It is often said that seeking psychiatric treatment can be scary. It may feel burdensome for patients to confide in their concerns for the first time, but medical staff must accurately understand a patient’s vast medical history and symptoms within the limited consultation time. South Korean researchers have developed artificial intelligence (AI) technology to support the initial psychiatric interview process, the first step in psychiatric care.
KAIST (Chairman Lee Kwang Hyun) announced on May 24 that a joint research team led by Professor Lee Wi-chin of the School of Computing and Professor Lee Tak-young of the Department of Industrial Design, and a team led by Professor Kim Eun-joo of the Department of Psychiatry at Gangnam Severance Hospital (Director Kim Yong-wook) have developed a large-scale language model (LLM)-based technology to support early symptoms in psychiatry. Interview.
This research was conducted in a way that allows patients to sort out their symptoms and conditions in advance by having a conversation with the AI before seeing their doctor.
The research team designed the system so that the AI could adjust the flow of the conversation depending on the patient’s reactions. AI analyzes patient responses in real time against expert medical knowledge from psychiatry to generate important questions to ask next. In particular, this system goes beyond simple Q&A and applies actual counseling techniques, such as expressing empathy, organizing and rephrasing the patient’s words, and clarifying ambiguous content. This is intended to make patients more comfortable talking about their condition.
Experiments with 1,440 virtual patients to verify performance confirmed that in most cases, the system was able to effectively capture the critical clinical information needed for treatment within just 30 minutes.
Based on the conversations it collects, AI generates and provides clinical dashboards to medical staff that show symptoms and underlying conditions at a glance. This allows doctors to understand the patient’s condition more systematically before the patient enters the exam room, and allows them to focus on deeper counseling with the patient during the actual consultation.
The core of this study is that it defines AI not as a replacement for doctors, but as a “teachable apprentice.” It is a collaborative model where AI processes iterative and structured information collection, and doctors make the final diagnosis and prescription based on that information.
The researchers found that AI still has limitations when it comes to understanding subtle emotional changes and dealing with sensitive topics, and stressed that final decisions should always be made by trained medical professionals.
Professor Lee Wi-Ching said:If AI reduces the burden of the initial consultation stage, medical professionals will be able to focus on deeper counseling with patients.” he added.This shows the possibility of developing a new medical model in which humans and AI collaborate in medical settings.”
The study, led by doctoral student Yoo-kyung Jeong, was presented on April 13 at ACM CHI 2026 (ACM Conference on Human Factors in Computing Systems), the most prestigious conference in the field of human-computer interaction.
sauce:
KAIST (Korea Advanced Institute of Science and Technology)
Reference magazines:
Young, Y. others. (2026). Toward flexible psychiatric history capture and visualization: Exploring clinician perspectives using large-scale language models. Proceedings of the 2026 CHI Conference on Human Factors in Computing Systems. DOI: 10.1145/3772318.3790970. https://dl.acm.org/doi/10.1145/3772318.3790970
