King's hosts experts to discuss how to ethically implement AI in healthcare

It featured a stellar line-up of expert speakers from across government, academia, the medical sector, industry and patient community. Dr Raquel Iniesta, Reader in Statistical Learning for Precision Medicine at the Institute of Psychiatry, Psychology & Neuroscience, King's College London, hosted the day-long event.

Thank you to all our speakers and audience for a fascinating and enjoyable event. We demonstrated that the conversation around making AI ethical in healthcare needs to involve many sectors, and valuable insights can arise from bringing together policy makers, academics, patients, developers, industry and clinicians.

What factors do we need for an ethical integration of AI in healthcare?

Sara Cerdas, a former Member of the European Parliament, delivered the first keynote speech, highlighting AI's potential to revolutionise healthcare systems and the establishment of the European Health Data Space (EHDS). She emphasized the necessity of fostering patient trust, alleviating technostress among clinicians, and addressing critical issues such as bias, fairness, data governance, and regulatory compliance. In addition to the EHDS, she discussed the legislative efforts surrounding the European Union's Artificial Intelligence Act.

She was followed by Dr Rupa Chilvers, Deputy Director for Life Sciences and Innovation at the Welsh Government, who used an example of cancer diagnosis to illustrate the risks of creating a two-tiered healthcare system. She raised potential issues around decision-making pathways and the need to test and evaluate service design early in order to deliver an ethical, AI-enhanced healthcare system.

The first panel discussed factors that are important to guarantee ethical integration, covering topics such as accountability and cybersecurity.

Ms Cerdas and Dr Chilvers were joined by Sarah Markham (Patient Representative, Visiting Researcher, King's IoPPN), Robin Carpenter (Head of AI Governance and Policy, Newton's Tree), Beatrix Fletcher, (AI Programme Manager, Guy's & St. Thomas NHS Foundation Trust) and Basab Bhattacharya (Clinical Informatics and Radiologist Lead, Barking, Havering and Redbridge University Hospitals NHS Trust). Panelists were split on whether there is currently enough regulation for AI to be implemented safely, raising questions about the training and support available for clinicians.

Human actions to align AI with ethical values in healthcare

Professor Payam Barnaghi, Professor of Machine Intelligence Applied to Medicine at Imperial College London highlighted in his afternoon talk that involving clinicians early in the modelling process will ensure AI tools reflect healthcare professionals' experiences and provide the most clinical utility.

Professor Susan Shelmerdine, Consultant AI Paediatric Radiologist, Great Ormond Street Hospital, advocated for more awareness and development of AI tools for children's healthcare, although cautioned that her research has shown that children, as digital natives, are more wary than their parents about using AI in their medical care.

A lot of the time we anthropomorphise AI and think it can do the same as humans, but they analyse the world in a different way to us.

The second panel discussion focused on the human side of implementation. Professor Barnaghi and Professor Shelmerdine were joined by Dr Nenad Tomasev (Senior Staff Research Scientist at Google DeepMind), Dr Ellie Asgari (Consultant Nephrologist, Guy's and St Thomas' Hospital), and two patient representatives from South London and Maudsley NHS Foundation Trust - Jennie Wilson Bradley and Emma Shellard.

Key themes emerged around the role of the patient and dealing with bias in data. Patients should be meaningfully involved throughout development, and their safety, care and wellbeing should always be put first. Although a holistic, 'human-in-the-loop' approach is desirable, it faces challenges on a large scale.

Wrapping up the panel, several speakers agreed that feedback loops and continuous retraining of models will be needed to implement AI successfully.

We need to remember that AI is not just one thing, it is a spectrum - and so is the risk.

Roadmap for an ethical AI in healthcare

Dr Iniesta leads the Fair Modelling Lab at the Department of Biostatistics & Health Informatics, King's IoPPN. Her research, supported by the NIHR Maudsley BRC, delves deeper into this area and she has published her work in the journal AI and Ethics which describes five facts that can help guarantee an ethical AI in healthcare. You can also read her two-part blog on the topic here: Part I and Part II.

The conference was part of an international partnership grant that aims to investigate the human role in guaranteeing an ethical implementation of AI in healthcare. Dr Iniesta is the Principal Investigator of this grant from Responsible AI UK. The event was also supported by UKRI - UK Research and Innovation.

Thank you for the collaboration of the Disruptive & Emerging Technology Alliance (DETA), the Catalan government and the UOC (Universitat Oberta de Catalunya) in Barcelona.

Recordings from the event will be shared shortly on this webpage; you can also contact [email protected] if you would like to be notified when they become available.