Disorganisation, social dysfunction, and LLM-based speech metrics 

Disorganised speech—when thoughts become hard to follow or jump from topic to topic—is a common and often disabling symptom in severe mental illnesses. It can make everyday conversations difficult and affect a person’s ability to work, study, or maintain relationships. However, traditional ways of identifying and measuring disorganisation are often inconsistent and influenced by other symptoms. In this project, we use advanced Large Language Models (LLMs) to analyze short samples of natural speech and detect subtle patterns that reflect disorganised thinking.

By applying these AI tools to the largest dataset of its kind, we aim to uncover how speech disorganisation relates to real-world social and occupational functioning. Our goal is to develop objective, reliable, and scalable tools that can help clinicians better understand and monitor mental health challenges—moving beyond subjective ratings to more precise, data-driven insights.