Jonny Jackson is a CDT PhD student with a research interest in how machine learning can be used to interpret medical images. In August he hosted a panel discussion to engage the general public with the debate around the future of Artificial Intelligence in healthcare. Jonny tells us about his experience here:

Earlier this year, I was fortunate to be given the opportunity to develop and host an event at the Barbican’s “AI: More Than Human” season. Together with the CDT’s Public Engagement Officer, Bella Spencer, I decided that I could make the most of my position within the research community by lining up a panel of experts for a public discussion on the increasing use of (so-called) Artificial Intelligence (AI) in the healthcare sector, which we titled “The AI Will See You Now”. I wanted a group of panelists whose expertise would cover as much of the breadth of the topic as possible, from the technical to the ethical, so over the course of about 4 months I built up a panel comprising of a doctor; Dr Ali Jomaa, an AI researcher; Prof Tom Vercauteren, a public policy researcher; Dr Christine Aicardi, and a science fiction theatre producer; Juan Echenique.

IMG_0918.JPG
The panel: (from left to right)  Juan Echenique, Dr Christine Aicardi, Dr Ali Jomaa, Prof Tom Vercauteren and Jonny Jackson.

My particular aim for the evening was two-fold. First, I believe there needs to be more of a flow of knowledge about the science and development behind the data technologies that are shaping society. That helped me define the four main subjects to be discussed during the evening: do we need AI in healthcare?; can it deliver on its promises?; will it always be used for good?; and who should control it?

Second, I also believe that it is vital to understand how the public feels about the potential use of AI technology in medicine in order to address any issues with public image before it causes a real problem for scientists (consider the public image crisis of GM crops). To really understand the audience’s preconceptions, hopes and fears, that meant enabling a two-way interaction between the panel and audience members from the opening minutes of the event to the end. Inviting questions from the audience is the best way to achieve this, but I also wanted to ensure a coherent structure during the event. In the end, I decided to test out a digital solution to this compromise, by providing a way to submit questions phone in real-time via mobile, which I could then pick, group and pose to the panelists. The questions posed by the audience were fascinating and led to a very dynamic discussion. They included:

  • ‘Is it possible to ensure that both the benefits and costs of AI in medicine are equally distributed across the country? How do you prioritise?’
  • ‘What about from the doctors’ perspective and well-being – could AI decrease the pressure on difficult decisions?’
  • ‘Often differences between each medical cases are very subtle, how would AI be “trained” to distinguish between these nuances? Would it adapt fast enough to aid?’

The event was a great experience for me as my first taste of chairing a panel debate and the feedback was great. There were 33 questions asked digitally and several more during the more traditional Q&A session at the end, with 100% of our 60 member audience saying that they would recommend the event to a friend. I was also able to get some great insights into how people felt about the issue of data privacy with their medical data. After the event 72% of respondents said they were comfortable with idea of AI being incorporated into healthcare, 11% said they weren’t comfortable and 17% were unsure if they were OK with that concept. The reasons given by the audience for feeling comfortable included ‘I think the benefits outweigh the risks which can be mitigated. It is an aid to more efficient healthcare, not a replacement.’ and ‘It can be used for even simple things. It can have an enhancing role rather than a dominant authority one’. Conversely, those who were not comfortable with the idea stated that it is because of ‘privacy issues’ and feeling ‘unconvinced of AI’s potential for empathy.  We also asked the audience what reason would most likely motivate them to give healthcare providers access to their data:

BarbicanPollAnswers.png

Though I wasn’t talking about my own research directly, organising this event has really consolidated my interest in engaging with the public in the context of the work I’m doing. It has given me a small but important insight into the sentiment of the public for artificial intelligence and the policy issues that will become increasingly important as the technology grows ever-more present in our society.

photo1.jpg
A series of short stories and illustrations developed around the theme of ‘the future of AI in healthcare’ for the event.