Would you trust health advice dispensed by a computer rather than a human physician? Despite the obvious challenges of using artificial intelligence (AI) in healthcare, researchers have long held that AI can give hard-pressed healthcare professionals a helping hand.
Artificial Solutions has several customers in the healthcare sector, such as Barcelona’s Sant Joan de Deu Hospital, which uses the technology on its website to provide advice on children’s health.
Today’s healthcare industry is making increasing use of Web-based virtual agents and avatars to provide information on benefits, advice on healthy living and so on.
But will computers ever be able to do medical diagnosis? Actually, they already can — and could forty years ago.
Back in the 1970s, Mycin, a pioneering medical expert system, managed to outperform members of Stanford medical school in identifying bacteria that caused severe infections.
It and a clutch of similar first-generation expert systems, promised to usher in a new era of computer-assisted healthcare, relieving physicians of the need to spend so much time diagnosing humdrum cases.
But today we are still waiting for the AI revolution to hit healthcare. So why did systems such as Mycin fail to capitalize on that early promise?
Then, as now, there were tremendous legal and ethical issues surrounding the use of computers rather than humans to perform medical diagnosis.
But the bigger hurdles facing those first-generation healthcare expert systems were technical rather than ethical. That’s because they were conceived in an era of centralized mainframes and primitive communications.
To perform a diagnosis, researchers had to book precious time on the hospital’s shared mainframe. Because there were no electronic medical records, the expert system had to be laboriously ‘fed’ with clinical data by a technician sitting in front of an IBM green-screen terminal –remember them?
In an era of smartphones and broadband communications, it is perhaps difficult to appreciate just how primitive early mainframe applications were.
As well as advances in computing power and communications, recent years have also seen major developments in the area of natural language technology. So, it is possible today for healthcare apps to have meaningful dialogs with patients – and maybe even suggest treatment.
But should we trust the information that an avatar provides?
“You have to be particularly careful with anything that involves human judgment or requires common-sense reasoning about the world, “says Timothy Bickmore, associate professor at Boston’s Northeastern University.
He is involved in several avatar research projects in the healthcare sector. One project uses avatars to ensure that patients that are about to be discharged understand their discharge information. Many do not and run a greater risk of being rushed back to the emergency room.
Using a bedside computer, the patients can ask the avatar questions if they do not understand some aspect of the discharge procedure.
A three-year trial of this” virtual discharge nurse” on patients at a Boston hospital reduced readmissions by 30 percent.
Mr Bickmore is one of a growing number of experts that believe that artificial intelligence could be poised to make a comeback in healthcare, thanks to advances in natural language technologies.
Geoff Nairn recently wrote on health avatars in the Wall Street Journal.