Nature: AI Effectively Motivates Patients to Get Vaccinated

Researchers in the United States conducted a large-scale experiment involving more than 90,000 patients at healthcare facilities to test whether artificial intelligence could persuade people to change their habits through personalized messaging. They found that informing people that they were at high risk of contracting influenza encouraged them to get vaccinated more actively. At the same time, it made virtually no difference to people who had provided the assessment — a machine-learning algorithm or a human doctor. The study was published in Nature Human Behaviour.

As part of three clinical trials, the researchers used a previously validated machine-learning model capable of accurately identifying people most vulnerable to the virus and its complications. These patients were sent text messages encouraging them to get vaccinated. Some notifications explicitly stated that the risk had been calculated by a computer, others included detailed explanations of the AI prediction, while a third group received a simple warning about the high risk of illness without any mention of modern technology.

The results were highly revealing and demonstrated the value of an individualized approach. People who received a notification informing them of their vulnerability were 1.1–1.4 percentage points more likely to get vaccinated than those who received a standard reminder about the need for vaccination. Compared with the control group, which received no messages at all, vaccination rates increased by 1.7–3.5 percentage points.

The study’s most unexpected finding concerned people’s attitudes toward the involvement of algorithms in their healthcare. The effectiveness of the messages did not change depending on whether artificial intelligence was mentioned. Patients showed neither rejection of the machine-generated prediction nor particular enthusiasm for detailed explanations of it. For them, the key factor was awareness of a personal health threat, rather than the source providing that information.

Laboratory-based public opinion surveys often produce mixed results, with many people expressing concerns about the use of AI in medicine. However, real-world data showed that in practice, patients are willing to trust automated systems when they provide a clear benefit.