ChatGPT tends to make overly cautious recommendations when it comes to health complaints
ChatGPT models tend to make overcautious recommendations when classifying health complaints. This is the result of a study by the Technical University of Berlin. The models are currently only suitable to a limited extent for a digital initial assessment and independent patient management.
The researchers tested 22 different ChatGPT model versions on the basis of 45 real patient cases. Each case was entered ten times per model, resulting in a total of 9,900 individual ratings. The models were to decide whether a case should be classified as an emergency, a case for medical clarification or a case for self-care.
Accuracy initially increased significantly with the early model generations, but has stagnated at a maximum of 74 percent since the third generation. A particularly large number of errors occurred in harmless complaints for which self-sufficiency would have been sufficient. 70 percent of all errors were attributable to this group. Most models consistently advised medical clarification, even in cases that did not require medical treatment.
In addition, the models showed some significant inconsistencies in their recommendations for identical inputs.
The study has been published in the journal Communications Medicine .
Original Paper:
Editor: X-Press Journalistenbüro GbR
Gender Notice. The personal designations used in this text always refer equally to female, male and diverse persons. Double/triple naming and gendered designations are used for better readability ected.
Most read
Wal Timmy: Environment Minister Backhaus ignored rescue technology that has been in place since 1983 The stranded humpback whale "Timmy" suffered severely for days in Mecklenburg-Western Pomerania beca...- Wal Timmy: Germany’s problem with euthanasia EDITORIAL. Letting Timmy suffer is more due to the German lack of experience and the bureaucratic hu...
- Impending fuel shortage: “If the laboratory fails, medical care will come to a standstill in large parts” DGKL CEO Jan Wolter explains the dramatic consequences of a potential fuel shortage for laboratory m...
