Artificial Intelligence: Why we mustn’t be ‘knobs’

Dr. Indra Joshi is Clinical Lead for Digital Experience, Digital Urgent & Emergency Care at NHS England
Dr. Indra Joshi is Clinical Lead for Digital Experience,
Digital Urgent & Emergency Care at NHS England

I love an awesome dinner party, the food is pungent, drink flows freely and conversations ripple through the table like a bubbling stream in spring. Then someone mentions two wonderfully topical words “artificial intelligence” and the bubbling stream turns into a raging frothy river, as heated comments and opinions swirl around the table struggling to stay afloat in this ocean of possibilities that is AI.

As a doctor, I often wonder how will intelligent systems impact my care on the frontline in the time pressured, resource scarce environment we find ourselves in the NHS. When we feel unwell often the thing we want most is that human touch. I’ve held countless numbers of hands, given reassuring hugs and even cried with the people I’ve encountered daily in the Emergency Department – and on those days, I wonder where in my world would AI fit in.

The Oxford dictionary defines AI as: the theory and development of computer systems able to perform tasks normally requiring human intelligence, such as visual perception, speech recognition, decision-making, and translation between languages. In healthcare, AI has to some extent been around for a long time for example with the use of Bayesian and neural networks for diagnostic reasoning.

Within the primary care setting we are seeing an abundance of symptom assessment “agents” appear on the market, currently they don’t quite fall into true ‘artificial’ intelligence however use elements of machine learning to identify data points and lead the user down a pathway.

To help me with my dilemma of how AI can help in my world in the ED, I recently attended CogX, a conference that brought together thought leaders from around the world to exchange views on the impact of the AI on a multitude of industries with two sessions focussing on mental health and healthcare in general. I was immensely pleased at the turnout for the Mental Health panel, despite having to compete with the topical cyber security and sexy robotics, the room was packed. The value AI could lend to those dealing with mental health problems showed great potential.

The panel kicked off with the recent winner of an under 18s hackathon who gave an inspiring and brave presentation on how an AI driven chat bot could help young people suffering with mental health problems understand how they are feeling and nudge them with their therapy. The founder of PsychApps explained how a clinically assured intelligent platform can give those dealing with depression set tasks and from their daily behaviours, offer tips that can help them with their condition.

The health panel, showed us the how the work of Cortana Intelligence from Microsoft in partnership with the Cochrane, is using AI and machine learning to text mine thousands of reports to automatically select ones to include in systematic reviews, thereby freeing up reviewer’s time for more valuable high level data analysis. Streams from Deepmind explained the value of machine learning in pulling together the multiple clinical pathways a doctor often must keep within their mind and by automating them we identify at risk patients earlier.

Machine learning and intelligent systems bring with them so many questions that we have yet to determine the answers to – who owns the data, is that data good enough, what of the governance, what happens when something goes wrong is it the clinician’s fault or the developer’s? The list goes on. However, they also offer some great merits in health: they can analyse data consistently inferring patterns, over time learn behaviours offering personalised insights, reduce human errors by considering multiple touch points, but for me most crucially they free up time.

Time is a rare commodity we lack in today’s NHS and in healthcare in general. Without time, we can fall into the trap of being knobs to our patients by underestimating what they have gone through to get to our door or by not hearing what they are not saying. So, despite some grumblings in my gut from the dinner party excitement, let us have a calm and open dialogue to explore the issues AI throws up. In the meantime, let’s embrace some of the approved intelligent solutions out there that offer us that rare commodity – time, and ultimately let us not be knobs.