At a packed panel debate at The Cell in Hagastaden, the matter of research funding and infrastructure evolved into a discussion on the use of AI, who has power over our health data and which human skills ought not to be ceded to machines. The conversation brought together perspectives from science, social science and science journalism.
One panel member was Elham Rostami, who researches traumatic brain injuries. She often uses AI to analyse large amounts of health data, in part to create improved diagnostic tools. However, the step from a functioning AI model to the broad application of AI in healthcare is longer than one might think.
“There’s no real guarantee that the tools we develop here can also be used in countries like Brazil, the USA or Norway,” said Dr Rostami , docent of neurosurgery at KI’s Department of Physiology and Pharmacology . “So its essential that what we develop is properly validated for patient groups, geographically and demographically.”
Achieving this also requires secure access to health data, ideally as near-patient as possible, added Johanna Furuhjelm , director of the Karolinska Institutet Centre for AI Innovation , who went on to stress how environments in close proximity to hospitals and clinics must also be made available to smaller companies.
“Rarely do such companies have the same resilience to make the innovation journey required to test their solutions, even when possessing technology perfectly suited to healthcare needs,” she said.
Retain critical thinking skills
There are many examples of how the ability of AI models to detect patterns in large volumes of health data has provided hospitals with new tools. That said, the results always depend on the data used. Emma Frans , author, lecturer and research specialist, raised the caveat that the technique per se creates the illusion of objectivity:
“Today’s language models come across as very convincing,” she said. “So we need to retain our critical thinking skills and consider such things as bias and false positives. And to scrutinise the data we feed in and how the results can be applied.”
The panel also stressed the need to include clinics, engineers, researchers and humanists at an early development phase of novel AI tools. At the same time, there must also be a clear understanding of who is to ensure the accuracy of the information and who will accept responsibility if something goes wrong.
Data is power
A similarly important question concerns who has control of our health data. According to the panel, the data ultimately belong to us citizens; however, it is far from an easy task for either clinics or patients to access them.
Yet the question of who owns the information is not a new one, as professor of the history of science and ideas at Stockholm University Staffan Bergwik pointed out. Access to it has usually been at the whim of mediaeval monarchs and dictators in one way or another.
“When we discuss who owns data, we inevitably get caught up in the question of how society is to be organised, which makes it a profoundly political matter,” he said.
The value of the interpersonal
So when AI tools take over more and more tasks, what then is left to us humans? One path for the healthcare sector to take could be to devote more time to interpersonal relations, something that Dr Frans noted it has decidedly not done, having seen the human as a machine required to produce the maximum amount in the minimum amount of time.
“When the machines take over certain tasks, it could be better to put more time and energy into interpersonal relations,” she said. “Maybe sometimes we should even strive to be slower and more reflective.”
Dr Rostami agreed that there was a strong desire among doctors to spend more time with their patients at the expense of their current administrative burden.
Important to uphold people’s trust
The discussion also addressed the aversion to AI that has emerged among young people, particularly in the USA. How can we avoid such a movement in Sweden? According to Dr Frans, there needs to be a concerted effort to ensure that the technology develops in a way that benefits the individual and society.
The debate was part of the National Museum of Science and Technology’s election special under the Samtal & Idé programme series co-arranged by The Cell and the Centre for AI Innovation. The moderator was Maja Fjaestad , strategic senior advisor to the president at KI and manager of the centre’s policy lab.
