Not all people have the opportunity to visit a doctor or a hospital, so the emergence of medical chatbots has given hope for easy and constant access to qualified medical advice. This will certainly become possible in the future, but so far, research has shown that medical chatbots are more likely to confuse patients than help them solve their health problems.
A group of scientists from the Oxford Internet Institute research center at the University of Oxford in the UK reported on the unsatisfactory state of affairs with the medical qualifications of artificial intelligence. To do this, they provided 1,300 volunteers with scenarios of various diseases (symptoms) prepared by professional doctors and suggested that they independently find answers on the Internet, including communication with medical chatbots.
“The study revealed a breakdown in two-way communication,” the authors of the study said. “Those who used [chatbots] made decisions no better than participants who relied on traditional methods, such as online searches or their own judgment.”
Users used personal skills, reference information, and consultation with medical large language models such as ChatGPT, GPT-4o, as well as Command R+ by Cohere and Llama 3 by Meta. Subsequent analysis showed that communication with chatbots reduced the likelihood of correct diagnosis and led to underestimation of the severity of the health condition.
Miscommunication and Risk of Misdiagnosis
Scientists note that participants in the experiment often omitted key details when contacting chatbots or received answers that were difficult to interpret. In its current form, the researchers note, using medical AI from open sources is not only unnecessary, but also often harmful.
Self-medication should not be practiced, and diagnosis with the help of chatbots is self-medication in its purest form, reminds NIX Solutions. If health problems are detected, you should contact a human specialist. And if medical chatbots are ever allowed to make real medical diagnoses, this will be done only after clinical trials, as is happening today with the adoption of new drugs.
Medical AI tools still have a long way to go before they can replace or even support qualified healthcare professionals in diagnosing and treating patients. We’ll keep you updated as more integrations and advancements in medical AI become available.