Currently viewing a development environment

The Genetic Lottery is a bust for both genetics and policy

Kathryn Paige Harden’s book tries to demonstrate how genetics can ameliorate societal ills. She falls well, well short

Kevin Bird

Evolutionary Biology

Michigan State University

The last decade has seen genetics and evolution grapple with its history; one composed of figures who laid the foundations of their field while also promoting vile racist, sexist, and eugenicist beliefs.

In her new book, The Genetic Lottery, Kathryn Paige Harden, professor of psychology at University of Texas at Austin, attempts the seemingly impossible task of showing that, despite a history of abuse, behavioral genetics is not only scientifically valuable but is an asset to the social justice movement.

In this attempt, she fails twice. For the first half of the book, Harden tries to transform the disappointment of behavioral genetics in the years following the Human Genome Project into a success that proves that genes are a major and important cause of social inequality, like educational attainment or income levels. In the second half, she tries to show that this information is not a justification for inequality, rather it is a tool to use in our efforts to make society more equitable and cannot be ignored if we wish to be successful. To say the least, this section too falls short. Harden refuses to engage with the history and trajectory of her field, and ultimately the science fails to uphold the idea that not considering genetic differences hinders our attempts to create a more equitable world.

In the book Misbehaving Science, sociologist Aaron Panofsky documents the history and progression of behavioral genetics, from its formal inception in the 1960s. Throughout its history behavioral genetics has responded to criticism in a variety of ways.

In 1969, the educational psychologist Arthur Jensen used behavioral genetics methods to argue that IQ gaps between white and Black Americans had genetic origins and, therefore, could not be remedied by educators or social policy. As criticism from mainstream geneticists and evolutionary biologists tied Jensen and behavioral geneticists to each other, the field attempted to hold a middle ground between Jensen’s racist conclusions and the belief that human behavioral genetics was fundamentally flawed. However, in this attempt to preserve their field from criticism, behavioral geneticists progressively defended the importance of race science research and adopted some core premises about the influence of genetic differences on the racial IQ gap.

In the following decades, Jensen and like-minded researchers like J. Philippe Rushton, Richard Lynn, and Linda Gottfredson received funding from the Pioneer Fund, an organization explicitly dedicated to “race betterment.” All the while, they were integrated into editorial boards of journals that published behavioral genetics work and treated as colleagues. Even mainstream behavioral genetics work like the Minnesota Study of Twins Reared Apart and the Texas Adoption Project would receive funding from the noxious Fund.

In attempts to justify their field against continued criticism, behavioral geneticists themselves used twin study results to argue social interventions would be ineffective. As Panofsky wrote:

“Behavior geneticists’ polemical style of valorizing their re­search led them to plant deep stakes that were tightly clustered around a particular, basically genetic determinist, interpretation.”

This history, including behavioral genetics' own role in generating, promoting, and defending scientific racism and determinist views of genetics is completely absent from Harden's book. This history matters; it is the source of the isolation of behavioral genetics from mainstream genetics research. This isolation has produced the intellectual and ideologically stagnant lineage that Harden operates in.

These biases are most pronounced in the early chapters walking readers through the science, which often leads to an incomplete, misleading, or mistaken account of genetic research and behavior. Harden presents an argument about the major causal role of genetic differences. These results span decades, including twin studies, and recent developments like genome-wide association studies (GWAS), polygenic scores (a single value combining individual estimated effects of genome-wide variations on a phenotype), and genomic analyses of siblings. Unfortunately, Harden often gives these results in such a misleading way that it obscures how damaging they actually are to her own core thesis.

For example, Harden extols sibling analyses as unassailable evidence of independent, direct genetic causation free of biases found in other methods. While it’s true that polygenic scores from sibling analyses resolve substantial problems that sometimes create inaccurate associations between DNA and a phenotype, Harden fails to mention several key differences between these sibling-based methods and other genomic or twin-based methods. It is rarely stated clearly that these family methods produce much smaller estimates of genetic effect, often nearly half the size as population-based methods, making the 13% variance explained by current education polygenic scores a likely overestimate. Harden also fails to mention that a commonly used method employed does not fully eliminate the problems from population structure or that estimates from siblings can still include confounding effects that create correlations between genes and environment.

Even worse, Harden moves between the less biased, but smaller, results from sibling methods to the more biased but larger estimates from population-based polygenic scores without being clear this is what she is doing. This happens frequently when discussing research claiming that educational polygenic scores substantially explain differences in income. The result is Harden obscures the fact that more reliable techniques result in lower predicted genetic effects. Readers may be wrongfully led to believe genetic effects are both large and reliable when in reality they are more often one or the other.

Harden’s failure to engage with critics of behavioral genetics, often from the political left, veers between simple omissions and outright misrepresentation. This treatment is in stark contrast to how she treats biological determinists on the political right. The work of Charles Murray, the co-author of The Bell Curve, which claimed that differences in IQ scores between the rich and poor were genetic, and whose research aligns neatly with Harden’s, is described as mostly true and his political implications are lightly challenged. The most prominent critic of behavioral genetics, Richard Lewontin, gets much rougher treatment.

In one of the three cases in which Harden bothers to mention Lewontin’s decades-long engagement with behavioral genetics, she gets it wrong, claiming that Lewontin merely said that heritability is useless because it is specific to a particular population at a particular time. In reality, Lewontin showed why the statistical foundation of heritability analyses means it is unable to truly separate genetic and environmental effects. Contra Harden’s characterization of her opponents, Lewontin recognized genetic factors as a cause of phenotypes; however, he stressed their effects cannot be independent of environmental factors and the dynamics of development.

Harden implies that giving people access to equal resources increases inequality and genetic influence. Lewontin explained why the outcome of equalizing environments precisely depends on which environment you equalize. As a toy example, a cactus and a rose bush respond differently to varying amounts of water. Giving both plants the same, small, volume of water is good for the cactus’s health and bad for the rose, giving both a larger volume of water is bad for the cactus and good for the rose. Equalized environments regardless of quality can reduce or increase inequality and can reduce or increase the impact of genotypic differences depending on the environment and the norm of reaction for a trait and set of genotypes. Heritability analyses cannot provide insight on this distribution or nature of genotype and environment interactions. These detailed, quantitative, and analytic arguments are entirely ignored by Harden.

In her story, people on the political left are ideologically driven to oppose behavioral genetics because they believe it invalidates their desire to ameliorate inequality. In the powerful book-length criticism of behavioral genetics, Not in Our Genes, Lewontin, with neuroscientist Steven Rose and psychologist Leon Kamin, all socialists, defy Harden’s characterization of her critics from the left, writing:

“The antithesis often presented as an opposition to biological determinism is that biology stops at birth, and from then on culture supervenes. This antithesis is a type of cultural determinism we would reject… Humanity cannot be cut adrift from its own biology, but neither is it enchained by it.”

They further write:

“Against this we counterpose a view not of organism and environment insulated from one another or unidirectionally affected, but of a constant and active interpenetration of the organism with its environment. Organisms do not merely receive a given environment but actively seek alternatives or change what they find.”

Not in Our Genes criticizes biological determinism for oversimplifying the processes that create diversity in the natural world. And the ways that biological determinism is employed for political and ideological reasons by people like Arthur Jensen, Daniel Patrick Moynihan, or Hans Eysenck, to undermine movements for social and economic equality on the basis of biological data. Lewontin, Kamin, and Rose did not oppose biological determinism simply on ideological grounds. They knew there was no true threat to egalitarian beliefs posed by biological data if one properly understands biology in a non-determinist way. Instead, they wanted to move beyond just a scientific critique and provide a social analysis of why the mistakes of biological determinism are made, persist, and gain in popularity. They write:

“The errors of the biologi­cal determinists’ explanation of the world can be explicated and under­stood without reference to the political uses to which these errors have been put. A large part of what follows in this book is an explication of these errors. What cannot be understood without reference to politi­cal events, however, is how these errors arise, why they come to characterize both the popular and scientific consciousness in a particular era, and why we should care about them in the first place.”

This lack of meaningful engagement with critics is not just poor scholarship, it weakens Harden’s case. Problems arise with Harden’s discussion of heritability, for example, which would be remedied with a genuine engagement with critics from mainstream genetics and evolutionary biology. Harden takes a hardline position that heritability is a measure of genetic causation within a sampled population; however, despite her attempt over two chapters to build this case, she is still fundamentally mistaken about the concept.

Early work in plant breeding and genetics can help shed light on the source of this confusion. The pre-eminent statistical geneticist, Oscar Kempthorne, in a 1978 critique of behavioral genetics, wrote that the methods employed by the field can tell us nothing about causation because all they really represent is simply a linear association between genetics and phenotypes, without any further ability to connect the two to each other.

The extent to which correlations can be interpreted as causation depends on properly controlling for confounding variables. In the context of heritability, this means that genetics and environment need to be independent of each other, but this cannot be the case without direct experimental manipulation. In fields like plant breeding, it is possible to experimentally randomize which environments a plant genotype experiences, and genetically identical plants can be put in different environments for extra control, so these inferences are safer to make. In human genetics, however, this is not possible even with the sibling and twin methods Harden focuses on. These processes that complicate causal interpretation of heritability estimates have been discussed ad nauseum by other behavioral geneticists, which is why Harden is one of the few who comes to her conclusions.

One final glaring omission worth noting occurs in Harden’s chapter on race and findings of behavioral genetics. Here, Harden does an admirable job trying to prevent the misapplication of behavioral genetics to questions of racial differences. Surprisingly absent though is the fact that across a variety of studies, genetic variation is much larger within races compared to between races. This finding undermines core perceptions about the biological nature and significance of race. It also has important implications for our assumptions about the role of genetics in phenotypic differences between races, namely that they will be small to nonexistent. One could speculate the omission is because the finding was from none other than Richard Lewontin. This case is particularly problematic because in randomized control trials, biology classes emphasizing Lewontin’s findings have shown very strong evidence of reducing racial essentialism, prejudice, and stereotyping. Few science education interventions against racism and prejudice have such strong evidence in their favor.

Above all, Harden desperately wants to impart one idea in the first part of the book: genes cause social inequality. Here she argues for causation as “differences makers” in counterfactual scenarios. In other words, X causes Y if the probability of Y occurring is different were X not to happen. As Harden notes, experimental science adopts a similar and in ways stronger,  “interventionist theory” of causation, based around experimental interventions. Here X is said to cause Y if there is a regular response of Y to an intervention on X.

Under the interventionist theory, Harden’s account of genetic causation runs into trouble. First, it requires us to be able to isolate a specific property on which we can intervene. This is possible in cases of simple genetic disorders with clear biological mechanisms and short pathways from gene to trait, like sickle cell anemia or Tay-Sachs. However, this doesn’t work for behaviorally- and culturally-mediated traits involving large numbers of genes, with small effects and diffuse associations between genetic and non-genetic factors. There is simply no method to isolate and intervene on the effects of specific genetic variants that holds environmental factors constant in a way we would normally recognize as an experimental intervention. This applies still to the sibling analyses that Harden tries to portray as randomization experiments. Contrary to one of Harden’s more bizarre claims, meiosis does not approximate a randomized experiment. All it does is randomize genotypes with respect to siblings, it does not randomize environments experienced by genotypes. Our broad array of social and cultural institutions still acts in a confounding way. Instead, we just have a polygenic score, which is more a statistical construct than a tangible property in the world.

Second, for Harden’s causal claims to hold weight, genetic and environmental factors must be distinct components that are independently disruptable. This reflects what the philosopher John Stuart Mill called the principle of the composition of causes, which states that “the joint effect of several causes is identical with the sum of their separate effects.” At the core, Harden assumes that genetic and environmental influences on human behavior are independent and separable. To say the absolute least, this is a highly dubious assumption. Based on the arguments from critics like Lewontin and the work from research programs like developmental systems theory, there is very good reason to think that biological systems are not modular, especially in the case of educational attainment. Genetic and environmental influences interact throughout development, the interactions are dynamic, reciprocal, and highly contingent. It simply isn’t plausible to estimate the independent effect of one or the other because they directly influence each other.

A further weakness of Harden’s book is that just because genes make a difference in phenotype, it does not mean that genes are even relevant to the analysis of these phenotypes. In reality, Lewis’s account of causation, that X is a cause if a different outcome would have occurred in the absence of X, can be a pretty low bar, and the causes it identified may not be very relevant. An obviously absurd example is that the argument could be made that the sun caused me to wake up this morning since it is the origin of the trophic cascade that nourished my body enough to continue necessary biological functions. Under Lewis’ account, the sun is a cause of my waking up, but it’s hardly a relevant or informative cause compared to my alarm clock or to the bus I need to catch at 8:35am.

In Biology as Ideology, Lewontin discusses the causes of the disease tuberculosis. He notes that in medical textbooks the tubercle bacillus, which gives people the disease when infected, is the cause of tuberculosis. Lewontin writes that this biological explanation is focused on the individual level and treats the biological sphere as independent from external causes related to the environment or social structure. While we can surely talk about the role of the tubercle bacillus in causing the disease we can also talk about the social conditions of unregulated industrial capitalism and its role in causing outbreaks and deaths by tuberculosis and can gain far more insight by analyzing the causes of tuberculosis in that way.

“...there have been complex social changes, resulting in increases in the real earnings of the great mass of people, reflected in part in their far better nutrition, that really lie at the basis of our increased longevity and our decreased death rate from infectious disease. Although one may say that the tubercle bacillus causes tuberculosis, we are much closer to the truth when we say that it was the conditions of unregulated nineteenth-century competitive capitalism, unmodulated by the demands of labor unions and the state, that was the cause of tuberculosis.”

This distinction of whether a cause is relevant for particular social and scientific issues becomes a problem for Harden in the climax of her book where she tries to convince the reader that genetic information is a crucial tool for addressing social inequality.

One example given by Harden is that children who perform well but are in poor schools are able to “achieve” less, and that poor people with higher education end up making less money than rich people in the same fields. These findings are neither novel nor do they require the use of potentially misleading genetic data. While Harden tries to defuse right-wing arguments about shortcomings of social science research, this isn’t a given. As research Harden herself presents shows, results from behavioral genetics bolster the far right and they regularly share this research to promote their beliefs and challenge egalitarian policies. Instead of engaging with this bad-faith criticism from the right, we can simply disregard them, just as Harden disregards their co-option of her field of research.

Finally, Harden expresses a general concern that social science and psychological studies are plagued by “genetic confounding,” that is the correlations they observe are actually due to unconsidered genetic forces that relate an individual to their outcome (i.e. low income doesn’t cause poor health, genes cause both low income and poor health). For this example, Harden is hard on these complaints, equating research that does not include genetic information as tantamount to robbing taxpayers, but light on evidence that this genetic confounding is a widespread problem, or that it can only be addressed with behavioral genetic research.

Surprisingly, all these examples abandon the earlier bluster about genes being crucial causal factors in our life and instead opt for genetic data as one of many methods for causal inference of environmental interventions. We no longer care about heritability estimates; instead, we use twins as an experimental design. In some cases this is fine, however using individuals who have similar genotype, environmental characteristics, and phenotype does not mean that genes are significant causes, it’s just a good experimental design. Here, some of Harden’s arguments about social science research are accurate. Observational and correlation-based studies are weak for a number of reasons, not simply because they ignore genetic differences. The goal should be strengthening causal inference in the social sciences, and we have some idea of how to do that from other fields. To strengthen the ability to identify causes, epidemiologists employ direct experiments, like randomized control trials, exploit “natural experiments” that can approximate experimental randomization, such as studies that observe changes in outcome shortly after changes in government policy are enacted, or designs that use statistical methods to match people based on background demographic information like income, neighborhood quality, family education, etc.

In fact, there are principled reasons to think genetic data has little to no benefit above and beyond the kinds of data we can collect from non-genetic social science experiments. Eric Turkheimer, Harden’s doctoral advisor, has articulated the “phenotypic null hypothesis” which states that for many behavioral traits the genetic variance identified from behavioral genetics studies is not an “independent mechanism of individual differences” and instead reflects deeply intertwined developmental processes that are best understood and studied at the level of the phenotype. This certainly appears to hold for the traits Harden talks about. Even with GWAS and polygenic scores, we are given no coherent biological mechanism beyond...something to do with the brain, they interact with and are correlated with the environment, and they are contextual and modifiable. Harden laments focus on mechanisms, but identifying specific causal mechanisms would be precisely how education polygenic scores could be actually helpful. For example, in medicine, GWAS have helped identify potential drug targets by identifying biological mechanisms of disease, and can double the likelihood of a drug making it through clinical trials.

However, this situation doesn’t exist for things like education. Instead, we can understand the role of correlated traits like ADHD, or the effect of interventions purely at the phenotypic level by seeing how educational performance and attainment itself change upon interventions from well-designed experiments. In fact, several polygenic scores, from educational attainment to schizophrenia, and even diseases like cardiovascular disease have been shown to have virtually no predictive power beyond common clinical or phenotypic measures, meaning we do not more accurately predict the outcome of those particular phenotypes even with robust polygenic scores. So why not focus our efforts on phenotypes instead of genotypes in cases like education, income, and health where we have some ability to do randomized experiments and a wealth of quasi-natural experiments?

There are existing studies that attempt some kind of true experimental manipulation related to education. Despite what Harden or the charter-school supporting billionaire John Arnold says, we do have some idea on what can improve schools. Research indicates that de-tracking education, that is ending the separation of students by academic ability and having all students engage in challenging curriculum, regularly improves student performance for those with lower ability and does not hinder students with higher ability.

Experiments have shown large benefits to those passing classes and the grades they receive when courses are structured around a more pedagogically informed curriculum that actively engages students. Detracking and active learning have the added advantage of greatly affecting racial gaps in educational performance. To achieve these goals it is likely that teachers will need to be better trained and compensated, and student-pupil ratios would need to change. These changes would likely be related to school funding, teacher salary and quality, and school resources even if those factors are not sufficient to improve educational outcomes in every situation.

Simply identifying that other methods can improve social sciences doesn’t mean we shouldn’t use every tool in our toolbox, as Harden says. However, there are convincing reasons we ought not to rely on genetic data for this kind of research. One reason is that polygenic scores are not very good as controls for experiments testing the effect of environmental intervention. Research has found that the pervasive interplay of genes and environment weakens their ability to control for genetic confounding or identify the efficacy of environmental interventions. Since polygenic scores can reflect contingent social biases without us knowing, it is possible, and likely, that by relying on them to identify effective interventions we are in fact reifying ingrained social and economic biases further in our systems.

One final concern is how this research is interpreted by people, were it to be widely adopted. Researchers found in online experiments that the very act of classifying someone based on their educational polygenic score led to stigmas and self-fulfilling prophecies. Those with high scores were perceived to have more potential and competence while those with low scores were perceived in the opposite way. Not only does this research suggest genetic data leads to essentialist beliefs that can re-entrench existing inequalities, but this kind of dependency can also create even more confounding influences that complicate the application of genetic data for social science questions.

Finally, we reach the last issue with The Genetic Lottery: we don’t need the concept of genetic luck to pursue egalitarian policies. Harden regularly remarks that the alternative is to perceive people’s outcomes as their individual responsibility. Either something is the result of genes they have no control over, or it is their fault for not working hard enough. However, progressive politics revolves around structural and systemic factors that are outside of people’s control and contribute to their outcomes. There is already a recognition of moral luck, or that people’s outcomes are not their fault, but due to the situations they find themselves in. This engagement with progressive motivations and philosophy is absent in Harden’s analysis.

In Harden’s penultimate chapter she contrasts “eugenic,” “genome-blind,” and “anti-eugenic” approaches to policy. What ultimately occurs is a strawman of “genome-blind” policy approaches and often anti-eugenic policies that are hard to distinguish from eugenic policies. For example, what is the difference between Harden’s description of the eugenic policy “Classify people into social roles or positions based on their genetics” and the anti-eugenic policy “Use genetic data to maximize the real capabilities of people to achieve social roles and positions”?  While the genome-blind position is described as “Pretend that all people have an equal likelihood of achieving all social roles or positions after taking into account their environment.“, all we really need to do to achieve our progressive goals is ensure that people’s ability to succeed and thrive in life is not conditioned upon their origin, preferences, or abilities. There’s simply no need to use genetic data on people at all.

In another case involving healthcare Harden suggests the genome-blind approach is to keep our system the same while prohibiting the use of genetic information, while the anti-eugenic approach is creating “systems where everyone is included, regardless of the outcome of the genetic lottery”. However, the system Harden describes is not universal social programs that ensure healthcare, housing, or education regardless of economic situations. Rather it is a system that resembles means-testing social welfare with genetic data. Of course, universal social programs do achieve exactly the anti-eugenic goal while still being genome-blind! Harden’s complete disregard for actual rationale and form of progressive policies when crafting the genome-blind caricatures is inexcusable from someone who claims to be progressive.

For a progressive that supports universal healthcare, a living wage for all, housing as a human right, or free education, it does not matter that people are different and it does not matter the cause for that difference. The fact that some people need healthcare to survive is the reason why it should be available for free, whether the need is from an inherited or acquired disease. It is acknowledged that people have different preferences and strengths, which ultimately results in them living different lives. The fact that for some people this means the difference between a living wage and poverty is what progressives take issue with, and it doesn’t matter what the cause of these differences are, simply that we address them.

Ultimately, Harden tries to sell us on research that we don’t need, based on faulty premises, and that is incapable of delivering on what she promises. Her failure to engage with the history of her own field, her scientific critics, or the actual content of progressive political goals leaves this book in a very poor place. In a way, The Genetic Lottery represents the fact that behavioral genetics no longer has a place to go after the tenets of genetic determinism and biological reductionism were shown to be untenable. If one wants to gain an understanding of modern genetics, or to learn how we may strengthen progressive causes, they should look elsewhere.