There is much enthusiasm about the potential of AI to improve healthcare. It is true that healthcare can probably be delivered both more efficiently and to a higher standard by eliminating unnecessary variation. This variation is introduced by failure of clinicians to recall accurately what treatment fits with what symptoms, for example, or sometimes unconscious biases and heuristics that may lead to the wrong conclusions being reached. However, the question is whether AI is ready to take the place of a clinician - or is it best seen as a tool that will help clinicians make decisions and remove unnecessary variation in how care is delivered?
One area that highlights the limitations of AI is around shared decision-making. Shared decision-making is about empowering patients to arrive at their own conclusions on which treatment is right for them. It recognises that what is a desirable outcome for one patient is not necessarily the right one for another. The decision should and can only be made by the patient himself or herself. A clinician's roles in the shared decision-making process is to explain as clearly as possible the treatments, likely outcomes and possible side effects. For example, a patient suffering with a urological problem may be offered treatment that would extend their lifespan but curtail an active sex life. Only that patient will be able to assess the costs and benefits to them of this treatment and decide whether it's the right one for them.
Of course neither doctors, nurses or other clinicians are immune from the problems of biases, heuristics and an inability to recall details with 100% accuracy. This is why computerised decision aids are gaining popularity for their ability to impartially explaining different treatment options and helping patients to reach a decision (http://sdm.rightcare.nhs.uk).
Is the future one where shared decision-making about our health is with a robot powered by an artificial intelligence so powerful that it has made the clinician obselete? How close is this to being a reality - where we are alone with our robot? In the age of isolationism - isn't this the ultimate example of that world vision?
Some cite recent advances in the prowess of AI - driverless cars and the success of AlphaGo. But these a literally two dimensional problems; impressive yes, complex - sort of, useful - only to illustrate how far AI has still to develop. Allow me to put it another way. A robot capable of helping us make high stakes decisions about our health has, as a bare minimum I would argue, to pass the Turing test - to be indistinguishable from a human - otherwise how can the robot demonstrate empathy that is essential for compassionate shared decision-making?
The Loebner Prize, held annually at Bletchley Park, is a coveted prize drawing competitors from all over the world by offering a Solid 18 carat gold medal and $100,000 to the creators of the very best AI bot capable of convincing four judges in conversation that it's a human. If the bot can convince half the judges - a silver medal and $25,000 is awarded. Tellingly, neither medal has ever been awarded. In it's 25 years, the judges of the Lorbner Prize have only awarded the best competitor of the day with a consolatory bronze medal for their efforts.
"...when the stakes are high and making the right decision is not straightforward - most of us will want to have that conversation with someone who can at the very least convince us they're human."
Of course, there are some situations where we don't really care whether our clinician (or the robot playing that part) is capable of demonstrating humanity at all - such as when we go to collect a repeat prescription for an asthma inhaler, for example. Nonetheless, when the stakes are high and making the right decision is not straightforward - I expect most of us will want to have that conversation with someone who can at the very least convince us they're human. Ideally, we can talk through the decision with someone who is knowledgeable about us, our background and family situation and someone who has an expert's understanding of our diagnosis and the treatment options available and what they will entail. Ideally this process will be supported by tools that I as the patient can use to reflect on the decision and discuss it with my loved ones.
The chasm that currently exists between primary and secondary care, between GPs, specialists and their patients in the UK and across the world makes compassionate, informed shared clinical decision-making difficult. The NHS may be wise to focus more on where the stakes are highest in healthcare, where the decisions are not straightforward - using technology to enable conversations to happen between patients, their GP and specialists and rather less on the controlled (2D) cases where AI may prove helpful.
To find out more contact me at owain@cinapsis.org
Almost done…
We just sent you an email. Please click the link in the email to confirm your subscription!