On genetics and justice: A reply to

“Let us make two assumptions, one a generalization of biologically ascertainable facts, the other a frank value judgement. People vary in ability, energy, health, character, and other socially important traits, and there is good, though not absolutely conclusive, evidence that the variance of all these traits is in part genetically conditioned. Conditioned, mind you, not fixed or predestined. … The second assumption is that genetic diversity is mankind’s most precious resource, not a regrettable deviation from an ideal state of monotonous sameness. … The problem is, then, not how to suppress the genetic diversity, but how to utilize it in a manner both socially advantageous and in accord with the ethical principles which we hold binding.”

(Dobzhansky 1962, p. 112).

Theodosius Dobzhansky wrote this passage in an article in Science on “Genetics and Equality,” and 60 years later, both his “generalization of biologically ascertainable factors” and his “frank value judgement” remain the objects of vigorous, and sometimes acrimonious, debate. In my book, The Genetic Lottery: Why DNA Matters for Social Equality (Harden, 2021), I consider both of Dobzhansky’s assumptions.

First, I ask: Given the tools and techniques of modern genomics, does Dobzhansky’s conclusion, that differences between people in their psychological and behavioral characteristics are genetically conditioned (“conditioned, mind you, not fixed or predestined”), still hold? Like Dobzhansky, I conclude yes. Genes cause (but do not determine) differences between people in their psychology and behavior, and these individual differences are indeed “socially important,” particularly because they help shape progress through formal education.

Second, I ask: What ought we do with this knowledge, so that it is used in a manner that is socially advantageous and in accord with the egalitarian ethical principles that I hold binding? Borrowing from the theories of John Rawls and other egalitarian political philosophers who have written about justice and luck, I describe how genetic differences can be understood as yet another form of chance in people’s lives. I then argue for the importance of “anti-eugenic” science and policy. Rather than using genetics to justify racist or classist ideas about innate superiority, I advocate for using genetics to reduce inequities in people’s freedoms, resources, and welfare. In terms of practical applications, I argue against ideas like “personalized education” and emphasize that genetics is currently most useful to social policy when it is used as a tool for improving basic research in psychology, education, and related fields.

In their review of The Genetic Lottery for this journal, Coop and Przeworski (2022) agree with my generalization of “biologically ascertainable facts.” They, too, conclude that genes influence human individual differences relevant for education (“we … fully grant the book’s starting point”; “GWAS [genome-wide association studies] undoubtedly capture some causal genetic effects”). But they disagree that this empirical finding raises any ethical problem with which to grapple: “We do not see what the field of genetics has to add to the conversation about redistributive justice.” This view is surprising on its face: As another reviewer astutely noted, “behavior genetics is a political science” (Panofsky 2021, my emphasis added).

To make the argument that genetics has nothing to add to the conversation about distributive justice, Coop and Przeworski’s review raises three substantive points, all of which have the same meta-message: “It’s complicated.” First, inequality can be understood at multiple levels of analysis (e.g., between-group, between-family, within-family) and what we learn at one level does not necessarily apply to a different level. Second, thousands of GWAS hits do not easily resolve themselves into clear mechanistic understanding. And, third, we often cannot feasibly or ethically exert experimental control over humans, complicating inferences about cause and effect.

Unfortunately, in explicating these points, Coop and Przeworski’s review distorts the arguments of The Genetic Lottery and overlooks the contributions of social scientists to the study of human behavior. A closer look at their critiques reveals that, yes, the study of genetics in relation to human behavior is complicated, but those complications do not obviate the relevance of genetics for justice. What does equality mean, in a world where people are not the same? How do we build a fair society, when people differ in their luck in life? The fact that science is difficult does not make these questions any less important. Dobzhansky’s problem is still our problem.

Response to Critique #1: What Do We Know about Within-Family and Between-Family Inequality?

Coop and Przeworski’s first substantive critique focuses on how the inferences allowed by within-family studies, which compare biological relatives and leverage the randomness of the Mendelian lottery, differ from the inferences allowed by between-family studies, which cannot fully control for genetic background nor the environmental effects that are correlated with genetic background. In particular, the review raises the possibility that the sources of within-family inequality might be distinct from the sources of between-family inequality. Results from studies that compare siblings are consistent with genetic effects on within-family differences in educational outcomes, but how then do we make the inferential leap to saying that genes are relevant for understanding differences between families? Their point is incisive, and it mirrors a point I make in The Genetic Lottery about the sources of within-group versus between-group differences. Their discussion about within- versus between-family studies, however, overlooks two important points.

First, as sociologists of inequality have long documented, over half of the economic and social inequality in America is within families, not between them (Jencks 1972; Conley 2005). Even if GWAS results ultimately inform our understanding of within-family differences in socioeconomic attainments, such within-family differences are hardly trivial.

Second, there are several methods that do leverage the randomness of the genetic inheritance from parents to children but do not depend on phenotypic comparisons between full siblings. Relatedness disequilibrium regression (RDR), for instance, was described by Dr. Przeworski in one of her previous co-authored papers as giving “unbiased estimates of direct genetic effects” (Young et al. 2019). RDR estimates the proportion of trait variation in a population that is due to genetic inheritance by testing the extent to which phenotypic similarity between pairs of people changes with their genetic relatedness (Young et al. 2018). Crucially, the RDR method depends on how related a pair is relative to the expectation based on the relatedness of their parents. This method thus cleverly leverages the randomness of Mendelian segregations, but it does not depend on phenotypic comparisons of full siblings reared together and thus is not limited to speaking only to the causes of within-sibling pair variation. Results from RDR suggest that, yes, genetic influences account for variation in educational attainment, and not just within-family variation (Young et al. 2018).

Other methods give further converging evidence. Studies that measure polygenic scores (PGS) in adopted offspring (Cheesman et al. 2020), for instance, leverage a different sort of randomness: To the extent there is lack of selective placement of adoptees, an adoptee’s PGS is decoupled from the genotype of their parents, and by extension with the environmental confounds that covary with parental genotype. As such studies typically include only one adoptee per nuclear family, observed associations between adoptee PGS and adoptee phenotypes cannot be driven by sibling contrast effects or other processes specific to sibling pairs, nor can they be attributed to the “legacy” of accumulated indirect effects.

Similarly, studies of parent-offspring trios estimate the association between offspring genotype and offspring phenotype, conditional on the parental genotype, but also typically focus on one offspring per nuclear family. Results from such studies are indeed informative about differences between nuclear families. A recent review of 38,654 parent-offspring trios or pairs across 8 cohorts found that there are “indirect” genetic effects on educational outcomes (β=0.08, 95% CI=0.07 to 0.09), but these indirect genetic effects are smaller than the “direct” genetic effects (β=0.17, 95% CI=0.13 to 0.20) (Wang et al. 2021). Again, estimates of direct effect effects from RDR, PGS-adoption studies, and parent-offspring trios do not depend on a phenotypic comparison between full siblings and cannot be waived as “just” speaking to within-family inequality.

Response to Critique #2: What Do We Know about the Psychological Development?

In Coop and Przeworski’s second major critique, they write that The Genetic Lottery “foster[s] a view of genetic causes of educational attainment as identifiable, intrinsic properties of individuals.” In fact, the book repeatedly makes the exact opposite point. Here are some representative passages:

“…the idea that the relationship between genes and social inequality is best understood at the level of a person’s cellular biology, rather than at the level of how societies organize themselves, resonates with the eugenic notion that some people are just inherently better than other people. … Such ideologically motivated talking points about the mechanisms linking genes and social inequalities can obscure the science itself” (p. 135).

“… interactions with the social environment are an essential part of the causal chain connecting genetics to psychological and social outcomes. Intelligence, curiosity, motivation, self-discipline: these do not emerge in a vacuum as some “inherent” or “inborn” property of a person’s nervous system” (p. 145).

Far from fostering a view of genetic causes of educational attainment as “intrinsic properties of individuals,” The Genetic Lottery argues that genetic differences between people always exert their effects on psychological development in interaction with environmental contexts, and that the ultimate consequences of these psychological characteristics for the human life course are dependent on how individual differences are refracted through social, political, and economic structures. This point is echoed in my other scholarly work (e.g., Raffington et al. 2020).

The review further misrepresents the themes and arguments of The Genetic Lottery by stating that: “it is hard not to interpret the book as saying that most of these GWAS effects are natural causes residing inside the brain, thereby nudging the reader towards genetic determinism.” In fact, the book contains multiple passages that directly contradict the claim of genetic determinism, and do much more than “nudge” the reader toward the opposite conclusion. Here is another representative passage:

“Hereditarian pessimism about the prospect of social change through social policy has typified eugenic thinkers for over a century. This pessimism grows out of a flawed genetic determinism, which imagines that people’s characteristics—their cognition, their personality, their behavior—are inexorably fixed by DNA. Genetic determinism is false, as the myriad studies that I’ve described in this chapter show…” (p. 171, emphasis added)

I could go on: The main text of the book is 256 pages long, and there are few themes I address more consistently than why genetic determinism is false. The passage I quote here, for instance, comes from a chapter titled, “Alternative Possible Worlds,” which is entirely devoted to the subject of gene-by-environment interaction and which gives multiple empirical examples of how genetic associations with behavior can differ according to social and political context, and in response to intervention and policy changes.

To justify their claim about the book’s genetic determinism, Coop and Przeworski point to part of a chapter where I discuss bioannotation of GWAS results. Such bioannotation results implicate—not surprisingly—the role of the brain in gene-behavior relationships. There are three problems with their discussion of this passage.

One, it is presented out of context. They don’t acknowledge in their review that this passage comes after an entire chapter devoted to explaining how causes are not necessarily deterministic, before an entire chapter devoted to explaining how even ostensibly “biological” causes can be moderated by environmental interventions, and that just a few pages later, this discussion of brain-behavior relationships is situated within the context of environmental transactions (“interactions with the social environment are an essential part of the causal chain connecting genetics to psychological and social outcomes”).

Second, the review does not use an accurate definition of “genetic determinism.” The American Psychological Association’s definition of “genetic determinism” is as follows: “the doctrine that human and nonhuman animal behavior and mental activity are largely (or completely) controlled by the genetic constitution of the individual and that responses to environmental influences are for the most part innately determined” (American Psychological Association n.d.). Nothing in The Genetic Lottery regarding the possible neurobiological mechanisms connecting genes with behavior states or implies that behavior is largely or completely controlled by genes, or that responses to environmental influences are innately determined. Thus, Coop and Przeworski’s review focuses on a short passage that might appear problematic only when taken out of context and then misdiagnoses the apparent problem.

Finally, their critique overlooks a vast literature on children’s psychological development, which informs our understanding of how genetic differences are connected to educational and socioeconomic life course outcomes. They, quite remarkably, claim that “we currently know next to nothing about the causal paths from GWAS findings to educational attainment,” because “all we actually have, at present at least, is a large number of genetic associations, individually of tiny effect, and a statistical enrichment for a tissue that makes sense for a behavior.” This might be all that we have from GWAS but, fortunately, our understanding of human development is informed by multiple scientific fields and methods, not just statistical genetics.

Developmental research has clearly identified a number of psychological processes important to children’s academic performance—including cognitive processes such as visuospatial and verbal reasoning, executive functions, processing speed, as well as “noncognitive” self-regulatory processes such as grit, motivation, frustration tolerance, impulse control, and intellectual curiosity (e.g., von Stumm et al. 2011; Tucker-Drob et al. 2016; Malanchini et al. 2018). These psychological characteristics show increasing heritability over time, which emerges in transaction with environment contexts (Tucker-Drob et al. 2013). It comes as little surprise, then, that an GWAS hits for educational attainment are associated with these same psychological characteristics, both between and within families (e.g., Krapohl et al. 2016; Selzam et al. 2019; McGue et al. 2020; Demange et al. 2021). Although we certainly have much more to learn, it is doubtful that any credible account of the causal paths from genetic differences between people to educational inequality will somehow circumvent psychological differences between people, or that these psychological differences will somehow not involve the brain.

Response to Critique #3: What Counts as Meaningful?

Coop and Przeworski’s third major critique is that my book relies on a “typological view of ancestry groups,” when, “in reality, there is no bright line demarcating comparisons ‘within’ versus ‘between’ ancestries: there is giant family tree of humanity, and people who share more ancestral paths through it than other, and more similar environments than others.” They are clearly correct, and I indeed make this same point in my book, citing Coop:

“As the population geneticist Graham Coop summarized, ‘Your family tree is vast and vastly messy; no one is descended from just one group of people.… Zoom closer in on an individual person, and the clarity you had from a distance dissipates. Wherever there are boundaries, there are people whose histories stretch across those boundaries.” (p. 75)

They go on to discuss the implications of the fact that human “populations” do not have clear boundaries, but rather are genetically and environmentally heterogeneous:

“Since human groups are never compared in an experimental setting or in randomized environments, nothing ensures that environmental effects are the same across groups. … In the end, the key difficulty is … all the confounding factors that exist in the absence of the ability to do experiments and how well we can measure and control them.”

Here, Coop and Przeworski have indeed identified the “key difficulty” in the study of human behavior. But, crucially, this key difficulty is not at all unique to the study of genetic influences on behavior. The study of environmental influences is every bit as difficult. For many environments of interest, experimentation is impossible. Even when true experiments are done, randomization can be compromised, assuming the generalizability of the estimated treatment effects is a dicey proposition is best, the mechanisms that instantiate the causal effects are unknown, and unexplained heterogeneity in treatment response is the norm rather than the exception (Heckman et al. 2010; Deaton and Cartwright 2018; Bryan et al. 2021).

Thus, the problems that Coop and Przeworski identify in the study of genetic influences of human behavior—that one must make the “big jump” from studies (like within-family GWAS) that have good internal validity but uncertain external validity, that studying humans is a “tangled mess”—beset nearly all research on the causes and consequences of human behavior, genetic and environmental. These are not the challenges of social science genetics; these are the challenges of social science.

Because of the difficulty of that “big jump,” Coop and Przeworski conclude that “current PGS for educational attainment are neither interpretable nor meaningful.” Further, in light of the limitations of nonexperimental social science, Coop and Przeworski characterize my argument that PGSs for human behavioral phenotypes are simultaneously meaningful, by virtue of the fact their associations replicate across a variety of quasi-experimental methods, and limited, by virtue of their contextual dependence and imperfect portability, as “trying to have it both ways.” To which I can only ask: What correlate of educational attainment would qualify as interpretable and meaningful, according to the criteria they lay out? Certainly, none of the classic environmental predictors of educational outcomes rise to that standard.

As the author Maggie Nelson wrote: “There is a lot to be learned from having it both ways” (Nelson 2016). Nearly all of the causal regularities discovered by social scientists are probabilistic rather than deterministic, exception-ridden rather than universal. Some even argue that “laws of nature” do not exist for social phenomena, as human societies are too dynamic and complex to allow for causal laws that are universally true. Given this complexity, many of us in the social sciences find the words of one Nobel-prize winning economist particularly resonant: “Better LATE [local average treatment effects] than nothing” (Imbens 2010).

What Genetics Has to Add to the Conversation about Justice

On the basis of these three critiques, Coop and Przeworski conclude that they “do not see what the field of genetics has to add to the conversation about redistributive justice, beyond confirming what has long been recognized–that life outcomes differ for all kinds of reasons beyond people’s control.”

But, who, exactly, has long recognized that life outcomes differ for reasons beyond people’s control? Certainly not the average American. The United States is one of the most inequality-tolerant countries in the world. American conservatives, in particular, are global outliers in their unwillingness to say that income inequality is even a problem to be solved. And, Americans’ tolerance for gaping inequalities in income and wealth is buttressed by their beliefs about the origins of inequality. Far from recognizing the role of luck, whether it be constitutive or environmental, Americans who oppose efforts to narrow income inequalities are more likely to see inequality in starkly moralistic terms: They believe that the rich are richer because they work harder, delay gratification, and take more risks (Rothwell 2020). This American tendency to see poverty and affluence in terms of just deserts—who deserves to be financially “rewarded” or “punished”—has material consequences for millions of people. Coop and Przeworski agree with me that findings from genetics help confirm what some people have long believed: “In the end, luck swallows everything” (Strawson 1998). But this belief is far from the predominant view in the American political conversation.

Coop and Przeworski underestimate the possibility that public understanding of genetics has any possibility of being effective at changing people’s attitudes about the causes of inequality (“we very much doubt that overstating our understanding of the genetics of behaviors is going to increase empathy”). This statement positions the relationship between genetic knowledge and attitudes as a matter of personal faith and doubt, rather than a psychological hypothesis that has been the subject of scholarly research. As I describe in the book, research has shown that “understanding of the genetics of behaviors,” ranging from sexual orientation to body size to serious mental illness, can indeed increase people’s empathic attitudes, reducing ascriptions of blame and increasing support for civil rights (Kvaale et al. 2013; Garretson and Suhay 2016).

Finally, I am wary of the review’s conclusion that “it matters greatly” whether the causal paths connecting genetics to educational attainment are “Jencks-style red hair effects,” aka, physical phenotypes such as skin or hair color that society discriminates against, rather than mediated through the development of neurocognitive phenotypes, such as executive functions and visuospatial reasoning. Does our responsibility to create a society where people share more equally in power, resources, health, leisure, self-determination, and freedom change, depending on whether genes exert their effects through hair color versus cognitive ability? Are people’s claims on society for their inclusion and participation, regardless of difference, any less valid if the difference is psychological versus physiological? I think not. If our egalitarian commitments to “elucidating” and “redressing” social inequality depend on which individual differences mediate genetic effects on education, I fear they are very unstable indeed.