Our Director, Dr. Christopher Shelton, and visiting researcher Jumin Seo spent a day at the University of Michigan’s AI & Mental Health Conference, hosted by the Michigan Institute for Data and AI in Society (MIDAS).
Much of the day stayed focused on the harder questions around actually using AI in mental health. The question wasn’t just whether these systems can do something impressive, but whether they help in the real world, where they fit alongside human care, what happens when they get things wrong, and how to protect the people using them.
During the Community Connection session, Dr. Shelton and Jumin brought several of our current projects and used them as starting points for conversations with clinicians, researchers, people working in AI and policy, and community members with very different perspectives on how these technologies should be used.
Privacy came up again and again. We’re especially interested in what becomes possible as smaller AI models become capable enough to run directly on consumer devices, rather than sending sensitive health information to the cloud. That shift could matter for more than privacy. Local models can also reduce dependence on cloud infrastructure and the recurring costs that come with it, which opens up new possibilities for building health tools that are both useful and much easier to make broadly accessible.
The team left with new questions, several potential collaborations, and a lot to keep working through.
Jumin Seo is spending the fall 2026 semester with us at Penn State Behrend as a visiting researcher from Sungkyunkwan University’s (SKKU) LAMDA Lab.
