May 11, 2023 -- Want to know what to expect at your upcoming colonoscopy? Ask ChatGPT.
A new study published in the journal Gastroenterology shows that the artificial intelligence (AI) chatbot generates easily understandable, scientifically adequate, and generally satisfactory answers to common questions people may have about colonoscopy.
Developed by OpenAI, ChatGPT is a natural language processing tool that allows users to have personalized conversations with an AI bot capable of providing detailed responses to any question posed.
It has already seen several potential applications across multiple fields, including medicine.
In a first-of-its-kind study, researchers assessed the quality of ChatGPT-generated answers to eight common patient questions about colonoscopy.
The questions focused on what a colonoscopy involves, why it’s performed, how to prepare for it, potential complications, what to expect after the procedures, and what happens when the test is positive or negative.
The researchers selected the questions (with human answers) from the websites of three top U.S. hospitals for gastroenterology and GI surgery and had ChatGPT answer the questions twice.
Four gastroenterologists rated 36 pairs of common questions and answers based on ease of understanding, scientific adequacy, and satisfaction with the answer.
Overall, the doctors rated the ChatGPT answers highly and on par with non-AI answers for all three quality indicators. In fact, some of the ChatGPT answers were thought to be better than the human-generated answers, Tsung-Chun Lee, PhD, visiting staff with Taipei Medical University in Taiwan, and colleagues found.
Interestingly, the doctors could only correctly identify AI-generated answers 48% of the time.
Although it’s still early days, the researchers say ChatGPT and other AI chatbots may represent a “transformative innovation” in how medical information is created by doctors and consumed by patients.
It could also be a huge time saver for health care providers.
Screening colonoscopies are often the subject of patient questions. Looking ahead, the researchers say future research should explore a broader sample of questions and include patient raters.
