Projects

​

​

Linguistic Markers of Mental State and Status

​​​

The goal of this line of research is to apply natural language processing

techniques (notably large language models) to detect clinically actionable

changes in mental state and status. Current application areas include

speech-based detection of deterioration in dementia; detection of cognitive

distortions; persecutory ideation and thought disorganization in data from

participants experiencing symptoms of psychosis; and detection of suicide risk

in search log histories contributed by participants. The long-term objective of

this work is to develop tools to support clinicians by enhancing their ability to

identify and act upon identified linguistic indicators.

​​

​​

​​

Cognitive Behavioral Therapy for psychosis (CBTPro):

​

Psychosis resulting from schizophrenia spectrum disorders and major mood disorders is one of the most disabling health concerns worldwide. Evidence-based psychotherapeutic interventions are recommended as standard of care by national psychosis treatment guidelines but are rarely accessible. Cognitive Behavioral Therapy for psychosis (CBTp) is the most well-researched psychotherapy for psychotic disorders, yet fewer than 1% of American mental health providers are trained in this intervention. To date, there has been no scalable way to offer high-quality and sustainable training to mental health providers in CBTp-informed care. 

 

BRiTE faculty have partnered with small business Lyssn.io to systematically develop, pilot, and rigorously test CBTpro, a novel Computerized Clinician Support Tool that will use cutting-edge spoken language technology to provide immediate feedback and coaching to mental health staff and trainees who wish to learn CBTp. CBTpro will provide a rapid means of scaling and sustaining CBTp in routine care settings across the US, resulting in more clinicians  across the country providing higher quality CBTp to individuals with psychosis.

 

This work is supported by a National Institute of Mental Health (NIMH) Fast-Track Small Business Technology Transfer grant (R42MH123215-01). For more information about this project, please visit our website at: https://uwspiritlab.org/cbtpro/.

​​

​

Automated Analysis of Speech and Language
 

Many of the cardinal symptoms of mental health conditions are detected in spoken language. For example, disordered thinking may manifest as speech that appears incoherent, lacking meaningful connections between phrases or sentences. Changes in mood may be expressed explicitly in language, or detectable from subtle linguistic indicators.

 

This project aims to develop and evaluate automated methods that detect linguistic indicators of clinically important changes in mental state, with the overarching goals of developing instruments to support continuous monitoring, and advancing scientific understanding of the transdiagnostic spectrum of symptomatology in psychiatry. 

​​

​

​

Development of an AI Digital Navigator System to Support Patients’ Use of Technology Based Interventions

​​

Clinical studies have demonstrated that when used correctly and consistently, digital mental health interventions (DMHTs) can be highly effective. However, a growing body of work suggests that many intended users struggle to use these novel technologies independently, apply them to their unique life circumstances, and maintain consistent usage over time. The objective of this project is to leverage Artificial Intelligence (AI) to create COACH: an on-device AI-driven digital navigator system that will support patients’ effective use of Digital Mental Health Technologies. We aim to: 1. Develop a prototype chatbot-based digital navigator; 2. Conduct preliminary evaluation of the system including lab-based usability testing with healthy participants and “red-team” stress testing with project confederates.  

​​

​

​

Optimizing Telemental Health with Live Artificial Intelligence Clinical Scaffolding and Feedback

​

Recent advancements in artificial intelligence (AI) enable the analysis of therapeutic interactions to identify intervention opportunities, provide feedback, and track progress. This project aims to develop a clinical scaffolding system that enhances telemental health care by offering real-time coaching and actionable suggestions during video-based sessions. Modeled after live supervision methodologies, the system supports clinicians by identifying intervention targets and providing text-based coaching prompts to guide care. Unlike automated chatbots, this approach allows clinicians to adapt suggestions to meet patient needs, balancing automation with oversight to ensure safer AI-supported mental healthcare. The proposed in-session support will facilitate the efficient implementation of strategies and help develop clinician skills. Additionally, this project seeks to enhance data privacy by processing all data on-device and avoiding external data transfers.