Illustrative image for Genomics Behavior and Social Outcomes

From Bioethics Briefings

Genomics, Behavior, and Social Outcomes

  • Within the past decade, sequencing of the human genome and the rapid development of large-scale DNA testing technology has given rise to a new era of genomic research.
  • Genome-wide association studies (GWAS) are used to scour large swaths of human DNA and identify very small associations between genetic variants and traits such as height and weight.
  • Polygenic scores (PGS), which are derived from GWAS results, permit modest genetic prediction of traits and outcomes from small samples of blood or saliva.
  • In addition to basic physical traits, such as height and weight, researchers are now using GWAS and PGS to investigate complex behavioral traits and social outcomes, such as intelligence (IQ), educational attainment, same-sex sexual behavior, and household income.
  • Genomic research on complex behavior and outcomes generates controversy, engendering myriad ethical concerns about the potential misuse or misunderstanding of genetic results; commercialization of genetic material; genetic essentialism; discrimination; and the use of genomic information to inform social, public, and educational policy.
  • In the interest of minimizing misunderstanding and misapplication of genetics and genomics findings, researchers are now publishing Frequently Asked Questions (FAQs) with their study results. The FAQs offer accessible explanations of the scope and limitations of their research and help clarify their findings and prevent misconceptions.

Framing the Issue

The completion of the Human Genome Project in 2003 ushered in technological advancements that have made genetic information more accessible to researchers and the public than ever before. While scientists have traditionally studied human genetics using twin, family, and adoption studies, the rapid development of DNA-testing technology has launched the new era of genomics. Instead of tracking observed differences between genetically related individuals, such as identical twins (who share the same genes) and fraternal twins (who share some genes), researchers are now using genome-wide association studies (GWAS). GWAS scan large swaths of DNA for genetic variants–namely, single nucleotide polymorphisms (SNPs)–that are statistically associated with a variety of traits and outcomes. GWAS results, in turn, allow for the construction of polygenic scores (PGS), which calculate the miniscule effects of thousands of genetic variants and allow for genetic prediction about particular traits or outcomes from small samples of blood or saliva.

While the application of GWAS and PGS to simple physical traits, such as height and weight, are more widely accepted, important and difficult ethical questions arise in light of recent applications of these genomic techniques to complex behavioral traits and outcomes that are known to be influenced by social factors. Scientists have conducted GWAS of socially fraught characteristics such as intelligence, educational attainment, social class mobility, same-sex sexual behavior , household income, mathematics ability, and reading ability. Pending research  is looking at externalizing behavior, a term for problems in self-regulation that include aggression, delinquency, and hyperactivity.  Moreover, some researchers have called for the integration of genetics into social sciences research to better understand the association of genetics with behaviors and socioeconomic outcomes, advocated for precision education, and even founded companies  that intend to empower prospective parents to use in vitro fertilization to deselect fertilized eggs with genetic traits that are thought to increase the odds of low intelligence or short stature.

Given the sensitivity and social complexity of these studies, researchers have been  engaging with press and publishing Frequently Asked Questions (FAQs) in conjunction with genomic studies. FAQs are designed to help general audiences understand the scope and limitations of the research and dispel common misconceptions of genetics. For example, in a brief FAQ on a genomic study of mathematics ability, authors emphasize that “Polygenic scores are NOT fortune-tellers,” nor are they “free of environmental or social processes.”

We created a repository of FAQs on Human Genomic Studies.

Ethical, Legal, and Policy Considerations

Genome-wide association studies and polygenic scores associated with highly complex, socially mediated behaviors and outcomes provoke a host of ethical, legal, and policy considerations, including:

  • Genetic essentialism and determinism
    • Some express concern that individuals tend to interpret genetic information with an “essentialist bias,” i.e., the belief that one’s genes define their essence or identity. Thus, genomic research on characteristics such as intelligence and educational attainment may leave some inclined to believe that their intellectual or educational abilities are immutable features of their genetic identity.
    • A related concern is that individuals tend to think of genes in fatalistic and deterministic terms. Thus, genomic research on complex behavioral traits may incline individuals to mistakenly believe that genes have strong causal effects that inevitably determine one’s future. Yet today geneticists routinely emphasize that the influence of genes is not immutable. For example, consider myopia, which is strongly influenced by genetics. Individuals with myopia are not destined for a life with poor eyesight; prescription eyeglasses provide an easy remedy.
  • Genetic discrimination and stigma
    • Health Care: Genomic research on complex behaviors could give rise to discrimination in health care. The Genetic Nondiscrimination Act of 2008 (GINA) makes it illegal to discriminate against genotype in health insurance. However, it does not provide protections for long-term care, disability, or life insurance. A life insurance company could decline an individual’s request for a policy after reviewing a direct-to-consumer test result indicating the applicant has a low-percentile polygenic score for income.
    • Education: Genomic research on complex behaviors could give rise to discrimination in schools.  Given that there have been proposals to conduct precision education through the development of “genetically sensitive schools” it is conceivable that polygenic scores for educational outcomes could be used to discriminate in admissions processes, denying admission to prospective students who do not have high-percentile genomic test results. Furthermore, institutions that are averse to spending money on learners with genetic predispositions to a host of costly developmental disorders could argue that financial constraints prevent them from serving all students. The disclosure of genomic information in educational settings could also result in the stigmatization of students.
  • Commercialization of polygenic scores
    • The consumer genetics movement is catalyzed by companies like 23andMe, signifying the consumerism around genetic testing and screening and the desire for self-discovery via direct-to-consumer (DTC) genetic testing. The U.S. government provides limited regulation of DTC.
    • Companies have begun offering health-related PGS tests to consumers, including predictive tests for heart disease and diabetes. This is despite ongoing debate over the validity and clinical utility of PGS.
    • Gene report marketplaces (e.g. GenePlaza, Genomelink, Promethease, MyGeneRank,, and DNA.Land) offer predictions for traits like math ability or depression based on unvalidated use of GWAS data.
    • Companies have started to market the use of polygenic scores for embryo selection in in vitro fertilization.
  • Privacy considerations
    • Genetic information is deeply personal. Privacy advocates have questioned whether DNA samples can ever truly be deidentified. Learn more.
  • Eugenics and racism
    • The development of PGS has reinvigorated long-standing debates about the viability of creating “designer” babies and the genetic bases of racial differences in intelligence. Bioethicists are concerned with questions about embryo selection using PGS. Moreover, just as researchers in the past appealed to traditional genetics to support racist and eugenic claims and policies, some people—including some scientists – are using GWAS results and PGS to rekindle old programs of controversial race science.

Understanding the Predictive and Explanatory Limitations of Social and Behavioral Genomics

Given potential ethical, legal, and social impacts, it is of the utmost importance for bioethicists, policymakers, academic researchers, and the general public to be familiar with the predictive and explanatory limitations of GWAS and PGS. Many of the ethical concerns about genomic research on human behavior, particularly the use of polygenic scores to predict individual outcomes, hinges on the scientific validity of the research in question. Here we’ll briefly review some of the key limitations of the research to help readers accurately and responsibly interpret and understand results of behavioral genomic research results.

Predictive limitations: problems of portability and individual prediction

Given that PGS can be used to try to predict behavioral traits and social outcomes from samples of DNA, it might be tempting to think that genomic research has identified genes that strongly determine these complex characteristics. This, however, is not the case, as the predictive power of genomic information is correlational and highly probabilistic. Although predictive accuracy is expected to improve with time and larger studies, PGS currently explain very modest amounts of variation. In a GWAS of over one million participants, PGS for educational attainment explain no more than 15% of the variance. Put another way, on a scale of predictive accuracy from 0 (not predictive at all) to 10 (perfectly predictive) PGS for education are no more than 2. That means that the vast majority of variation in educational attainment (>80%) is not predicted by PGS. Given this very limited predictive capacity, many scientists emphasize that PGS cannot accurately predict the outcomes of individuals, and are better understood as risk factors. Rather, PGS are more appropriately predictive of average differences in a single population for a given behavior. For instance, on average, persons with high-percentile PGS for IQ are likely to score higher on IQ tests than persons with low-percentile PGS for IQ. Given the limited accuracy of these scores, however, this means that there will be many individuals with low-percentile PGS who score very high on IQ tests,  and vice versa. Simply put: PGS are not useful predictors for individuals.

Understanding the limited predictive capabilities of PGS calls for attention to an emerging issue known in the genomics community as the “problem of portability”: PGS scores don’t “port,” or generalize, from genetic characteristics of the original study to novel populations. That is, PGS are most predictive in populations whose genetic profile is very similar to the original GWAS study and least predictive in populations whose genetic profile is very different from the original GWAS. For example, PGS derived from GWAS of European ancestries participants are most predictive of European DNA samples, and least predictive of African-ancestries DNA samples.

Yet issues of portability go beyond limitations for predicting across populations of different genetic ancestry to within-population generalizability. Here is a brief overview of the various ways in which PGS are subject to problems of portability and confounded by non-genetic factors:

Ancestry: The genetic ancestry of an original population study from which PGS are derived influences the predictive accuracy of PGS across populations with different genetic ancestries. For example, PGS derived from genomic studies of European ancestry populations will be most accurate in European ancestry DNA samples, and least accurate in non-European ancestry samples, such as those from individuals of African or Asian descent. Given that the vast majority of GWAS to date have been conducted in individuals of European ancestry–as of 2016, over 80%–PGS are currently most accurate in people of European ancestry. 

Age: The age of the participants in an original GWAS study also influences the predictive accuracy of PGS within populations of the same ancestry. For instance, PGS for body mass index (BMI) derived from samples of young participants are less predictive of BMI in older participants.

Sex: The biological sex of participants in an original GWAS study also influences the predictive accuracy of PGS within populations. For example, PGS for diastolic blood pressure derived from a GWAS of woman are less accurate in men.

Environment: Environmental factors associated with an original GWAS population sample also contribute to the limited portability of PGS. For example, PGS for educational attainment derived from participants of low socioeconomic status (SES) are less predictive in high SES participants.

Explanatory Limitations:  Biological Etiology and Missing Heritability

Growing enthusiasm for GWAS and polygenic prediction of complex behaviors and outcomes does not mean that scientists today understand, or have meaningfully elucidated the causal relationships between, DNA and human behavior. While many geneticists are optimistic that they will eventually find genetic variants in GWAS data that cause particular behaviors or outcomes and help explain biological mechanisms and etiologies of behavior, even the most cutting-edge results today are fundamentally correlational. For instance, although GWAS have discovered over 1,200 genetic variants statistically associated with differences in educational outcomes, researchers have not identified specific biological mechanisms that would explain why some individuals are more successful in school than others. Moreover, there is some concern that the genetic variants identified may be attributable to noncausal, random genetic variation due to a phenomenon known as population stratification. Individuals with shared genetic ancestry (e.g., from the same ancestral region) may be more likely to share values, cultural practices, or exposure to other unobserved environmental experiences. Population stratification can give rise to associations driven by shared environmental factors which are incorrectly attributed to the shared genotype.

Put another way, although large-scale genomic studies can be used to identify statistical associations between DNA and human behavior, they have yet to explain human behavior in any meaningful sense of the term. One’s intelligence or success in school cannot be reduced to one’s DNA. Findings on the genomics of behavior contrast with many examples across the sciences in which researchers have elucidated meaningful biological pathways and mechanisms linking genes to traits, such as phenylketonuria, a metabolic disorder, or macular degeneration.

Finally, the development of GWAS have given rise to a conceptual issue known as the “missing heritability problem”: for any given behavioral trait or outcome, estimates of heritability derived from genomics are significantly lower than estimates of heritability derived from traditional genetic methods (e.g., twin studies). Although “heritability” is often interpreted as a measure of genetic influence on a trait, it actually refers to a statistical analysis of the amount of variance in a trait. When one calculates heritability of the same behavioral trait using traditional and genomic tools, the outcomes are quite different: genomic heritability (estimated by SNPs) is significantly lower than traditional “twin heritability.” The missing heritability problem raises questions about the meaning of genomic results and their implications for advancing scientific understanding of the genetic influences on human behavior.

It’s also worth noting that the mere notion of heritability itself, whether derived from twins or SNPs, can be misleading. Given that heritability is most commonly treated as a measure of genetic influence, even in scientific publications, it can be tempting to believe that highly heritable traits are strongly influenced by genes, or even determined by them. This is not the case, as even the most highly heritable traits are not immutable. Height, for example, has shown a consistently high twin heritability of more than 90%, yet an individual’s height is also influenced by environmental factors, such as nutrition, during early stages of development. Heritability does not equal destiny.

Moving Forward

Although genetic research on human behavior is not new, the scientific techniques, methods, and technologies are advancing rapidly. The transition from traditional twin and family genetics to social and behavioral genomics has ushered in a host of new ethical, legal, and social challenges. We believe that there are two important issues to be addressed by bioethicists, policymakers, and social scientists to ensure that behavioral genomic research is done responsibly. The first issue regards communication: how can experts communicate to nonexperts, especially the general public, the results of controversial genomic research in a manner that minimizes harm? Given the potential for social and behavioral genomic studies to be misinterpreted with genetic essentialist biases, we believe it is of the utmost importance that researchers investigate methods of communicating their findings to broader audiences that help avoid misconceptions or misapplications.

The second issue regards future genomic studies. To minimize potential harm, bioethicists, researchers, and policymakers should consider whether there  are genomic studies in which the draw of scientific curiosity simply fails to outweigh potentially negative consequences.

In short, growing interest in the genomics of socially complex behaviors and outcomes raises a host of important social and ethical considerations. Bioethicists, policymakers, researchers, and the greater society will need to grapple with the implications of social and behavioral genomics for public services such as health care and education while also working to proactively safeguard against genetic essentialism, determinism, discrimination, and stigma. Socially responsible communication and widespread awareness of the methodological limitations of GWAS and PGS are key to ensuring that genomic research is done responsibly. For social and behavioral genomics to benefit all and not just some, the research community will need to come together to consider their social and ethical responsibilities as stewards of knowledge.

Daphne O. Martschenko, PhD, is a postdoctoral fellow at the Stanford University Center for Biomedical Ethics. Lucas J. Matthews, PhD, is a postdoctoral researcher at The Hastings Center and the Columbia Center for Research on Ethical, Legal & Social Implications of Psychiatric, Neurologic & Behavioral Genetics. 

  • Eric Turkheimer, PhDHugh Scott Hamilton Professor of Psychiatry, University of Virginia
  • Erik Parens, PhDSenior Research Scholar, The Hastings Center
  • Ben Domingue, PhDAssistant Professor, Graduate School of Education, Stanford University
  • Sam Trejo, PhDAssistant Professor of Sociology, Princeton University