Prospective parents faced with choosing which embryo to implant appear to weigh non-medical traits just as heavily as medical ones when given concrete statistical estimates about genetic risk, according to new research. The study, reported in Nature Human Behaviour, suggests that the boundary between medical and non-medical traits softens once people must make practical reproductive choices.
Preimplantation genetic testing is a medical procedure used during in vitro fertilisation. Before an embryo is transferred to a uterus, doctors extract a small number of cells and analyse the genetic material. Polygenic testing is a specific form that assesses thousands of genetic variations at once to produce a risk estimate rather than a single disease mutation.
These scores rely on statistical models that compare an embryo’s DNA with population data, offering a probabilistic picture of how likely a person is to develop complex traits, from heart disease to intelligence. They do not determine an innate destiny; rather, they provide a probability, not a guarantee, about future outcomes.
Earlier surveys showed public openness to embryo screening using such complex statistical scores. For example, a 2023 study found broad willingness in the United States to screen for a range of physical and behavioural traits, while later research indicated that non-Western populations might be even more receptive to selecting for non-medical traits, with Singaporeans appearing more open to educational and cognitive outcomes than Americans.
The new work builds on these findings by testing whether people’s stated moral attitudes match how they act when balancing medical risks against personal traits. Led by Edmond Awad, a senior lecturer at the University of Exeter and senior research fellow at the University of Oxford, alongside Julian Savulescu and Dominic Wilkinson, the team sought to discover whether providing these statistical estimates shapes prospective parents’ choices, even if they initially say they would not test for non-medical traits.
“I have argued for a long time that when parents are choosing between embryos they have reason to choose the one expected to have the best life, and that the medical label is the wrong place to draw the line,” Savulescu told PsyPost. “This study was about what people actually do.”
“My hunch was that people would not sort attributes into medical and non-medical once a real choice was in front of them, and that they would go instead by how much an attribute would affect the child,” Savulescu explained. He noted that Awad had previously run the Moral Machine, which guided millions of people through online moral dilemmas about automated vehicles, making a similar approach ideal for this question. “Those pilots turned up a gap between what people said they wanted tested and what they did with the information once they had it.”
The researchers also considered a psychology known as deliberate ignorance — people may avoid learning about non-medical traits to sidestep ethical dilemmas. But if a clinic provides the information anyway, those same people might still use it to decide which embryo to implant. The team recruited 1,467 adults in the United States for an online experiment, focusing on four attributes that a polygenic score might estimate: two medical conditions (heart disease and vision impairment) and two non-medical traits (low cognitive ability and antisocial behaviour).
In the first group, participants were simply asked whether they would be willing to use polygenic testing to obtain risk estimates for each attribute, answering yes, no or not sure. The results showed 66.7 per cent were willing to test for the medical conditions, while 61.4 per cent would test for the non-medical traits, indicating that a sizeable minority would consider screening for cognitive ability and behavioural tendencies.
Awad emphasised that the study was conducted as a Registered Report, with hypotheses and the analysis plan reviewed and approved by the journal before data collection commenced. He said: “On the basis of our own pilot studies, and of other recent work, we registered a prediction that most people would not want testing for low IQ or antisocial behaviour. The opposite happened. Sixty-one percent said they would.” He added that wording could influence responses, noting a possible shift toward greater willingness to test when questions framed avoidance of negative traits rather than selection of positive ones were used.
The second group faced hypothetical embryo-implant decisions after genetic testing had already been completed on two viable embryos. Participants had to choose which embryo to implant based on the estimated risk for two of the four traits, with the scores indicating either an average chance, a 5 per cent above-average chance or a 5 per cent below-average chance of developing the trait.
When forced to choose, participants avoided embryos with a higher-than-average risk of undesirable non-medical traits as strongly as they dodged those with medical conditions. Wilkinson, a professor of medical ethics at the University of Oxford and consultant neonatologist, described the magnitude of the effect: “What we found was that in the US sample, an embryo with a higher-than-average chance of developing heart disease in later life was about 40 percentage points less likely to be chosen than one with an average chance. For antisocial behavior the gap was 36 points, for low IQ 32 points, and for vision impairment 19 points.”
He added that researchers had defined “above average” as five percentage points above the average, noting that a small shift in the odds could produce a large change in the choice. Savulescu observed that people care deeply about the life prospects of a child, a high-stakes issue that can distort probabilistic reasoning in ways clinicians recognise.
Across the four traits, participants tended to avoid higher risks for vision impairment less than for the other conditions or attributes. The team also included direct trade-off scenarios, pitting a higher risk for a medical condition against a higher risk for a non-medical trait, with vision impairment often tolerated in preference to higher risks for cognitive ability, antisocial behaviour or heart disease.
To assess cultural variation, the researchers replicated the study with 623 participants in China, who completed the embryo selection task without the initial willingness survey. The results largely mirrored the American findings, with a pronounced aversion to higher risks for non-medical traits, and an even stronger aversion to low IQ among Chinese participants.
Participants were also asked to rate the four attributes on various dimensions, including controllability and impact on life. Rather than avoiding traits perceived as more controllable, the researchers found that perceived seriousness of the outcome drove decisions, regardless of whether the trait was medical or non-medical.
“What people say they want tested does not tell you what they will do with the answer,” Savulescu commented. “The label on an attribute is doing less work than people assume. What guided the choice was how serious the outcome seemed for the child. Doctors and regulators care about disease; people care about well-being, of their child and others.”
Wilkinson stressed that the study does not suggest people want designer babies. “What we found is that people want to avoid their children having genetic risk factors that might affect how well their life will go in the future. It is important to note that these were hypothetical scenarios and we don’t know what people would do in real life, especially when there are risks or financial costs involved.”
The researchers caution that the study used simplified, hypothetical scenarios with uniform five-per-cent changes in risk and that polygenic scores for behavioural traits are currently less precise than those for medical conditions. Awad noted that real-world decisions involve different base rates and margins of error, and that in a clinic, individuals may choose not to view certain genetic information at all. He emphasised that real consent processes may influence outcomes differently from the study’s design.
Wilkinson also warned that within a family, polygenic predictions for behavioural traits such as cognitive ability are weaker than the scenarios imply, meaning the potential gains from selecting on them would be far smaller in practice. The researchers suggest that the way information is presented can influence decisions, highlighting that signing a consent form is not the same as making a fully considered choice.
The project forms part of ANTITHESES, a Wellcome Trust-funded programme examining deep value disagreements. Savulescu described the work as helping to reveal how public values shape ethical policy, arguing that public preferences should inform—not settle—policy decisions. He said that the findings challenge the traditional medical versus non-medical dichotomy and prompt renewed consideration of how traits are categorised in policy and practice.
Awad noted plans to expand the project with Tinker Tots, an open online experiment offering embryo-selection dilemmas across more traits and countries, with the aim of moving closer to clinical contexts and studying people who are actually navigating IVF decisions.
The study, titled Public perceptions of polygenic testing and embryo selection for non-medical traits, lists Awad, Colombatto, Demaree-Cotton, Earp, Everett, Liu, Schaefer, Singh, Wilkinson and Savulescu as authors.
