Dartmouth will serve as a leading partner in a new artificial intelligence institute backed by the National Science Foundation, building on the university's pioneering research in developing digital therapeutics for treating addiction and behavioral and mental health disorders.
Dartmouth joins the multi-institutional AI Research Institute on Interaction for AI Assistants, or ARIA, as its center for "use-inspired" research. Faculty and researchers based in the Center for Technology and Behavioral Health in the Geisel School of Medicine, as well as the Department of Computer Science and the Program in Cognitive Science, will lead the implementation of AI into devices and wearable sensors that can provide users with personalized assessment and intervention.
Based at Brown University, ARIA is supported by a five-year, $20 million NSF grant and aims to develop next-generation AI assistants that can interpret a person's unique behavioral needs to provide safe and effective feedback in real time.
ARIA was announced July 29 as one of five National Artificial Intelligence Research Institutes—plus a hub to coordinate and expand the institutes' work—launched through a $100 million investment by the NSF. The institutes align with the White House AI Action Plan to ensure the United States' global leadership in AI development.
“Dartmouth is the birthplace of AI, so we are especially excited to bring our long-standing expertise to this national AI institute and work with an extraordinary group of partners across the nation,” says Lisa Marsch, the founding director of CTBH who will serve as ARIA's behavioral and mental health lead.
"The goal is to generate the science that will inform best practices for developing AI-powered agents that are capable of trustworthy, sensitive, and context-aware interactions with people about their mental health and substance-use-related needs," Marsch says.
In ARIA's first year, Dartmouth researchers will lead a project focused on identifying and adapting to the rapid shifts in physiology, behavior, and cognition in people with major depressive disorder that come before the onset of serious symptoms.
For the second year, the Dartmouth team will focus on substance use disorders, particularly opioids, by examining the complex interplay between physiological states, environmental triggers, and neural connectivity patterns that lead to drug use. These insights will be coupled with sensor, neurocognitive, and neuroimaging data to produce real-time interventions that can help prevent relapse and support long-term recovery.
"We will build on our established and ongoing work of integrating data from behavioral sensing, AI, and theoretical and cognitive neuroscience to understand the complex interplay between physiological states, environmental factors, and neural connectivity patterns that contribute to mental health and substance use behaviors," Marsch says.
Dartmouth faculty central to ARIA's work include:
- Lisa Marsch, the Andrew G. Wallace Professor of psychiatry and biomedical data science, who will ensure that practical use is the focus of all of ARIA's research, education, and outreach activities. She also will tap CTBH's extensive partnership network to coordinate ARIA's work with patients, mental health professionals, treatment facilities, and advocacy organizations.
- Andrew Campbell, the Albert Bradley 1915 Third Century Professor in computer science, who is the lead for ARIA's mental health testbed. Campbell, the director of CTBH's Emerging Technologies and Data Analytics Core, will oversee the integration of new technical and scientific ideas into the AI prototypes behind ARIA's virtual assistants.
- Nicholas Jacobson, an associate professor of biomedical data science, psychiatry, and computer science, who is a co-investigator for ARIA. Jacobson, directs the AI and Mental Health: Innovation in Technology Guided Healthcare (AIM HIGH) Lab in CTBH, and will build largely from his lab's development of Therabot, the first fully generative AI psychotherapy chatbot to undergo a clinical trial.
- Steven Frankland, an assistant professor in cognitive science and ARIA co-investigator, who will coordinate the basic cognitive science research and computational modeling conducted at Dartmouth. His Mind, Brain, and Computation Lab focuses on the computational mechanisms underlying the cognitive flexibility necessary to build more adaptive and sympathetic AI systems.
CTBH, which is the only Center of Excellence designated by the National Institute on Drug Abuse focused on digital health tools and therapeutics, is home to many of the projects that will inform ARIA, says Marsch, who led the development of the first FDA-authorized prescription digital intervention for addiction treatment.
She also led the first study, known as D-TECT, that used behavioral sensing to project the clinical trajectories of people in drug treatment and design more precise, personalized options for recovery support.
Campbell has led several projects demonstrating the ability of mobile and wearable tech to evaluate mental health and predict the onset of symptoms. The four-year StudentLife study, the longest of its kind ever conducted, showed how an app on students' phones could be used to measure and understand their mental health.
Campbell and Jacobson collaborated on an app called MoodCapture that paired AI with facial-recognition software to reliably detect the onset of depression using the front-facing camera of a person's phone. They also developed MindScape, the first system to combine behavioral sensing with ChatGPT to deliver personalized mental health interventions, and Time2Lang, a framework that makes mobile sensing data usable by the learning models that underpin AI.
"At Dartmouth, we're specifically focused on developing the foundational science to understand, predict, and improve mental health outcomes through real-world interaction across student and clinical populations nationwide," Campbell says. "It's clear that where we stand today, the science of AI assistants for mental and behavioral health is in its infancy, and Dartmouth's prior and ongoing research is well-positioned to help advance this emerging field as part of ARIA."
Jacobson's AIM HIGH group also works to create digital biomarkers for predicting and responding on an individual level to changes in mental health, which is essential for the next generation of AI systems, Jacobson says. His group also works from the user's perspective, including how people organically use models such as ChatGPT for mental health support and evaluating the clinical knowledge and biases of AI.
"This research directly informs our goals within ARIA," Jacobson says. "While our work shows the promise of generative AI, it also underscores the scientific work needed to ensure these systems are safe, reliable, and truly personalized."
The cognitive science arm of ARIA aims to bridge the gap between the strengths of large language models and those of human intelligence, Frankland says.
"It remains unclear whether the models we interact with daily learn and reason in ways that are most useful to humans," Frankland says. "By advancing our understanding of both human and machine cognition, we aim to build systems whose grasp of the world and their users is both more flexible and more reliable. Nowhere is this goal more critical than in the treatment of mental health disorders."
ARIA is led by Ellie Pavlick, an associate professor of computer science at Brown. Pavlick was keynote speaker for the AI and Student Mental Health Symposium sponsored by CTBH and held at Dartmouth on March 17.
In addition to Brown and Dartmouth, ARIA includes numerous leading research institutions: Colby College; New York University; Carnegie Mellon University; the University of California, Berkeley; the University of California, San Diego; the University of New Mexico; the Santa Fe Institute; and Data and Society, a civil society organization in New York. ARIA researchers will meet annually with a different partner hosting each year—Dartmouth is scheduled to host the institute's fifth meeting in 2030.
ARIA also will prioritize training the next generation of AI scientists through educational opportunities and curricula for middle and high school students and students of two- and four-year colleges. Partner institutions will be able to provide mentoring for undergraduate and graduate students, postdoctoral fellows, and visiting scholars who can work across multiple universities and organizations.
ARIA also is dedicated to engaging mental health professionals in workshops, roundtables, and college-level certificate programs intended to educate practitioners and implement ARIA's innovations into a clinical setting more efficiently.