Skip to main content Skip to Footer


December 15, 2016
By: Kaveh Safavi

How artificial intelligence will augment doctors, not replace them

One question I get asked by healthcare colleagues is: What will become of doctors when the machines take over? It’s not as crazy-sounding a question as it used to be.

Recently there has been attention around supercomputers that identify specific treatment approaches for cancer. These computers find options that elude even experienced oncologists, for whom time and memory are more finite propositions. These technologies can also compensate for a physician’s cognitive biases, such as the “recency bias,” where you are more likely to recall the last case you saw to guide the next patient’s treatment.

There’s enormous excitement around this concept. There is also concern. Could machines replace doctors?

My short answer is no. Too many things patients want from doctors go beyond the realm of processing data.

As a former clinician, I know first-hand the critical role empathy plays in treating serious illness. Often, I found myself caring for the patient and their loved ones. Part of medicine is also just being present in the patient’s journey when science has little to offer. Understanding who patients are, their preferences, their experiences, their hopes and fears, are all critical components behind effective therapy.

I don’t think we are anywhere near the point where artificial intelligence can replicate that. But there are places where advances in cognitive computing are making a huge difference already.

Let’s go back to cancer, which is not a single disease but a multitude of diseases that may appear the same but are biologically distinct. In oncology, scientific discovery is proceeding rapidly, and doctors need help processing it to find what they need to know. Artificial intelligence can bring greater precision to medicine by identifying treatments the doctor may have missed. It’s still up to the doctor to decide whether a treatment makes sense for a particular patient.

Machines can present options. They can create better outcomes and save us money and time. Sometimes they can even treat a patient who wants care but is too embarrassed to talk to a live person about their concerns. What they can’t do is make choices people are willing to trust with their lives.

As far as another big fear goes, that machines could displace the economic position of doctors, I actually see cognitive computing having the opposite effect. The more routine tasks are automated, the more time the doctor can spend on those services society values most. I do have questions around the strain this could produce on clinicians operating at peak efficiency for full eight-hour days, so the nature of work will likely change as these tools become widespread. On balance, though, the possibilities for better care and a more economically sustainable care system are profound.

What is your thinking around the role of cognitive computing in healthcare? I would like to hear from you.

More blogs on this topic