Back in 1996 the leaders of medicine gathered in Washington D.C. at the Institute of Medicine (IOM) headquarters to discuss something revolutionary and important: health information becoming available to the public on the web.
I was probably the youngest person in the room.
One of the doctors I revered stood and addressed the assembled crowd, telling us that the Internet is dangerous, the risks are too great, and we should warn all patients never to look up health information or answers to their health questions online. Applause filled the room.
When the room quieted, I stood. I had already been answering questions at DrGreene.com since 1995 and putting the best of what I knew online. I said that people searching the web for answers to their health questions was absolutely inevitable. It would soon be unthinkable not to learn online.
Our job was to help create and point to reliable information online – and to teach how to find, judge, and use it wisely. It was to open doors, not close them.
That was then.
Now there’s a new kind of doctor coming to town.
It will be there to help when your child wakes up screaming in the middle of the night. It will reassure you when you’re in pain and scared and don’t know what to do. It might even be able to figure out diagnosis and treatment options for you down the road.
But it won’t be human.
The wave of chatbots is on its way to health, and I’m very excited about it.
You may have interacted with a customer support chatbot before, or had a conversation with the artificially intelligent Cleverbot. There are already more than 30,000 bots that exist just for the Facebook Messenger platform now, and roughly $200M in funding went to AI and chatbot startups last year. Even the Pope has a chatbot.
Yes, health is a lot more complicated than asking about the weather or getting shopping recommendations, so the first medical chatbots will need to be built in very close partnership with human doctors and patients.
It’s like self-driving cars still having real human drivers as they’re learning the ropes. When human lives are at stake, it’s important to get it right.
Why is this great?
Adapting chatbots to medicine has the potential to help millions if not billions of people around the world, right when they most need it.
If you have a cell phone, you have access to a doctor.
Imagine the positive impact on people’s lives with this equalizing force. Especially in developing countries where doctors can be few and far between, or in developed countries where healthcare can be expensive and not immediate.
Even if we can help reduce people’s anxiety about something going wrong with their body (or their child’s body) until they can get to a doctor, that can be an immense relief of global suffering.
Medical chatbots can offer relevant high-quality information, reassurance, answers, and ways of thinking about the situation that might be more useful. They wouldn’t replace human doctors, but they could help set a new, increased standard of care.
What are the risks?
You’re probably wondering, what about liability? What if the chatbot gives someone inappropriate information and they end up being harmed somehow?
This is a valid concern.
But it’s really not that different from patients searching for health information online. In fact it might be safer than that if it’s being trained by doctors.
If we look to the original guidelines developed by the American Medical Association for online health information standards in 2000, many of the same principles can apply to chatbots.
A key principle is that of transparency. People interacting with the chatbot must know they’re not talking to a real person, and be shown the proper disclaimers about health information not being medical advice.
Another important standard is that any medical chatbot present statements of fact rather than directives or opinions. So instead of saying, “you should do this” it could say something like, “people with this condition have been shown to benefit from this.”
And obviously it will be critical to maintain the highest standards of privacy and confidentiality.
By following these guidelines, we can put forth our best effort to protect patients while helping them make informed choices.
Where is this all going?
Real transformation can happen when we look back to honor the past and look forward to embrace the future at the same time.
Medical chatbots mash up the ancient practice of healing with the modern technology of artificial intelligence, which gives them the power to recreate and amplify the doctor-patient relationship. They can make access to healthcare more immediate, more accessible, and more equitable.
And this is only the very beginning.
Sometimes people need to see all the pieces before making a plan and moving forward, and sometimes we just have to start moving forward with a preliminary plan and trust that we will adapt and find the pieces as we go. I think we have enough of the pieces now to move ahead and create a real scalable positive impact on people’s lives around the world.
It’s time to start building, training, and testing chatbots for medicine in a massive way.
Let’s do this!
I’d like to put out a call for a gathering of minds to make chatbots and AI in medicine be of most benefit to the most people.
Let’s make chatbots a complement to doctors, not a replacement. Let’s have them answer medical questions, provide relevant high-quality information and reassurance, be taught by real doctors and patients reviewing what information goes out, and eventually also help diagnose, treat, and even prescribe. And let’s have them reach as many people in the world as we can.
If you’re in this with me, please drop me a line. I’d love to hear your thoughts!