The Center for Responsible and Effective AI Technology Enhancement of Treatments for PTSD (CREATE) is a National Institute of Mental Health-funded multidisciplinary center that develops and evaluates LLM-based tools to support evidence-based mental health treatment implementation.

Access to evidence-based mental health care is very limited in public mental health contexts that grapple with severe financial constraints while treating a high volume of diverse individuals. Innovation is needed to promote the uptake and effective delivery of evidence-based care in such settings. Large language models (LLMs), a form of Generative Artificial Intelligence (AI), are promising tools to support this effort. They have the potential to provide scalable and just-in-time support to administrators seeking to implement effective treatments, therapists who use them, and to patients between sessions. However, given the unique risks and considerations for mental health intervention there is an urgent need to establish that essential criteria are met before such tools can be responsibly deployed and scaled in an equitable and ethical manner. Our team’s Readiness Evaluation for AI Deployment and Implementation (READI) framework posits that these interventions must be safe, private, engaging, and effective; developed and deployed with attention to equity and effective implementation.

CREATE’s Methods Core brings together experts in clinical psychology, implementation science, computer science (AI, human-computer interaction), cultural adaptation and culturally responsive trauma treatment, ethics, biostatistics, and economics. Across three exploratory projects and a signature project, we will develop and evaluate tools to support effective and culturally responsive EBP delivery across the care delivery continuum. CREATE’s resulting products and the pilot funding, resources, training and mentorship provided through our Administrative Core will foster advancement in the nascent field of LLM-based interventions to support EBPs.

CREATE’s projects include development and evaluation of:

• ImpleMentor. An interactive chatbot to facilitate development of an implementation plan and support teams as they implement EBPs in under-resourced treatment settings
• Therapy TrAIner: Simulation-based tools and conversational agents to support therapists as they learn new treatments after initial workshop training
• COACH-AI. Ongoing LLM-based consultation and support on specific cases and challenges
• Worksheet Helper. A tool to provide support for patients as they complete EBP homework between sessions (EP#4).

Additional planned projects include:

• Working with chatbot users and therapists to understand the best strategies for how chatbots could therapeutically interact with individuals at higher risk of suicide or self-harm as they are triaged to a higher level of care
• “Reimagining therapy”. With LLMs and other technologies likely to have greater and greater capacity to provide mental health interventions–what do individuals who seek care and the individuals who treat them believe are the right timing, sequence, frequency, and combination of interventions and tools?
• Chatbot-based coaching for digital mental health apps (research has shown that support or coaching from a therapist, peer, or non-licensed professional greatly increase engagement and effectiveness)

Stanford Team: Shannon Wiltsey Stirman and Johannes Eichstaedt (MPIs), Eric Kuhn, Jane Kim, Michael Bernstein, Betsy Stade, Pippa Kennard

Partnering Systems: STRONG STAR Training Initiative, Beck Community Initiative, VA National Center for PTSD

Contact: sws1@stanford.edu

Our neuroscience-informed clinical research lab is dedicated to advancing the understanding of the neurobiological foundations of atypical cognition and behavior in psychiatric disorders, with a focus on the most prevalent—autism, addiction, ADHD, and schizophrenia. Our research combines advanced computational methods, including artificial intelligence, with multimodal approaches such as human brain imaging, extensive phenotypic profiling, and molecular and cellular data analysis. By integrating these diverse and complementary data sources, we aim to uncover the complex, multilevel neurobiological mechanisms underlying psychiatric disorders from early development through adulthood.

At the forefront of developing innovative neuroscience-based AI-driven solutions, we strive to transform mental health care by enhancing the precision and accuracy of diagnostics, identifying novel therapeutic targets, and improving personalized treatment strategies. Our research holds promise for reshaping how psychiatric disorders are understood, diagnosed, and treated across the lifespan, ultimately leading to better outcomes and improved quality of life for those affected.

Our lab is interested in using AI applications to improve mental health and wellbeing. This includes employing AI to help in identifying and predicting mental health symptom onset and worsening to developing and optimizing technology-based interventions. We have ongoing projects that employ active and passive sensing of psychological, behavioral, and physiological data and will develop AI algorithms to understand how and when mental health symptoms are likely to disrupt functioning for a given individual. Our team has developed numerous, publicly available evidence-based mobile mental health apps so we are actively exploring ways in which AI can be used to enhance user engagement (e.g., by employing LLM chatbots), optimize use and effectiveness of self-management tools, facilitate psychotherapy homework completion, improve matching of user interests and what is most helpful for them, and deliver interventions with maximal benefit in the moment of need. Lastly, our lab is interested in using AI to improve precision treatment, which involves delivering the right treatment to the right patient at the right time. We are excited about these AI applications as we believe they have vast potential to make a tremendous impact on the mental wellbeing of the U.S. population and beyond.

Contact: Dr. Eric Kuhn (ekuhn@stanford.edu).

Unfortunately, there is, on average, a 17-year delay between the onset of OCD symptoms and treatment initiation. We aim to apply generative artificial intelligence (AI) to augment timely and accurate clinical diagnosis of OCD and related disorders. A more thorough understanding of the potential role of AI in automating aspects of mental health care is urgently needed, particularly as the demand increases for real-world implementation in clinical settings. One example is our study showing large language models (LLMs) outperform mental and medical health care professionals in identifying OCD. We will continue to rigorously study sensitivity, specificity, test-retest reliability, and other key needs before the application of untested technologies in clinical settings; particularly to ensure HIPAA compliance, awareness of potential biases or inaccuracies, and ensuring oversight of AI-driven clinical decision support.

We are a village of interdisciplinary team members and community partners who highly value inclusive team science, increasing representation in study participants and the scientific workforce, and advocacy and education about mental illness and clinical research. Our goal is to innovate targeted, rapid-acting treatments and to drive insights into the brain basis of OCD and related disorders to bring compassionate relief to the suffering of patients. Our mission is to lead cutting-edge, high-quality research studies that accelerate and transform treatments for severe mental illness.

If you would like to participate in a research study or join our team, we encourage you to reach out to ocdresearch@stanford.edu or call us at 650-497-2577.