With personalised medicine and AI, doctors get more and more responsibility and might face a hard time proving they are right if they don’t agree with the machine.
Let’s imagine a fully automated Danish artificially intelligent healthcare system. It is very efficient, and our lifespan goes from 80 years to 90 years. Because of the AI, we are being measured and screened. And thus, we will constantly look at ourselves as potentially sick. We are now patients all the time. Would you give these 10 years you’ve gained in lifespan for freedom from thinking about your health during most of your life?
These are some of the ethical questions Ezio Di Nucci, professor of bioethics at Section of Health Services Research at the University of Copenhagen, ponders over when he is asked to assess the advantages and disadvantages of new technology, and not least artificial intelligence (AI).
“Usually, most of us will start thinking about our health when we are 50. Now with AI and constant tracking and screening, we might do it already when we are 20. Is that worth it?” asks Ezio Di Nucci. He has himself gone off grid, having no phone “But the economic incentive is so strong, that we probably cannot resist it in the health sector, even though it might not be worth it.”
As part of the Personalised Medicine and MG-PerMed Project, Ezio Di Nucci has been discussing ethics in medicine for years. Another ethical consideration is the following:
“Let’s say that without AI for every positive hit on breast cancer, there are three false positives. Three women will be faced with thoughts of the possibility of dying. Then we get AI. The optimists will say, the false positives will go down from three to two. But with the more precise – and possibly cheaper results – the absolute numbers of screenings will rise. We might not start screening when women are 50, but instead 40 or even 30. Which will mean that even much more accurate outcomes will result in more absolute false positives, for example… and then the question is again: is it worth the trouble? But the economic pressure might make it really hard to say no.”
Many of the answers to these questions would be better to give in hindsight. But before buying new tech and implementing it, we should at least have these discussions, he believes.
Increasing Responsibility
Most AI in the health sector today is still mainly based on machine learning (ML) which operates only with human involvement, whereas the newer version of AI, generative AI, is less used, but has the potential to be used without human involvement.
Doctors are faced with increasing responsibility when using ML and AI. There always has to be a human in the loop, according to EU laws and ethics – that is, the doctor. But what is meaningful human control asks Ezio Di Nucci:
“All AI is approved as only decision support. But the human is at the same time under increasing pressure to make more and more decisions. If the human doctor just ticks the box, it is probably not meaningful. If the doctor does not understand – how do they then say no?”
One ethical principle is thus very important: Explainability. But it is more important for the doctors who make decisions than for the patients, he believes and says that another worry is the fact that big tech companies today are sitting on most AI services.
“We risk taking the human out of the loop – for the sake of profits to big AI tech companies.”
Automation Bias
As AI is increasingly being embedded in the health sector, we need to be observant when it comes to so-called automation bias. We are becoming reliant on the machine, so when the machine says cancer and the doctor says no, it is not cancer, you will probably favor the machine.
“The burden of proof is shifting. Now it is the human who has to prove that she is right about the decision, and this might overwhelm the doctor.” says Ezio Di Nucci.
Human doctors might end up having to prove themselves against AI systems the way when we use the Internet, we often have to prove to the machine that we are humans via a Captcha. Just in case you were wondering who is really in charge.
This is the last article in a series about Personalised Medicine and the MG-PerMed Project
The first was about Ethical Aspects of Personalised Medicine. The Second was AI Will Always Be a Tool, Not a Substitute for the Neorologist, and the third was about Patients’ Expectations to Personalised Medicine. We’ve used Myasthenia Gravis (MG) as a case, as we followed the EU-financed project (MG-PerMed) dealing with personalised medicine for MG patients. However, the same ethical aspects of personal medicine exist in the personal treatment of all other diseases. MG makes muscles weak because the body’s immune system attacks them. People with MG often need medicines that weaken their immune system to control the disease. But it varies a lot from person to person, what medicines they need. This project, Prevention in Personalised Medicine, aims to tailor treatments to each person’s unique needs instead of a one-size-fits-all approach.
