Recipient organisationDouglas Hospital Research Centre
Funding£5.5M
PeriodApr 2025 — Apr 2030
In plain English
AI plain-English summary
People with psychosis often speak in ways that are hard to follow—jumbled, tangential, or disconnected—and this disorganisation of speech can be as disabling as hallucinations or delusions, yet its underlying causes remain largely unknown. DIALOG is an international collaboration, designed with input from people who have lived experience of psychosis, that aims to change that. The project will use large language models (the same type of AI behind tools like ChatGPT) to analyse thousands of hours of natural, everyday speech from patients, extracting objective markers of disorganisation that go beyond traditional clinical ratings. These speech markers will then be linked, for the first time, to brain imaging data—MRI, PET, and MEG scans from over 3,000 patients—to identify the specific neural processes, from synaptic density to oscillatory dynamics, that break down when speech becomes disorganised. If successful, DIALOG will provide a mechanistic, molecular-to-systems-level account of disorganisation, pinpointing precise targets for new treatments. This is fundamental science with a clear translational path: by cataloguing existing treatments and building infrastructure for multi-site trials, the project aims to accelerate clinical testing and, ultimately, improve social connectedness and recovery for people with severe mental illness.
View original technical description
Social connectedness is critical to recovery from psychosis in mental disorders and underpins vocational success. Persistent disorganisation of speech and language powerfully disrupts social interaction and perpetuates stigma, yet we know little about its causes. DIALOG, an international interdisciplinary initiative co-conceived with lived-experience experts, aims to address this and provide clarity on the neurobiological underpinnings of disorganisation. By focusing on patients’ everyday language use, rather than traditional clinical ratings, DIALOG seeks to identify the predictive computations implemented ‘on-the-fly’ by our brain when we interact with others, and how they break down to cause disorganisation. We leverage state-of-the-art Large Language Models (LLM) to extract objective markers of disorganisation from naturalistic speech, sampled longitudinally from patients with severe mental disorders. Combining LLM with large-scale new and legacy data from neuroimaging tools (MRI/PET/MEG; >3,000 patients) for the first time, we will identify the cardinal neurophysiological processes (synaptic density, neuronal connectivity and oscillatory dynamics) that underlie the computational failures behind disorganisation. In parallel, we will catalogue potential treatments and build the capacity for their multi-site evaluations, facilitating future trials and accelerating clinical translation. DIALOG will pioneer a computationally informed, molecular-to-systems-level account of disorganisation, identifying the precise mechanisms that can be targeted with novel treatments.
Applying neuroscience to understand symptoms in anxiety, depression & psychosis
Plain English summaries and category classifications on this site are generated by AI and may not perfectly reflect the original research.
Is something wrong? Let us know