J Meth Res in Psychiat. 6, 143-151.

Theme
Common Mental Health
Primary Approach
Digital Technologies & Artificial Intelligence
Institutional Requirements

Professor Traolach (Terry) Brugha
Professor of Psychiatry