AI replacing doctors. What do doctors think about that?

Last week, I was teaching health IT to clinicians in Mthatha, South Africa. Interventional cardiologists. Pulmonologists. Psychiatrists. Family physicians. Surgeons. Orthopedic doctors. More.

As senior specialists - sometimes the only ones in town - their perspective of clinical care was deep. Yet they enthusiastically wore hats of students and had a fun time.

Mthatha happens to be Nelson Mandela’s hometown. A beautiful semi-urban town in the Eastern Cape province. Not too far from the Indian ocean. Locals speak the Xhosa language, x-clicking their tongues.

On the final day of class, we discussed artificial intelligence. We saw what IBM’s been up to with Watson (they are now the lead sponsors at many healthcare conferences!). What direction Amazon could take with Alexa. And others.

Will AI replace doctors?

Naturally, the discussion gravitated towards artificial intelligence replacing doctors. I didn't prompt this. They asked it themselves. As with other industries, there was a sense of threat from technology. Will doctors lose their jobs?

IBM Watson has been training under oncologists at Memorial Sloan Kettering for a long time. 300 medical journals. 200 textbooks.

At a Singularity University executive program, I recall asking the person leading the Watson project (at that time) if they would open up their APIs. They soon did. Surely expecting a burst of adoption. However, I don't yet find many Watson-enabled applications that the average healthcare client will buy.

During the same program, I also recall Vinod Khosla say that 80% of doctors will be replaced by technology. Though I don't hear much of that now.

And then I've heard this somewhere: If a doctor can be replaced by AI, he must.

All that was 5 years ago.

Today, this is how Watson for Oncology pitches itself (on its website):

"Spend less time searching literature and the EMR, and more time caring for patients. Watson can provide clinicians with evidence-based treatment options based on expert training by Memorial Sloan Kettering (MSK) physicians."

The AI tune now is more that of assisting doctors than replacing them.

Even if the industry's goal is a machine that can see patients all day, everyday, without sleeping or eating, and without making any diagnostic errors.

So I gave this information to the doctors. Firing up an active debate.

Some of them shook their heads saying, "No, no, no, no, no...this is so wrong."

Some other doctors said, "Oh well, we can't stop the sun from rising. So we got to do what we must."

FOR or AGAINST AI?

Then we played a game.

We divided the group into FOR and AGAINST artificial intelligence. 

And then we did something interesting. We swapped roles. Compelling doctors to think the opposite of what they believed in.

Doctors who were AGAINST AI: They role-played as a large multinational vendor selling a healthcare AI platform to the local hospital. Now they had to think positively about something that they were originally against.

Doctors who were FOR AI: They role-played as the hospital board and doctors’ union. Evaluating and resisting a complex AI implementation proposed by the vendor. Now they had to think against what they thought was a good thing.

Swapping their mindset allowed them to see things differently. From the other side.

Over the next 40 minutes, there was an amusing, fierce debate among clinicians. Nervousness. Excitement. Confusion. And reconnection back to present-day reality. 

What were the takeaways?

From the debate, emerged a few takeaways.

1. There’s no stopping the sun from rising. People inherently got this. That whether they liked it or not, artificial intelligence will play a strong role in healthcare. Because many needs continue to go un-addressed.
2. Local issues are very pertinent. Different from global needs. The “hospital board” kept questioning the “vendor” on local issues. Does your system speak Xhosa? It won't understand our patients. Your system won't understand the unique combination of South Africa's issues. Of unemployment, poverty, lack of education and infectious disease like TB and HIV.
3. Garbage in, garbage out. Most of the data generated by healthcare systems is garbage. Templated medical records. Unusable lab results. Excess imaging. Even if an AI were to read and interpret them, it's dangerous to allow clinical decisions merely on data. There's a role for intuition in healthcare.
4. It’s never about technology alone. While role-playing, people often forget about time and space. They forget that AI for their hospital isn’t there yet. They extend present-day reality to the future. So they brought up political issues. Policy concerns. Challenges with health IT implementations. Medical fraud. Legal problems. None of which has got to do with technology.
5. AI or not, interoperability is the dominant problem. Systems talking to each other (interoperability) will be a problem. For a long time to come. Even if hospitals take to AI like they've taken to EHRs. Because of the hundreds of thousands of legacy systems around. Those won't be replaced soon. They won't generate relevant data easily. Many vendors don't talk to each other either for business reasons.

Remember, a knife is also technology

The recurring theme in our class was to remind ourselves of what technology is. That a knife is also technology. Something that's been put to good use in history.

A knife that can cut an apple can also cut your finger.

The answer for how AI is used in healthcare, in the end, is simple.

How will we use this new knife?

It's up to us, isn't it?