When the Algorithm Meets the Counselor

Somewhere in a psychiatric hospital in Honduras, there is a counselor with a patient in crisis. The counselor has training, commitment, and the best intentions. What they don't have is easy access to the 900-page textbook sitting in an academic library thousands of miles away — the one with exactly the guidance they need, right now.

John Piette wants to change that.

Piette, a Professor of Global Public Health and Research Scientist affiliated with the Ann Arbor VA Center for Clinical Management Research, has spent years working at the intersection of behavioral health and technology in Latin America. His current work, anchored in deep partnerships with clinics, NGOs, and psychiatric hospitals across Honduras, is trying to solve a problem that's both simple and enormous: we know what good mental health care looks like, but too much of that knowledge sits behind paywalls and cost barriers, inaccessible to the providers and systems working to deliver it.

Motivational interviewing and cognitive behavioral therapy are two of the most widely studied approaches to behavior change and mental health treatment, and both have important benefits for people struggling with common disorders, including depression, substance use, and anxiety. After hundreds of rigorous controlled trials, dozens of meta-analyses, and the experience of tens of thousands of patients globally, the science is settled. Despite that, after decades of evidence generation, most practitioners delivering community mental health care in under-resourced settings lack the resources they need to deliver high-quality services.

Piette and his collaborators are developing a suite of AI tools designed to bridge that gap without bypassing the people on the ground. One platform teaches counselors reflective listening skills through interactive practice they can do privately, on their phones. Another program uses AI to score clinical conversations in real time against established fidelity metrics, giving health systems an objective window into care quality across different facilities. A third, built on retrieval-augmented generation (RAG), unlocks decades of buried clinical knowledge and makes it queryable in the moment — in both English and Spanish — for a counselor who needs guidance right now, not a library trip.

The technology is sophisticated. But Piette is quick to point out that it only works because of who's at the table. Psychologists in community-based organizations across Honduras. Computational linguists from the National Institute for Astrophysics, Electronics, and Optics in Mexico. Government health systems and international aid organizations connected through the Red Cross. "A strength of this work," he says, "is that we're connected with community organizations on the ground."

He calls himself an AI pragmatist. The cynicism about AI in global health, he argues, often comes from people who aren't facing the very real, very large barriers to delivering care in communities around the world. His partners are. And they're hungry for tools that actually help.