Richard White

I had better begin by confessing that I am not an enthusiast for rational choice theory (hereafter RCT). It is not uncommon for devotees of RCT to suggest that it provides a quite general tool for understanding human behaviour, and should largely supersede traditional historical, cultural, and related theoretical approaches. While I am happy to acknowledge that RCT may provide a useful set of tools for certain specific kinds of problems, the attempts to apply it to almost every sphere of human activity seem to me somewhat grotesque. Indeed, I am generally sceptical of attempts to discover a general theoretical approach to human nature, and claims for this sort of application of RCT strike me as some of the most exaggerated. In an ambitious recent book, economist and philosopher of science Don Ross (2005) acknowledges that rational choice does not really have application to humans, but opines that it is perfectly suitable for describing the behaviour of many insects. But rather than concluding, as I have, that rational choice theory is of very limited use for understanding humans, Ross argues instead that the human mind is actually a loose confederation of insectlike components. So once one has found the right structural level at which to apply the RCT machinery, humans turn out to be perfectly appropriate subjects for RCT after all. Unintended reductio ad absurdum is not an infrequent feature of imperialistic academic enterprises.

I have explained many of my reservations about RCT in some detail in my 2001 book, Human Nature and the Limits of Science, and I won’t repeat them here. Many of the central worries will undoubtedly be explored by other contributors to this collection. Probably the biggest problem is the failure of RCT to say anything about the origins of preferences. It is perfectly open to its proponents to deny that this is part of its remit, of course, as economists have consistently been inclined to do: De gustibus non disputandum est. But since understanding why people value the things they do is surely a central question for almost any social enquiry, this refusal makes hyperbolic claims about the significance of RCT obviously unsustainable. Given that there is at the same time increasing evidence that even the most general and abstract statements of RCT, extensively investigated by the recently thriving project of experimental economics,[1] have turned out to be systematically empirically false, it might seem that there was little left to be said for the approach. Even admitting these empirical anomalies, however, it is still often claimed that the theory remains normatively more or less all right, useful or even indispensable for practical decision-making.

Before considering the relevance of RCT to the novel contexts of decision to which genomics gives rise, it will help to say a little about what I understand by genomics. I can best start by mentioning two things that it isn’t. First, genomics is not genetics. Although a good deal of what I shall discuss in this paper fits comfortably under the traditional rubric of genetics, it is important to understand that science is moving rapidly into quite different arenas and I mark these with the less familiar term, genomics. Classically, genetics was the study of heritable differences. For much of its history the causes of these differences were entirely hypothetical entities. The heyday of classical genetics was in the Drosophila labs of T. H. Morgan, his students and collaborators, in which thousands of flies with heritable characteristics were bred and counted, and their traits compared with those of their parents. This work provided a remarkable amount of insight into the hereditary relationship, which was conceived in terms of the transmission of hypothetical factors, or “genes,” from parents to offspring. Many (though not all) the scientists involved in this work assumed that these factors would eventually be identified with material entities, and when, in the second half of the twentieth century, scientists began to understand the chemical nature of the hereditary material it was naturally supposed that these hypothetical causes would be identified and described. But this is not what has happened. Classical genes can be seen as differences—very often defects—in the genome that make a difference to the phenotype, but they are a very poor guide to the real functional constituents of the genome. The relation between genome and developed organism, and the variety of other factors on which this relation depends, has proved far more complex than could have been predicted.

When molecular biologists say, as they often do, that only 3% of the human genome comprises genes, they are not referring to the causes of differences, but to the causes of development. Specifically, they are referring to the sequences of DNA that contain coding sequences that are involved (often in quite complex ways) in the production of proteins. Since many of these proteins must function properly for the organism even to be viable, much of genomics is about things that have to go right in development if development is to happen at all, so it has little or nothing to say about differences. Most alterations to the coding sequences that are referred to in genomics as genes, if they have any effect at all, cause things to go wrong, so that most of human genetics, as opposed to human genomics, is medical genetics. And such entities in medical genetics as the gene for cystic fibrosis, for example, are not determinate bits of DNA sequence, but sets of hundreds of possible errors in the sequence that cause the failure to make a particular protein. Even the familiar gene for blue eyes is similarly one of a large number of errors that prevent the production of brown pigment in the eyes.

Second, by genomics I do not mean primarily the sequencing projects that have provided such an explosion of genomic data in the last ten years or so. Some people do use the term “genomics” this way, and refer to the current era of science as post-genomic. But since no one supposed that sequencing was an end in itself, it seems to me much more appropriate to look at sequencing as a preliminary part of the project of understanding what genomes actually do. And it is the latter that is the area of science in which the exciting developments in genomics are happening.

As a final introductory remark, let me mention one very clear result that is emerging from contemporary genomics. The genome is not, as is still very widely supposed, anything like a blueprint, a code, or a programme. It is not a blueprint because, despite the continued misinterpretation of loose talk about genes for this or that, there is no systematic correlation between parts of the genome and parts of the developed phenotype. It is not a programme because how it functions depends on much else besides merely the sequence of nucleotides. It is at most a constituent of a programme. The genome, in short, is one crucially important part of the cell, but just one. The fundamental unit of biological systems is not the genome, but the cell. Our understanding of the complexity of interactions between the genome and other substances and structures in the cell grows by the week. Perhaps the most important thing about the genome, from a scientific point of view, is not its centrality to the functioning of the cell, but its suitability as a lever for manipulating the cell. It is relatively stable; a crucial property of it, its nucleotide sequence, is measurable in various ways; and we have increasingly powerful techniques for manipulating its activity and even its basic sequence. These technical possibilities give rise to the actual and conceivable choices I want to consider in this paper.

Undoubtedly the most controversial area for genetic choice is in human reproduction. Though the nature of these potential choices is often widely and grossly misrepresented, there is no doubt that they exist. There are two technical contexts in which such choices can be, and to some extent already are being, exercised: selective abortion following amniocentesis (or related forms of fetal diagnosis) and pre-implantation genetic diagnosis (PGD). A somewhat different technology, likely to become much more widely available soon, is sex selection through sperm sorting.

By far the best established of these technologies is selective abortion, reflecting decisions about what kinds of fetuses are and are not desirable. This has mainly been limited to the detection and abortion of fetuses with major chromosomal abnormalities, such as Down syndrome, though use in some cultural contexts for sex selection may be even more significant. The obvious limitation here is of course the drastic nature of the technology. Abortion is not a trivial medical procedure and no one will want to go through a long sequence of conceptions and abortions seeking the genetically perfect baby—even apart from the fact that there is no guarantee that the sequence can be continued successfully, still less that the perfect baby will eventually be conceived. Although the amount of genetic information that could, in principle, be applied to abortion decisions is substantial and growing rapidly, the small number of fetuses to which the decision to abort or not can be applied makes the relevance of this information limited. It seems unlikely that this technology will ever be applicable to anything but the screening of fetuses with very serious genomic defects. And American readers, especially, will not need to be reminded that the technology is an ethically controversial one.

Fantasies about designer babies look a little more realistic in the context of pre-implantation genetic diagnosis. This involves the selection among a set of in vitro fertilised eggs of those thought to have desirable genetic characteristics. Though this is currently an expensive and unreliable technology, and is unlikely to become accessible to the majority of the world’s population, it could quite imaginably become an option for substantial proportions of the populations of affluent societies. To assess this scenario further, I need to say a bit more about the kinds of variation that might imaginably be screened for. There is likely to be little technical limitation to the number of genomic variations that might be screened. Current gene chips can look for thousands of specific sequence segments, and this number could be increased. The problem, which takes me back to my earlier comments about the nature of genes as understood in genomics, is that genes are not simply related to developmental outcomes. There are, as I mentioned, drastic genetic diseases such as cystic fibrosis or Huntington’s disease, which causes rapid neural degeneration in midlife. These are all uncommon, mostly extremely rare. Screening for these whether by PGD or prior to therapeutic abortion does not raise major questions beyond broader general ethical objections to any technique whatever for reproductive screening. The cutting edge of medical genetics, however, is more concerned with predispositions to major common diseases. Here we find a full spectrum of risk-increasing genetic variants, from the BRCA genes that indicate a lifetime risk of up to 80% for breast cancer, to the thrombophilia genes, such as Factor V Leiden,[2] that confer a substantially increased, but still absolutely small, risk of deep vein thrombosis, or a gene that increases by about 30% the risk of smokers’ contracting lung cancer.

It should be said that the point of screening for thousands of these risk-conferring genes remains dubious at present, since the number of embryos available for implantation is small. Still, one can imagine an algorithm for choosing among available fertilised eggs on the basis of the prevalence of such risk-conferring features. More problematically, in the longer term it is entirely possible that feasible techniques will be developed for editing genomes to “correct” large numbers of such variations.

A serious concern, especially with this last prospect, is that these are only in a very technical and misleading sense genes for disease at all. Because of the causal distance from gene sequence to developed organism, genetic differences are likely to have a wide range of effects. Nobody knows, for example, the range of consequences of the so-called thrombophilia genes. It is entirely possible that they may have beneficial effects in addition to the known predisposition to disease. These concerns become even more pressing when one imagines attempt to screen for behavioural genes. There is surely no gene for homosexuality, still less a gene for intelligence but there are very probably genes with measurable statistical effects on one’s IQ or the probability that a child will grow up gay. Apart from obvious ethical concerns about screening for such differences, the overall consequences of attempting to reduce human variability along these dimensions are quite unknown. More important, they are largely unknowable, unless perhaps by the grossly unethical attempt to carry out a large scale experiment in human genetic selection. The central reason for the unpredictable upshot of particular genetic variations is that development involves a cascade of interactions between countless biological and environmental factors. A variant that statistically tends to increase intelligence, say, in the existing set of environments, may have the opposite effect in a different set of environments or, for that matter, in different set of genetic contexts. That different genomes tend to produce quite different phenotypes in varying environments has been well known to plant breeders, for example, for decades. It is ironic that this seems not always to be understood for such enormously complex traits as human intelligence. But it is perhaps reassuring that biological knowledge confirms the widely held intuition that we would do much better to improve IQs (if that’s a good idea) by exploring possible environments than by selecting genomes.

But I have noted the rapid rate of progress in fundamental biology, and it is rash to make predictions about what will remain infeasible in the foreseeable future. So let us imagine for a moment that techniques exist for increasing the probable intelligence of our offspring, or the chances that they will be altruistic, gay, straight, staunchly Republican, seven feet tall, and so on. The high and low quality children envisaged in Becker’s (1991) new family economics takes on a whole new meaning. How would we conceive these possibilities as objects for rational choosing?

Choosing future people poses very sharply some of the major problems with RCT. First of all, there is the question of individualism. RCT sees choice as a utility maximising exercise by an individual pursuing his or her own maximum benefit. Social outcomes are derived from aggregation of individual choices, though it is acknowledged in the extension of RCT to Game Theory that the dispositions to choose of others may sometimes provide essential background to finding optimally self-interested choices. Notoriously, at any rate, it has difficulty dealing with the external effects of individual choices. Given that societies, though they may not be reducible to the individuals that make them up, are in some sense composed of individuals, it is easy to see that technologies of the kind I have been considering, that technologies that determine what kind of individuals will compose a society in the future, pose particular difficulties for the individualistic perspective. Consider the technically simplest choice, the selection of the sex of one’s offspring. I have seen it claimed that the growing sex imbalance in China, with a substantial predominance of young males, has made a major contribution to an exponential rise in prostitution and a concomitant rise in sexually transmitted disease—though no doubt other causes are easy enough to imagine. It is anyhow likely that this will not be the end of the problems caused by large numbers of young males unable to find female partners.

It may reasonably be argued that problems like that just mentioned are no different from a host of similar difficulties with market outcomes. It is extremely desirable for most of us individually to own and use a car, but when we all do so we can barely drive down the congested roads and the planet is in danger of being rendered uninhabitable. But even apart from the point just mentioned, that reproductive choices affect the constitution of a society, not merely its conditions, there is a further important difference in relation to what one might think of as the natural equilibrium that exists prior to the exercise of these individual choices. The “natural” equilibrium for transport is non-existence, or anyhow pedestrianism. The market in automobiles provides something—mobility—that most people see as desirable, even if it ends up providing a markedly suboptimal solution to the problem. The equilibrium in human sex is very close to 1:1, and this is very plausibly socially optimal. Individual choices in offspring sex selection can at best do no social harm, and experience suggests that they won’t do this well. Hence from a social perspective there is no benefit and considerable potential cost in introducing a market-like system of individual choices.

It may seem that genetic “enhancement” of future children, unlike the case of sex ratio, offers by definition an improvement on the natural equilibrium. But consideration of the expense of the technologies envisaged suggests that an even more seriously undesirable social outcome is much more likely. Genetic screening of eggs prior to implantation, let alone editing of genetic “mistakes” are almost certain to remain extremely expensive technologies within the reach of only an economic elite. This raises the familiar dsytopic vision of a bifurcation of the human species into streams of genetically enhanced and genetically unenhanced. Most people will find this prospect horrifying. [3]

Equally clearly, the systematic tendency of some societies to produce, given this aspect of reproductive choice, large excesses of male children raises the question of the origin of these choices. I am not an expert on the relevant anthropology, but I take it that no one supposes that a taste for male children is just a randomly endogenously generated aspect of individuals. Evolutionists might imagine that if there is an endogenously generated preference it will be for the sex that is currently least frequent, as this will give the best reproductive prospects. But since this was not something people were in a position to choose until a very late stage in history (at least prior to a great deal of unrecoverable investment in cases where infanticide is the only technique to hand), no one imagines that such a conditional preference has actually evolved as an innate factor affecting conscious choices. More realistically, the combination of a particular gendered socio-economic system with new technological possibility is producing a social outcome that nobody would rationally choose. Individual choices, from this perspective, are merely intermediate links in this unwanted causal chain.

Perhaps the most problematic ethical issues arise from the fact that rational choice is generally seen as a choice between available and exchangeable goods. But children are not normally conceived of as the property of their parents. It is of course precisely for this reason that the talk of investing in children of various qualities in Becker’s family economics seems distasteful to many people. New possibilities for reproductive choice add a new dimension to what might have been seen as no more than an economist’s unpleasant fantasy. It is not simple to diagnose the moral discomfort engendered by choosing the characteristics of one’s children and I won’t claim to be able to resolve the philosophical difficulties that arise from such possibilities. I certainly don’t want to claim concern for the possible people who fail to exist because a particular possible person has been chosen for actuality. And certainly by providing various inputs into their children’s development, most notably educational inputs, parents quite legitimately contribute to determining properties of their offspring. Still, choosing, for example, the sex or sexuality of one’s child differs markedly from choosing the colour or engine capacity of one’s car and even from choosing whether to provide the loved one with piano lessons. One such respect which offers a certain irony in the present intellectual context is that one is likely to be choosing the determinants of the preferences of these people to be. Even if it is appropriate to treat the preferences of the present generation of parents as “exogenous to the model,” it will be difficult to do so for the next generation if their preferences were determined by choices in the first iteration of the model. Preferences for other peoples’ preferences are no doubt very common—homophobes, for instance, would presumably prefer that others did not prefer same sex partners—but such preferences are a perennial embarrassment to theories of liberal democracy. Systematic implementation of such preferences in the production of future generations, predetermined limitations of the freedom of future individuals to develop their own preferences, seem highly undesirable, and especially so from a perspective that attaches pre-eminent importance to individual choice. Piano lessons, it is true, may have some slight effect on the child’s future tastes, but it is important that one does have the choice of consulting the child’s antecedent desires in deciding whether to provide the lessons at all. There are, indeed, related issues in the philosophy of education, but there are also important differences between the cases.

Of more practical importance, given our understanding of contemporary biology, by trying to determine one feature of a planned child, one is surely making many unintended choices about the characteristics of one’s children. The problem here cuts to the heart of the biological misunderstandings so prevalent in this area. If one is impressed by constant media announcements of the gene for homosexuality, alcoholism, musical genius, or whatever, it may be natural to think that these are things one would like one’s child to have or not have. But these genes, if they exist at all, are not necessary and sufficient causes for the traits mentioned, but merely factors which, averaged over the whole range of possible genetic and non-genetic contexts, increase the probabilities of such outcomes. And they may very well increase or decrease the probabilities of many other outcomes at the same time. The notion of an organism as a bundle of discrete traits each with its distinct cause, an image all too readily promoted by much loose gene talk, encourages an entirely mistaken analogy between the assembly of a child and the loading of a supermarket trolley.[4] An important issue then, to which I shall briefly return, is biological ignorance and, more importantly, ignorance of our ignorance.

One further very important point about genomics as an object of choice is the extent to which it is inherently relational. Genomics inherits from genetics a fundamental concern with inheritance, which is, in this context, a relation of similarity. The consequences of this fact are perhaps most striking in the context of a rather different area in which genomics presents new kinds of choices, choices about knowledge.

One rather remarkable thing to which genetic testing has given rise is the right not to know. A paradigm case to which this applies is that of Huntington’s disease, a dominant genetic condition that leads with great predictability to devastating neurological deterioration in middle age. Children of parents with this genetic disease will know that they have a 50% chance of succumbing to the same horrifying fate in middle age. Tests for the condition are relatively routine, and the development of the disease, apart from the exact time of onset, is virtually inevitable. Many people prefer not to know whether this fate is in store for them, and in some places they have been accorded an explicit right not to be told this. But this decision to know or not to know is something that cannot, in principle, be a fully isolated decision or, therefore, the exercise of a fully individualised right. If I learn that I have the gene for Huntington’s, my children, if I tell them, learn that they have a 50% chance of sharing my fate. Knowledge about myself is at least probabilistic knowledge about my relatives.

More stark still are the issues of interconnection that arise through the possibility of paternity testing. I may, for whatever reason, feel a pressing need to confirm that my children are indeed genetically my own. (Given that genetic studies have shown that from around 5% to as high as 20% of children don’t have the father they think they have, one can imagine this becoming an increasingly common concern.) The technology to address this question is readily available, and can now be purchased over the internet. (A quick check on Google at the time of writing reveals that at genetree.com you can currently find a special offer on a standard paternity test for $199.) But if I learn that my son is not my genetic child, he is likely to discover that I am not his genetic father, something he may very well have preferred not to find out. It is also, of course, something his mother may not have wanted to know. In Germany there is currently a heated debate about making it illegal for a putative father to solicit such a test without the explicit consent of the child’s mother, a piece of legislation apparently known fondly by activist fathers’ groups as (roughly translated) “the slut protection law.” Since this increasingly routine technology is explicitly designed to provide information about at least two other people it cannot be treated straightforwardly as the expression of preference of the autonomous buyer.

The dilemmas of the interconnectedness of knowledge about paternity are strikingly illustrated by the very possible dilemma of the doctor, asked to advise parents with a genetically diseased child whether they should take the risk of another pregnancy. The doctor knows that there is no such risk, as the genetic disease originates with a person not party to the discussion—the actual but absent genetic father. Should she, for example, perform a sterilisation procedure on the basis of the mistaken assumptions of the couple? Or risk destroying the existing family unit by revealing the unsuspected parentage of the first child? I offer no answer to these questions, only stressing again the impossibility of localising the value of this information to one individual chooser.

Genomics and genetics have had impacts in areas quite distinct from health and reproduction. In fact in the UK the most contentious applications of genetic technologies have been to food.[5] Genetics and genomics have the potential to transform radically the kinds of organisms we cultivate to eat. By far the best known of such technologies is transgenic genetic modification, insertion of DNA sequence from one species into another. This is a rather crude technique, and very likely will not be the most important for transforming food supplies, but it will do well for the present discussion. At first sight the reception of GM foods might seem a paradigm for the successful operation of RCT. Producers offer a new product and consumers choose whether they prefer to purchase it. In the U.S. they have generally been happy to do so, in Europe they have largely rejected the opportunity. Everyone gets what he or she wants.

The most pressing problem here, and surely it is a very general problem with complex new technologies, is that the basis for these decisions seems typically to be anything but rational. In Britain, for instance, there have been widely publicised reports of large scale studies of the environmental impact of herbicide resistant crops. These experiments have in fact addressed the impact on wildlife diversity of particular regimes of herbicide use associated with the use of herbicide resistant crops, and the results have been varied. They have been widely interpreted, however, as measuring some direct environmental damage done by the crops themselves, for example, as somehow poisoning wild creatures. There has been extensive public consultation about how people feel about these crops, but these unfortunately have tended to elicit responses such as, “I’m certainly not going to eat anything with DNA in it.”

These are complex technical issues and consulting the feelings of the public (i.e. their raw preferences) does not seem a very useful way of deciding the appropriate policy. Of course I don’t mean to say that these decisions should just be left to experts and the public should do what they are told. The public is understandably and probably rightly suspicious of the ability of governments to make good decisions about such matters, and plausibly tends to suspect that governments are strongly influenced by the huge corporations that produce these new biotechnologies. My point is just that RCT does nothing to address these real problems. The central problems are, I think, providing high quality information for the general public, and providing it from people whom the general public find credible. The valorisation of individual decision-making does nothing to address either of these problems.

Let me summarise my general point. Biotechnology provides extraordinary possibilities for changing human lives, and hence faces us with important new choices about how we and future generations will live. The diversity of the impact of biotechnology is reflected in the diversity of problems involved in making these new choices wisely. These problems include, for example: determining the ethical status that should be accorded to unfamiliar biological objects, particularly in reproductive technology; balancing the dangers of giving too much authority on the implementation of new technologies to experts with the dangers of taking too much account of public feelings that may be based on profound misunderstanding, and the correlative problem of providing publics and interested parties with accessible and good quality information; weighing the desirability of individual autonomy, again particularly in reproductive decisions, against potentially undesirable aggregate social decisions; exploring the implications of vital personal knowledge that is, by its nature, also knowledge about others. These problems range across the philosophical domains of ethics and metaphysics, and require inputs from several areas of sociology. No doubt there is a part to be played by Rational Choice Theory, Game Theory, Decision Theory or, generally, the cluster of academic disciplines that deal with the problems of optimal individual choice under scarcity, uncertainty, in strategic interactions and so on. But as far as I can see this is a relatively minor part.

I take this to emphasise a more general point. There are interesting problems that can be addressed under such rubrics as, for example, choice under scarcity. But suggestions that this, or anything else, is the Rosetta Stone that will be central to understanding the whole range of problems that arise with human behaviour or human choice is silly and irresponsible. Academic disciplines, I suggest, provide deep but partial insights. Major human problems, as, for example, assessing the risks and benefits of a potentially life-transforming science and technology, require various perspectives with complementary and overlapping insights. Rational Choice Theory is fine in its place. But when its overenthusiastic supporters attempt to impose it on an ever-expanding range of problems far from its original home, it becomes an intellectual cancer. This essay is intended as brief excursion into intellectual oncology.[6]

John Dupré is a philosopher of science whose work has focused especially on issues in biology. He is currently Professor of Philosophy of Science at the University of Exeter, and since 2002 he has been Director of the ESRC Centre for Genomics in Society (Egenis). He has formerly held posts at Oxford, Stanford, and Birkbeck College, London. In 2006 he held the Spinoza Visiting Professorship at the University of Amsterdam. His publications include The Disorder of Things: Metaphysical Foundations of the Disunity of Science (Harvard, 1993); Human Nature and the Limits of Science (Oxford, 2001); Humans and Other Animals (Oxford, 2002); and Darwin’s Legacy: What Evolution Means Today (Oxford, 2003). Most recently he co-authored Genomes and What to Make of Them (Chicago, 2008) with the sociologist Barry Barnes.

References

Becker, Gary S. A Treatise on the Family. Enl. ed. Cambridge, MA: Harvard University Press, 1991.

Dupré, John. Human Nature and the Limits of Science. Oxford: Oxford University Press, 2001.

Guala, Francesco. The Methodology of Experimental Economics. Cambridge: Cambridge University Press, 2005.

Moss, Lenny. What Genes Can’t Do. Cambridge, MA: MIT Press/Bradford Books, 2003.

Ross, Don. Economic Theory and Cognitive Science: Microexplanation. Cambridge, MA: MIT Press/Bradford Books, 2005.

Saukko P., S. Ellard, S. Richards, M. Shepherd, and J. Campbell. (2007). Patients' understanding of genetic susceptibility testing in mainstream medicine: A qualitative study on thrombophilia. BMC Health Services Research 7:82.

Silver, Lee M. Remaking Eden. New York: Avon Books, 1997.