LLM-based mobile teaching assistant and anxious-child OSCE performance in undergraduate paediatric dentistry: a cluster quasi-experimental study

In the early years of dental school, students spend countless hours mastering the mechanics of their craft. They learn how to drill, fill, and extract teeth, often practicing on plastic models or mannequins that never move, speak, or cry. Yet, the reality of treating children is far more complex than technical skill alone. A young patient who is frightened, in pain, or simply unwilling to sit still can derail even the most skilled procedure. Success in pediatric dentistry relies heavily on the ability to calm a terrified child, explain a scary situation to a worried parent, and guide a family through a stressful experience without causing further distress. Traditionally, students have had to learn these soft skills through trial and error once they begin seeing real patients, a high-stakes environment where mistakes can cause lasting trauma. To bridge the gap between the classroom and the clinic, educators are increasingly turning to digital tools that allow students to practice these difficult conversations safely and repeatedly.

Researchers at Wenzhou Medical University in China recently explored whether a specific type of artificial intelligence could help undergraduate dental students prepare for these emotional challenges. They focused on a tool built using a large language model, a type of computer program capable of holding complex, human-like conversations. The researchers created a mobile teaching assistant that students could use on their phones or computers. This digital partner was programmed to act as both a crying, fearful child and an anxious parent. It allowed students to practice calming a distressed child, explaining treatment options, and managing difficult behaviors. Crucially, the tool did not just chat; it provided feedback on the student's reasoning, pointing out where their approach might have been too harsh or too vague, and suggesting better ways to handle the interaction. The study aimed to see if students who practiced with this AI assistant would perform better when faced with a real-life simulation of a difficult pediatric patient compared to those who relied only on standard classroom teaching.

The study involved 128 fourth-year dental students who were about to begin their clinical rotations. These students were divided into two groups based on their existing class schedules. One group continued with the standard eight-week pediatric dentistry course, which included lectures, videos, and practice on plastic models. The other group received the same standard instruction but was also given access to the mobile AI assistant. These students were encouraged to spend about 15 minutes a day, three times a week, interacting with the virtual child and parent. They could practice scenarios ranging from a child screaming in the dental chair to a parent questioning the safety of a procedure. The AI would respond in character, reacting to the student's words with fear, anger, or relief, and then offer a detailed critique of the student's communication strategy.

At the end of the eight-week course, all students faced a rigorous test designed to measure their readiness for the clinic. They entered a room to see a standardized patient: an actor playing a five-year-old child with a toothache who was crying and afraid, accompanied by a worried parent. The students had 15 minutes to examine the child, explain what was wrong, and propose a treatment plan while managing the child's fear and the parent's anxiety. Their performance was scored by four senior faculty members who did not know which group the students belonged to. The scoring system was comprehensive, evaluating everything from how well the student built a connection with the child to how clearly they explained the risks and benefits of treatment.

The results showed a clear difference between the two groups. Students who had used the AI assistant scored significantly higher on the overall test than those who had not. The most dramatic improvement was seen in the area of behavior management. These students were better at calming the crying child, using a gentle tone, and keeping the child cooperative during the examination. While the standard course helped all students improve their confidence and knowledge over the semester, the group with the AI assistant demonstrated a distinct edge in handling the emotional and interactive parts of the visit. They were more effective at navigating the chaos of a frightened child and an anxious parent, turning a potentially chaotic situation into a manageable one.

The students who used the tool also reported feeling more confident in their abilities. They found the system easy to use and felt it helped them understand how to communicate better with young patients. Many noted that the ability to practice these difficult conversations repeatedly, without the pressure of a real patient, was invaluable. They appreciated the immediate feedback that told them exactly what they had done well and what they could improve. However, the researchers were careful to note that the benefits were specific to the skills the tool practiced. The AI did not teach them how to drill a tooth or perform a surgical procedure; it taught them how to talk to a child and a parent. The study found that while the course itself raised the confidence of all students, the extra practice with the AI specifically sharpened their ability to manage behavior and plan treatment in a way that the standard course alone did not.

Despite the positive results, the researchers emphasized that this was a single study conducted in one location with a specific type of test. The tool was designed to supplement, not replace, traditional teaching. The students still needed the lectures and the hands-on practice with plastic models to learn the technical side of dentistry. The AI served as a safe space to rehearse the human side of the job. The researchers also pointed out that they could not track exactly how much time each student spent with the tool, relying instead on the students' own reports. They acknowledged that the novelty of using a new technology might have played a role in the students' engagement. Nevertheless, the findings suggest that integrating artificial intelligence into dental education can provide a unique and effective way to train students for the emotional realities of treating children. It offers a way to practice the art of empathy and communication in a controlled environment, preparing students to be not just skilled technicians, but compassionate caregivers.