Doctors are less empathetic towards patients than artificial intelligence, an experiment showed – Human – Science and technology

Artificial intelligence treats patients better than human doctors, smaller research shows. The ChatGPT bot not only provided better quality information, but was even more empathetic in its responses.

Artificial intelligence (AI) could help in the preparation of conversations between doctors and patients – that is, before the actual communication takes place, the machine would get answers to important but essentially routine questions from the client. Unlike a person, he will never be under stress, he will not be sleepless, and it will not even happen to him that the patient is unsympathetic to him, for example. “The possibilities for improving health care with the help of artificial intelligence are enormous,” said Dr. John Ayers of the University of California, San Diego.

The results of the research she reported on news television Thursday 24 they do not yet mean that ChatGPT is a better doctor than a human. Experts warn against machines gaining responsibility for human health, if only because AIs very often simply make up answers or provide completely false information. But if they were primarily just asking, this risk would not be so high.

Read also Artificial intelligence will enable non-invasive reading of thoughts, converting them to text

Study published in journal JAMA Internal Medicine used data from the AskDocs forum on the Reddit social network. It is a place where members of this community can ask medical questions, which are then answered by certified health professionals. The team randomly selected 195 exchanges from AskDocs where a doctor answered a public question. The original questions were then posed to the ChatGPT artificial intelligence language model. How well the robot and the person answered was then evaluated by a committee of three professional doctors – they had no idea whether the answer came from a human doctor or a machine. They graded the answers in three categories – according to their quality and according to the degree of empathy that was in them.

In total, in 79 percent of cases, the commission gave preference to artificial intelligence answers over human answers. ChatGPT responses were also rated as good or very good 79 percent of the time, compared to 22 percent of physician responses, and 45 percent of ChatGPT responses were rated as empathic or very empathetic—compared to just 5 percent of physician responses.

Read also ChatGPT provides therapy, can lead to a revolution in mental health

Christopher Longhurst of the University of California, San Diego summarized the results: “They suggest that tools like ChatGPT can effectively deliver high-quality, personalized medical advice.”

Professor James Davenport of the University of Bath, who was not involved in the research, added to the British newspaper Guardian: “The article does not say that ChatGPT can replace the doctor, but it quite rightly calls for further research into whether and how ChatGPT can help doctors generate answers.”