Parents may say they do not want to select a child's personality, but would the choice remain the same when a score appears beside an embryo?
Original Chinese title: 父母嘴上說「不想選孩子的性格」，但當那個分數真的擺在胚胎旁邊，選擇還會一樣嗎？
People may distinguish between wanting to receive a polygenic score and using it when it is displayed beside an embryo. That gap makes interface design, uncertainty, population limits and the right not to receive information central to reproductive technology ethics.
鍾靜蓉
A digital education PhD and specialist in digital teaching strategy, metadata reasoning and analytical interpretation.
Public discussion often reduces polygenic embryo testing to a dramatic question: will future parents choose a child's intelligence, height or personality? That framing is memorable, but the ethical problem is more precise. A 2026 Registered Report in Nature Human Behaviour studied attitudes toward polygenic testing and embryo selection in samples from the United States and China. Respondents were generally more comfortable with disease-related testing than with non-medical traits. Yet when scores appeared in a forced-choice implantation task, the medical versus non-medical boundary did not fully predict what people would choose.
In other words, wanting access to information and using information once it is visible are different questions. This distinction shifts the ethical focus away from model accuracy alone and toward the environment in which a reproductive decision is made. A number can be statistically uncertain and still become a powerful action signal when it is placed beside an embryo.
A polygenic score combines statistical associations across many genetic variants into an estimate of risk or tendency. It is not a single gene determining an outcome and it is not a forecast of destiny. Disease, height, educational performance and behaviour are shaped by genes together with environment, family, society and opportunity. Even when a score has predictive value in a population, the practical difference between individual embryos may be small.
Model performance is also affected by the training population. A score built with data from people of one ancestry may not perform equally for other populations. This creates a serious risk in embryo selection: probability can be mistaken for certainty. A report that displays a ranking may hide the fact that the ranking is unstable, population-dependent or too small to have a meaningful individual interpretation.
Human decisions respond strongly to information architecture. Highlighting, ordering or colouring a result changes attention. If an embryo report displays disease risk, height, education-related scores and other non-medical traits by default, the information has already entered the decision environment even if the clinic says that the numbers are only advisory. Choices do not occur in a vacuum; they are shaped by interfaces, defaults, labels and explanations.
Medical ethics often emphasizes informed consent, and it can seem that more information must always be better. But a result that is highly uncertain, easy to misunderstand and psychologically pressuring may not increase autonomy simply because it is provided. If two embryos have a small difference in a non-medical score, a prominent rank may make the difference feel important. The key questions become which information should be shown by default, which should require an active request and which should not enter a clinical report at all.
In 2026, ASRM stated that polygenic embryo screening is not ready for routine clinical use, especially for selecting non-medical traits. The concerns include prediction limits, population fairness, interpretation, disease prioritization and ethical risk. ESHRE has also urged caution about PGT-P because individual-level predictive ability and clinical utility remain insufficient. These positions do not say that polygenic science has no value. They recognize that embryo selection has a direct consequence: the information is used to choose which embryo will be transferred, not merely to discuss a future possibility.
Policy often uses a medical versus non-medical boundary, but many traits do not fit cleanly into it. Height may appear cosmetic while extreme height can relate to health. Anxiety, attention, learning and behavioural traits can involve medicine, function and social values at the same time. One genetic variant may also affect several traits through pleiotropy. Trying to reduce one risk can alter another characteristic, so the policy phrase select disease but not personality may be less clear in real polygenic architecture.
One sensitive ethical issue is the meaning of normality. If society treats some physical or cognitive differences as traits that should be removed at the embryo stage, what message does that send to people who already live with those characteristics? Disability rights and disability justice distinguish the burden of a health condition from the barriers created by education, buildings, employment and institutions. Embryo-selection ethics therefore has to examine not only individual risk but also the values a society communicates.
Reproductive selection is often described as a private parental choice, but markets, clinical advice and social pressure all shape decisions. If a clinic offers a premium embryo-ranking package, if peers begin using it or if insurance and health systems recommend certain scores, choosing not to participate may require an explanation. A decision that began as voluntary can become an implicit obligation through normalization pressure.
The interface is not neutral. A score coloured red looks more dangerous than one in grey. A list ordered from high to low invites the user to search for a first-place embryo. A percentile without a confidence interval looks more certain than the underlying evidence. Ethical evaluation must therefore examine the user interface as well as the algorithm. A clinical report should show uncertainty, population limits, effect size and possible pleiotropy instead of presenting one simplified rank.
The cross-national design of the Nature Human Behaviour study matters because reproductive attitudes are shaped by family expectations, health systems, social values and regulatory cultures, not scientific knowledge alone. Cross-national comparisons still require restraint: samples do not represent an entire country, and a survey scenario is not the same as a real IVF decision. The useful question is how different societies understand disease prevention, family responsibility and a child's future, and how those interpretations change the decision architecture.
Future genetic counseling should ask more than whether a person wants to know a score. A fuller consent process would explain predictive accuracy, the size of differences between embryos, population bias, effects on other traits and the option to ignore a result. It should also disclose whether the clinic will rank embryos and what defaults will be used. Consent then becomes agreement about a decision environment, not merely permission to receive data.
The central public question is not only how accurate a prediction technology is, but which information is allowed into clinical practice. Once a number is placed beside an embryo, it is no longer just scientific data; it becomes a prompt to act. Regulation should ask whether the information should appear, how it should appear, who sets the default and whether parents can refuse a category of information. More information does not necessarily mean more autonomy. Mature reproductive technology should protect the right not to be pushed by information that is unnecessary, uncertain or socially loaded.
The same principle should guide digital records and clinical interfaces. A transparent system can make uncertainty visible, preserve the reasons for a recommendation and record when a person declines a result. It can avoid visual rankings that imply a precision the model does not possess. It can also give disability communities, patients and counselors a voice in deciding which categories are harmful or misleading. Good information architecture does not remove the decision; it makes the limits around the decision harder to ignore.
International studies and professional positions can inform public debate in Taiwan, but they are not a substitute for local legal, medical and social review. Any future policy would need to examine access, insurance, counseling standards, data protection and the possibility of discrimination. The ethical baseline should remain clear: a score is an uncertain population estimate, not a verdict about a child's worth or future.
Evidence and applications
The durable lesson is that technology changes choices not only through what it predicts, but through what it makes salient. A responsible reproductive system should allow people to question a score, decline a category, see uncertainty and revise a decision without treating refusal as irrational. That is how information can support agency rather than quietly becoming a new form of pressure.
Sources:
AI use and content-safety disclosure
This English edition is an AI-assisted translation of the Chinese article, checked for source parity and evidence boundaries. It distinguishes public research and professional positions from policy interpretation and does not provide medical advice or endorse embryo selection for non-medical traits.