At Seoul's Gangnam Severance Hospital, patients arriving for their first psychiatric appointment now have the chance to talk through their symptoms with an AI before they ever sit down with a doctor. Korean researchers have built a conversational system that listens, asks clarifying questions, and helps people organize their emotional distress in advance—making those intimidating first consultations feel less burdensome and more productive.
The research team, led by Professors Uichin Lee and Tak Yeon Lee at the Korea Advanced Institute of Science and Technology, alongside psychiatrist Eunjoo Kim, developed a large language model designed specifically to support initial psychiatric interviews. The system doesn't replace doctors; instead, it preps patients and their charts before the real conversation begins. This matters because psychiatric care has always faced a central tension: patients struggle to articulate months or years of emotional struggle in limited time, while doctors rush to capture extensive histories within brief consultation windows.
The AI conversational system adjusts its flow in real time, responding to what each patient says. It draws on specialized psychiatric knowledge to identify which questions matter most, then asks them naturally. Crucially, it goes beyond simple interrogation. The system applies genuine counseling techniques—expressing empathy, restating what patients say in organized language, and gently clarifying confusion. These touches help people open up more freely.
When researchers tested the system with 1,440 virtual patients, they found it reliably gathered essential clinical information within 30 minutes. The AI then generates a clinical dashboard summarizing symptoms and potential conditions, giving doctors a clear picture before the patient enters the room. This shift changes how doctors work: instead of frantically scribbling notes during intake, they can focus on deeper, more therapeutic conversation.
The research philosophy is crucial. Rather than positioning AI as a replacement for doctors, the team defines it as a "coachable apprentice"—a collaborator where AI handles the repetitive, structured work of information gathering while doctors retain full authority over diagnosis and treatment decisions. Professors Lee and Kim made clear that the system has real limits. It struggles with subtle emotional nuances and sensitive topics that require human judgment and presence. Final decisions always belong to trained psychiatrists.
Professor Uichin Lee captured the vision plainly: "If AI reduces the burden of the initial consultation stage, medical staff can focus more on deeper counseling with patients." For clinicians already stretched thin, that shift could reshape psychiatric care itself.
The research, presented in April 2026 at ACM CHI—a major conference on human-computer interaction—points toward a future where technology doesn't replace care but clears the path for it. By letting AI shoulder the structured, time-consuming work of history-taking, doctors gain space for what they do best: listening deeply and making wise clinical judgments. For patients, the benefit is equally real: a chance to gather their thoughts with a patient, responsive listener before facing what many find to be one of medicine's most daunting conversations.