Generative artificial intelligence in medical postgraduates: a qualitative study of perceived risks, benefits, and responsible use

Imagine the world of medicine as a massive, ancient library where every book contains a secret to healing. For centuries, doctors and students have had to read every single page by hand to find the right answer. But recently, a new kind of magical assistant has arrived: Generative Artificial Intelligence, or GenAI. Think of it as a super-fast, super-smart robot librarian who can read millions of books in a second, summarize them, and even write a story for you. It sounds like a dream come true, right? But here's the catch: this robot librarian sometimes makes things up, forgets to check its facts, and if you let it write your story for you, you might forget how to write your own.

This paper dives into a very specific group of people who are just starting to use this magical robot: medical graduate students in China. These are the young doctors in training who are learning to become experts. The researchers wanted to know: Do these students love the robot? Are they scared of it? And most importantly, how do they decide when to let the robot help and when to do the work themselves? They didn't just ask for a "yes" or "no" answer; they sat down and had deep conversations to understand the students' real thoughts, fears, and hopes about this new technology.

The Study: Listening to the Future Doctors

The researchers interviewed 18 first-year master's students from five different medical specialties, like neurology (brain doctors), psychiatry (mind doctors), and radiology (image doctors). All of these students had already tried using AI tools like ChatGPT or DeepSeek for their schoolwork. The team listened carefully to what they said, looking for patterns in their stories. They found that the students have a very clear, balanced view: they see GenAI as a "double-edged sword," meaning it has a sharp side that can cut through problems quickly, but also a dangerous side that can hurt if you aren't careful.

The Good Stuff: The Super-Powered Sidekick

The students were excited about how much time the robot saves them. They described it as a turbocharger for their brains.

• Speeding Up the Grind: One student said the AI helps them "quickly locate needed literature," while another noted it can "quickly analyze review articles." It's like having a friend who can find the best books in the library in seconds instead of hours.
• Seeing the Unseeable: The students were amazed that the AI could spot tiny details humans miss. One student explained that for "small stenoses [narrowings] that are hard to see with the naked eye, AI can directly calculate the stenosis rate." Another mentioned it could analyze huge piles of protein data to find disease markers. It's like giving the students X-ray vision for data.
• Brainstorming Buddy: The AI also helps them connect dots they hadn't seen before. One student said it can "analyze 4,000 medical articles... in 10 minutes," helping them come up with new ideas faster than they could alone.

The Scary Stuff: The Robot's Bad Habits

However, the students were not blindly trusting. In fact, their biggest worry was that they might get lazy.

• The "Brain Rust" Fear: This was the most common concern, mentioned by 16 out of 18 students. They feared that if they let the AI do all the thinking, their own brains would stop working hard. One student put it bluntly: "Using AI directly leads to a lack of one's own thinking." Another worried that "long-term use of AI can lead to mental laziness." They compared it to using a calculator so much that you forget how to do basic math.
• The "Fake News" Problem: The students didn't trust the AI to be 100% truthful. They noted that the AI tends to "make things up" or "hallucinate." One student complained that "most of the literature sources shown by AI are fake." It's like a librarian who confidently points you to a book that doesn't actually exist.
• Privacy and Blame: They were also worried about patient secrets. If they type patient data into the AI, they fear the data could leak, leading to discrimination or privacy breaches. Furthermore, they were confused about who is responsible if the AI gives a wrong answer. As one student said, "AI cannot take responsibility – the developer or the user institution must endorse it." If the robot doctor makes a mistake, who gets in trouble? The human? The machine? The answer is currently unclear.

The Rules: How to Play Fair

Despite the fears, the students did not want to ban the robot. They used a Chinese proverb to explain their stance: "We shouldn't stop eating for fear of choking." Instead, they drew a very clear line in the sand about what is okay and what is not.

• Green Light (Okay to use): They are happy to use AI for polishing their English, checking their computer code, or generating a first draft of a paper. They see this as using a tool to help them work faster.
• Red Light (Never use): They are firm that AI should never generate a patient's diagnosis, handle private patient data, or create the original idea for a research study. One student stated, "I will use AI to generate a first draft, but I will definitely revise and verify it manually." They want to be the captains of the ship, with the AI as the navigator, not the captain.

What They Want: The Perfect AI Assistant

Finally, the students told the researchers exactly what they would need to feel safe using these tools. They don't want a "magic black box" that gives answers without explanation. Instead, they want an AI that:

1. Knows the Rules: It must base its answers on real, high-quality medical evidence and guidelines, not just random guesses.
2. Protects Secrets: It must be as secure as a bank vault, following strict privacy laws (like HIPAA or GDPR) to keep patient data safe.
3. Explains Its Work: It shouldn't just give an answer; it should explain why it gave that answer, so the student can understand the logic.
4. Stays Humble: It must clearly position itself as an "assistant" to the doctor, never a replacement.

The Bottom Line

The study suggests that these future doctors are not just passive users of technology; they are thinking critically about how to use it. They see the huge benefits of AI for efficiency and discovery, but they are deeply concerned about losing their own ability to think creatively and critically. They aren't asking to throw the robot away; they are asking for a robot that is honest, safe, and clearly defined as a helper, not a boss. Their voices suggest that for AI to truly help in medicine, we need to build tools that respect human judgment and keep the human in the driver's seat.