AMA encourages thoughtful development of 'augmented intelligence'

The American Medical Association passed its first policy on so-called "augmented intelligence," encouraging the development of augmented intelligence tools that are free of bias and improve patient outcomes and physician satisfaction.

"As technology continues to advance and evolve, we have a unique opportunity to ensure that augmented intelligence is used to benefit patients, physicians and the broad healthcare community," AMA board member Dr. Jesse Ehrenfeld said in a statement.

By referring to the technology as "augmented intelligence" rather than "artificial intelligence," the AMA reinforced its position that the technology must be combined with a human healthcare provider.

"The reason to call 'artificial intelligence' 'augmented intelligence' is to reassure clinicians that technology shouldn't be taking an authoritative role over the practice of medicine," said Susan Etlinger, an analyst with the Altimeter Group. "That's a really important first step and philosophy about the role of technology in the practice of medicine."

To make sure that augmented intelligence benefits both providers and patients, AMA members will consider how to best design, evaluate and implement augmented intelligence. They'll try to involve physicians in the process.

"I agree with the AMA that the opinion of practicing physicians is needed, but I would extend that to other groups, including health systems and other providers," said Lidia Fonseca, senior vice president and chief information officer of Quest Diagnostics.

AMA members will also promote transparent and unbiased algorithms look into related liability issues.

"There are a lot of issues around AI that we need to learn about very much the way we needed to learn how the internet works," Etlinger said.

This year may be a turning point for AI as chatbots, machine-learning algorithms for imaging, and other tools begin to take off, she said.

In April, the Food and Drug Administration for the first time approved marketing a medical device that uses AI and doesn't require clinicians to interpret the input. The technology, IDx-DR, is made by IDx to detect retinopathy.

Hospitals already use other forms of artificial intelligence in limited settings. Ochsner Health System physicians are applying a machine-learning model to patient data to predict adverse events.

No AI-based tool in practice has yet obviated completely the need for clinician involvement.

As such, user experience matters—and AMA members have taken note. Pointing to physicians' frustrations with electronic health records and their usability, members of the AMA stressed the need for user-centered design in augmented intelligence. User testing should involve people who are "demographically representative" of those who will ultimately use the tools, according to the group.

Just as medical students now receive EHR training, clinicians in the future will need to be trained in augmented intelligence, according to the AMA.