When is AI preferred for health advice? Public preferences for AI versus physician advice vary by use case beyond general attitudes
A study of 1,799 UK participants reveals that public preference for AI versus physician health advice is not determined by general attitudes toward AI, but varies significantly depending on the specific health use case and whether the advice is sought as a first or second opinion.
Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer
Imagine you're standing at a crossroads in a giant, bustling library of human knowledge. On one side, you have the wise, experienced librarians (doctors) who have spent years studying books, diagnosing problems, and holding hands with patients. On the other side, you have a super-fast, super-smart robot assistant (Artificial Intelligence) that can read every book in the library in a split second. For a long time, people have been asking: "Who should I trust?" Is the robot a magic wand that knows everything, or is it just a fancy calculator that might get the details wrong? This question sits at the heart of a field called "health decision-making," where researchers try to figure out how people choose who to listen to when they feel sick or worried. The big mystery isn't just whether people like robots or humans; it's about when they pick one over the other. Do we want the robot when we just need a quick fact, but the human when we need a hug and a deep diagnosis?
A new study from researchers at Technische Universität Berlin and the University of Milano-Bicocca decided to stop guessing and start testing. They didn't just ask people, "Do you trust AI?" Instead, they set up a massive experiment with nearly 1,800 people from the UK. They gave each person a specific, real-life health scenario—like "I have a weird rash" or "Is this medication safe?"—and asked them to choose: Would you rather get your first answer from a doctor or an AI? Then, they asked who they'd want to double-check that answer. The researchers wanted to see if people's choices changed depending on the type of problem, or if they just had a blanket rule like "I hate robots" or "I love robots."
Here is what they found, and it's a bit more nuanced than a simple "yes" or "no."
First, the big picture: When it comes to getting a first opinion on a health issue, most people (about 83.2%) still want a human doctor. They want the real thing. However, the researchers discovered that this preference isn't a solid wall; it's more like a sliding scale that changes depending on the job the AI is being asked to do.
Think of it like hiring help for different chores. If you need someone to move a heavy couch (a high-stakes, complex medical task like deciding whether to go to the emergency room for acute symptoms), almost nobody wants a robot. In these cases, only 4.0% of people chose AI. But if you need someone to look up a specific fact, like "Is this infection risk real after a trip?" or "Is this vitamin safe to take?", people are much more open to the robot. For these "information retrieval" tasks, the preference for AI jumped up to 28.5% for infection risks and 26.4% for medication safety.
The study suggests that people aren't just being "anti-robot" or "pro-robot." Instead, they are acting like smart managers. They seem to match the tool to the task. If the task feels like a simple information search (like checking a fact), the AI looks like a good fit. If the task feels like it needs a human touch, a physical exam, or a complex judgment call (like dealing with unexplained body changes or deciding on care), people stick with the doctor.
The researchers also looked at "second opinions"—what happens when you want to double-check the first answer. Here, the most popular choice (49.0%) was a doctor checking a doctor's work. But the second most popular choice (34.4%) was an AI checking a doctor's work. Interestingly, getting two opinions from AI was the least popular option, chosen by less than 1% of people. This suggests that people see AI as a useful sidekick or a fact-checker, but rarely as a full replacement for a human doctor, even for a second look.
One of the coolest parts of this study is that these preferences didn't disappear even when the researchers accounted for how much people generally liked or disliked AI. Even if someone said, "I love AI in general," they still preferred a doctor for serious, ambiguous health problems. This means the decision isn't just about a general attitude; it's about the specific situation.
So, what does this all mean? The paper suggests that we shouldn't think of AI as a single character that is either "trusted" or "distrusted." Instead, people are making fine-tuned judgments. They are willing to let AI handle the "data crunching" parts of health advice, like checking medication safety or explaining a symptom, but they want a human in the loop for the messy, scary, or ambiguous parts of being sick. The study doesn't prove that AI is safe or unsafe for these tasks; it just tells us what people want to use. It suggests that for AI to be truly helpful in healthcare, it might need to be designed not as a replacement for doctors, but as a specialized tool that works best when it's part of a team where a doctor is still the captain.
Drowning in papers in your field?
Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.