Caution Urged in Distinguishing AI-Generated Medical Advice from Human Doctors in New York

Have you ever wondered if you could tell the difference between medical advice from a human doctor as opposed to advice from artificial intelligence? A recent study from New York University suggests that it could be more challenging than you think.

Get our free mobile app

According to the study by NYU's Tandon School of Engineering and Grossman School of Medicine, we are having difficulty telling the difference between the two. They say that chatbots, like ChatGPT, could be more helpful with healthcare providers when talking to their patients.

ChatGPT is part of the Generative Pre-trained Transformer (GPT) series and helps to find  the most likely next word in a conversation. It responds to the users question and learns from human feedback. But keep in mind that, ChatGPT has its limitations and sometimes it generates inaccurate information.

Researchers found nearly 400 participants over the age of 18. They were asked ten patient questions, with half of the responses from humans and the other half generated by ChatGPT. They had to tell if it was human or ChatGPT and rate their trust in the chatbot's answers  from "completely untrustworthy" to "completely trustworthy."

The participants correctly identified  the chatbot-generated responses 65.5 percent of the time, which is pretty much the same as the 65.1 percent accuracy for identifying human-generated responses.

Can You Trust Chatbot

The researchers noticed the level of trust in chatbot responses depended on the nature of the question. Participants showed the highest level of trust in chatbots when it came to handling things like making appointments or answering insurance questions, with an average trust score of 3.94.

Questions concerning vaccines or cancer screenings, was next with a trust score of 3.52. But when it came to diagnosing or treatment advice, the trust was lower with scores of 2.90 and 2.89, respectively.

What does this all mean? Well, the study highlights the potential of chatbots in helping patient-provider communications, especially in administrative matters or taking care of chronic diseases. But take note, the researchers of the study expressed caution when it comes to chatbots taking on more clinical types of roles.

"Providers should remain cautious and exercise critical judgment when curating chatbot-generated advice due to the limitations and potential biases of AI models," the researchers said in a statement.

ChatGPT and other chatbots may have their advantages, it's important to remember that they have their limitations too. In my opinion, when in doubt seek human help