The robot will see you now: can doctors survive the rise of AI?

Have you ever phoned 111? If you have, you will have been asked a series of questions by a helpful human. This human will try to ascertain whether you need to go to hospital, see a GP or rest easy.

But for residents of some London boroughs, that experience is about to change. The service will soon be provided by a robot. In a pilot scheme, patients will be encouraged to use a ‘chatbot’, developed by Babylon Health, which is claimed to diagnose patients ‘faster and more accurately than a doctor’.

This is no one-off, either. A growing number of such computer programmes are in development or in operation. For example, Alphabet, Google’s parent company, has developed a system that can apparently detect cancers on scans more accurately than doctors. Another system, DXplain, generates a set of possible diagnoses after a doctor feeds it data on things like symptoms and investigation results.

Then there’s Dr Watson (nothing to do with Sherlock Holmes). He is a computer system developed by IBM and named after its first CEO, Thomas Watson. IBM has a history of pushing the boundaries of artificial intelligence. Some may remember the victory of Deep Blue, another of its computer systems, over chess champion Garry Kasparov. Watson is part of this pedigree, developed as a technology platform to answer questions posed in natural language. He (it?) was first put to the test on the TV quiz show Jeopardy!, but now IBM wants him to diagnose and treat medical illness.

What makes Watson different is that, according to IMB, he ‘processes information more like a human than a computer’. If this is true, it would mean Watson could process completely unstructured information such as patients’ medical notes, research journal articles, or human speech. Doctors would not need to sanitise and enter data themselves, as the computer would be able to make sense of all necessary information for itself, potentially in their absence.

It is unclear how far IBM has succeeding in developing Watson to this level. What is more, diagnoses and clinical decisions are often more complicated than the questions on Jeopardy!, made on shaky foundations full of uncertainty and conflicting ideas. There is rarely a ‘right’ answer and we have to rely on expert opinions that are hard to verify. In the absence of cold, hard facts, it is extremely difficult to train Watson.

On the other hand, computers have several clear advantages over doctors. A computer never suffers a memory lapse, needs not rest, has never missed a lecture, reads every bit of new research and is unrelentingly logical.

As a lowly medical student, sitting in a library, surrounded by reams of pages listing facts I do not know, I feel quite intimidated. Are computers on a trajectory to replace doctors?

Perhaps I am biased, but I think artificial intelligence in medicine has limitations, too. Doctors are not solely diagnosticians or decision makers. The word ‘doctor’ is from the Latin docere, meaning ‘to teach’, so doctors are, in essence, educators. The modern doctor’s role is to help patients understand their bodies and ways that they might improve their health. Actually identifying what is wrong and identifying potential treatments is merely to facilitate that education. This emphasises that patient autonomy should be at the heart of clinical decisions. For example, the most effective treatment in terms of life expectancy may not be what a patient wants if the side effects are undesirable. Helping a patient understand a treatment’s likely consequences will allow them to make an informed choice.

This job becomes all the more important in today’s age of information. A pet hate of many doctors is when patients go away and read excessively about their condition on the internet. Often this can lead to warped, half-true ideas formed without a way of processing or critically appraising information. Doctors would much rather be part of the education, lending their expertise to help a patient navigate the vast seas of online knowledge.

As a patient, you can imagine similar problems surfacing if your advice were coming straight from a computer. To be told you have had a miscarriage is not the same as understanding that you have had a miscarriage. And to be told the treatment options for a miscarriage is not the same as understanding what they are.

A computer would be able to inform, perhaps better than a doctor, but it would not be able to educate. Plonk a physics textbook down in front of a 10-year-old and you won’t turn them into Einstein — you need a teacher as an interface between the information and the student. As medicine seeks to embrace the awesome power of computing, this is how doctors should define themselves.