Psychological frameworks help AI models to provide better health care advice

May 11, 2026 - 22:26

Researchers at Technische Universität Berlin have found that training large language models to simulate human intuition and logical reasoning can sharply improve the quality of medical guidance they offer. In a series of experiments, the team taught AI systems to apply two well-known psychological frameworks: one based on fast, intuitive thinking and another on slower, deliberate reasoning. The results showed that models using this dual-process approach gave more accurate and context-sensitive advice about when to seek medical care compared to standard AI systems.

The study addressed a common problem in AI-driven health tools: they often provide generic or overly cautious recommendations, such as telling every user to see a doctor immediately. By incorporating psychological principles, the researchers enabled the models to weigh symptoms, urgency, and patient history more like a human clinician would. For example, the intuitive framework helped the AI quickly flag obvious emergencies, while the reasoning framework allowed it to analyze ambiguous cases step by step.

This approach could lead to more reliable virtual health assistants and triage tools, reducing unnecessary visits to clinics while catching serious conditions earlier. The Berlin team emphasized that the goal is not to replace doctors but to make AI a safer, more useful first point of contact for patients. Further testing is planned to see if the method works across different languages and healthcare systems.