By Wendy Tuohy
Artificial intelligence is being used in Australian fertility clinics to help choose which embryos should be transferred into patients, with the potential to be dehumanising for parents and babies, researchers say.
Allowing machine learning to make decisions about “who is brought into the world” without ethical oversight of its introduction could erode public trust in fertility clinics, say authors of an Australian paper that raises bioethical concerns.
Fertility clinics in Australia are introducing machine learning into the process of selecting embryos during IVF.Credit: iStock
Professor Catherine Mills, head of Monash University’s Reproduction in Society research group and one of the authors of the paper, said IVF patients and partners might not know if AI had been used to help select which of their embryos to use, or how algorithms were trained to make their choice.
Artificial intelligence carries a risk of unintended bias, according to the paper published by the European Society of Human Reproduction and Embryology. This includes that “ML [machine learning] algorithms will perform better for members of some groups than others (eg based on ethnicity)”.
The technology may “take into account features that patients would not want to influence the choice of embryo (eg if the AI system is more likely to recommend transferring embryos of a particular sex or, theoretically, embryos with disease traits that happen to correlate with a higher chance of implantation)”.
Mills said it was difficult to know how widespread use of AI in embryo selection in Australia’s booming fertility industry was. Revenue for the local sector was estimated at $US922.9 million ($1.49 billion) in 2023 and is expected to reach $US1.63 billion by 2030.
“Some clinics are using it; there is some empirical data some are using it routinely,” said Mills. “We think it’s going to be used more widely.”
The technology aims to improve selection of embryos and the chances of successful pregnancies, and to reduce the time to pregnancy and cost of treatments. A national fertility group, Virtus Health, recently ran a clinical trial of the predictive success of artificial intelligence in analysing embryo development. Virtus Health was contacted for comment.
Mills said there had not been enough discussion about ethical implications and consent.
“There are so many examples of technologies being introduced into assisted reproductive technology clinics that have dubious benefits, in some cases no benefits, and in some cases are harmful or are costing people money without any real benefits,” she said.
“It’s because of the way innovation drives [services]. We often don’t know how effective something is before it’s actually used. I don’t think this is good for people using assisted reproductive technology.”
Bioethics lecturer Dr Julian Koplin, lead author of the paper, said while there were benefits to the technology there were no guidelines, regulations or requirements concerning disclosure of AI’s use to patients. “And it may be the case there is someone who doesn’t want decisions about what children they are having being made by AI rather than a human embryologist assessing it themselves,” he said.
That computer algorithms “are beginning to make decisions about who is brought into the world” involved interfering with a particularly sensitive area of human life, Koplin said, “which needs to be handled carefully and respectfully”.
“Particularly if [AI] helps people have cheaper, faster and less emotionally draining IVF, then that seems good, but given it is intruding on something very important to a lot of people, their family plans, it’s important they are informed, have alternatives and are not required to use the technology – and give informed consent,” he said.
Patients of IVF clinics must be given the right to opt in, or out, of the use of artificial intelligence in embryo selection, researchers say.Credit: iStock
Fertility Society of Australia and New Zealand president Petra Wale confirmed that AI was being used in Australian fertility clinics, “primarily as a decision-support tool to assist embryologists in ranking embryos based on various parameters”.
The final decision regarding embryo selection always rested with a trained clinical embryologist, she said, ensuring oversight and human judgment remained central to the process.
“AI algorithms offer several proven benefits,” Wale said, including standardised assessments, reducing variability between embryologists, efficiency of lab workflow by automating parts of the process, and decision support for embryologists.
“Regarding transparency, clinics inform patients about the technologies used in their treatment, including AI tools where applicable. However, as AI continues to evolve, ongoing discussions around ethics, guidelines and public awareness remain critical,” she said.
Reproductive scientist Amy Webb, of Monash Bioethics Centre, said some patients interviewed for the group’s research hoped AI might produce more information that would lead to a successful pregnancy sooner than may otherwise happen, and were excited about its introduction.
Some embryologists felt the extra information offered by AI was useful, Webb said, “as long as patients are informed of the use it and have their say, and the choice to opt in or out”.
However, “a lot of patients do feel a bit disconnected when they go through IVF and there was fear among some around making something that is difficult feel more difficult by reducing human contact,” she said.
The Morning Edition newsletter is our guide to the day’s most important and interesting stories, analysis and insights. Sign up here.