How to—and how not to—consult with ‘Dr. ChatGPT’ and its peers
Do: Use large-language AI to help analyze your specific symptoms. Don’t: Ask a chatbot if you should go to the emergency room.
The pointers are from Ashley Merryman, JD, an attorney, author, speaker and performance coach who specializes in popularizing scientific and technical advances.
She regularly contributes content to U.S. News & World Report, which published her Do’s and Don’ts for dealing with “Dr. ChatGPT”—and, by extension, other medically precocious LLMs—May 11.
The tips are aimed squarely at healthcare consumers, not healthcare workers, but as such they’re readily shareable with friends and family.
Besides, we’re all healthcare consumers every now and again.
Here’s more from Merryman’s list, which leans heavily on the thinking of famous Eric Topol, MD, crediting him as it goes.
DO use AI to prepare for a scheduled doctor’s appointment.
Since physicians have only so much time to meet with patients, reviewing health issues with an AI chatbot beforehand may help you make better use of that time, Merryman suggests.
“Use it to clarify concerns you have or prepare a list of questions to ask—something especially useful if it’s a topic you find difficult to talk about,” she adds.
DON’T use AI tools to diagnose your overall health or fitness.
Don’t depend on it for an overall health assessment, Merryman advises.
“For individuals, the results are too inconsistent,” she clarifies. “You can even get different outcomes when inputting the identical information in the same AI model.”
DO ask it to explain lab tests, identify drug side effects and more.
Ask AI to explain the need for a lab test that your doctor ordered, Merryman writes.
“Have AI review an inventory of your medications to identify potential negative drug interactions. Ask how treatments or side effects may impact your participation in favorite hobbies or your attendance at social events.”
DON’T consider an AI diagnosis as the final authority.
In a study conducted with Microsoft, Topol’s team found that AI chatbot diagnoses can be “brittle,” thrown off by even a small change in a prompt, Merryman points out.
“Instead of correcting errors, the chatbots often came up with compelling (but false) reasons to defend their flawed diagnoses. So think of a chatbot analysis as a starting point and not a final diagnosis.”
Dave P. has worked in journalism, marketing and public relations for more than 30 years, frequently concentrating on hospitals, healthcare technology and Catholic communications. He has also specialized in fundraising communications, ghostwriting for CEOs of local, national and global charities, nonprofits and foundations.