A recent study reveals that patients often provide less detailed descriptions of their symptoms to a digital assistant, which could harm the accuracy of automated diagnoses.
Medical chatbots and self-assessment tools are increasingly used as the first point of contact between the patient and the healthcare system. They help determine the urgency of a situation before a professional intervenes. However, this new research shows that the effectiveness of these tools does not depend solely on their computing power: it also relies on the quality of the information provided by users.
The study was conducted by the team of Professor Wilfried Kunde at the University of Würzburg and published in
Nature Health. It involved 500 participants, who were asked to write symptom reports for two common conditions: unusual headaches and flu-like symptoms. They were told that these reports would be analyzed either by an AI chatbot or by a human doctor.
The results are clear: the descriptions intended for the AI were less useful for medical evaluation than those written for a professional. Among other indicators, the reports for doctors averaged 255.6 characters, compared to 228.7 for the chatbots. Even though the difference seems small, it can have real consequences. AI systems, no matter how advanced, risk providing inaccurate advice if important details are missing.
A phenomenon called "uniqueness neglect" partly explains this withholding of information. Many people believe that AI cannot grasp the individual nuances of their situation and merely recognizes standard patterns. Concerns about privacy and skepticism toward algorithmic diagnoses also lead patients to provide vague or incomplete data.
The researchers emphasize that improving technology alone will not be enough. They recommend designing interfaces that encourage richer communication. For example, giving examples of detailed descriptions and actively asking for clarification when information is missing. These adjustments could reduce diagnostic errors and ease the pressure on healthcare systems.
User trust and perception play an important role in the performance of digital tools. For these technologies to fulfill their promises, it will be necessary not only to refine algorithms but also to better understand patient behavior.