Dr. Ezekiel Emmanuel, from the Department of Medical Ethics and Health Policy at the University of Pennsylvania, and others, just published an article in the Journal of the American Medical Association (JAMA) entitled: Will Autonomous AI Exceed AI-Aided Physicians as the Best Medical Care? The article, relating studies done on simulated patient encounters, answers with a resounding YES! Dr. Emmanuel is part of a distinguished family, whose brothers are Rahm Emmanuel, former mayor of Chicago and potential Democratic party candidate for president in 2028, and Ari Emmanuel, Hollywood mogul. He is a bit of a gadfly. He has famously said that people over the age of 75 should not receive active medical treatment. He is almost 69 years old. The clock is ticking, Zeke.
The article shows data that purports to show the superiority of AI treatment in cognitive care. It also argues that using a hybrid human-AI model actually shows poorer outcomes than with AI alone.
Here are some of my beefs with this article:
The data revolves around using simulated patients only.
The comparator group to AI are “physicians without access to colleagues, textbooks, or the internet.” In other words, not a real life clinical decision making process.
One study quoted used a grand total of 16 patients to evaluate the differences between AI and humans for insulin management in diabetes.
The authors do point out that, as a result of dependence on AI, medical professionals will experience “deskilling”. Is that a good thing?
The conflict of interest disclosures reveal that the authors are chock full of financial remunerations from IT companies.
I would love to see a study using real life patients. I am going out on a limb here, but I think patients would rather see a trusted physician with whom they have developed a long term relationship, who can lay hands on them, and actively empathize with them, rather than a machine. Patients are less likely to cry to a machine.
I would also like to see how an AI gets information from a patient who may not trust them, who may lie or say half truths, or who will withhold information out of embarrassment. A patient I saw today was confused about a specialist’s change of medications and decided to stop all of them. How would AI handle that?
Of course, like all articles about AI, authors write that systems are only getting better as a way of discounting errors, slop, and outright hallucinations.
Then there is the issue of responsibility. AI’s will not get sued. Their human “handlers” will.
Will studies also show that all human interactions are inferior to human-AI interfaces? Why have friends, lovers ?
I am not a Luddite. I use AI frequently. What I don’t like in this piece is the way it insinuates that physicians using AI as an adjunct somehow will lead to inferior patient outcomes. Further, better designed studies are needed. Preferably ones where the humans doing the studies are not in the back pocket of AI industries.
References:
https://jamanetwork.com/journals/jama/fullarticle/2852952#google_vignette
https://www.ezekielemanuel.com/news-articles/blog-post-title-two-mt7l6
No posts
Comments
Nothing yet. Say the first thing.
Sign in to join the conversation.