AI Predicts Schizophrenia Via Hidden Linguistic Patterns

Scientists at the UCL Institute for Neurology have utilized AI language models to analyze the subtle speech patterns of individuals with schizophrenia, potentially revolutionizing psychiatric diagnosis and assessment. Their research suggests that the language model could predict word choices in control participants more accurately than in those with schizophrenia, indicating possible connections to how the brain formulates memory relationships known as ‘cognitive maps.’ This novel application of AI in psychiatric study holds promise for a more nuanced and data-driven approach to understanding and diagnosing mental disorders.

Currently, psychiatric diagnosis heavily relies on patient interviews, with limited tests such as blood tests and brain scans. However, this lack of precision hinders a comprehensive understanding of mental illnesses and monitoring of treatment effectiveness. The research involved participants with schizophrenia and control participants completing verbal fluency tasks. Using an AI language model that was trained on extensive internet text, researchers found that the model could predict responses from control participants more effectively than those with schizophrenia, especially in patients with severe symptoms.

The researchers believe that these differences in predictability may be related to how the brain creates and stores cognitive maps. The team plans to expand the use of this technology in larger patient samples and various speech settings to evaluate its practicality and effectiveness in a clinical environment.