Read Part 1 If you were a woman applying to graduate school at the University of California at Berkeley in 1973, you would have had a 35 percent chance of being accepted. If you were a man, your chances would have been significantly better. Forty-four percent of male applicants got accepted that year. Apparently, at this early stage of the feminist movement, even a school as notoriously progressive as Berkeley still discriminated against women. But not surprisingly, when confronted with these numbers, the women of the school were ready to take action to right the supposed injustice. After a lawsuit was filed charging admissions offices with bias, however, a department-by-department examination was conducted which produced a curious finding: not a single department admitted a significantly higher percentage of men than women. In fact, there was a small but significant trend in the opposite direction—a bias against men.
What this means is that somehow the aggregate probability of being accepted into grad school was dramatically different from the probabilities worked out through disaggregating the numbers with regard to important groupings, in this case the academic departments housing the programs assessing the applications. This discrepancy called for an explanation, and statisticians had had one on hand since 1951.
This paradoxical finding fell into place when it was noticed that women tended to apply to departments with low acceptance rates. To see how this can happen, imagine that 90 women and 10 men apply to a department with a 30 percent acceptance rate. This department does not discriminate and therefore accepts 27 women and 3 men. Another department, with a 60 percent acceptance rate, receives applications from 10 women and 90 men. This department doesn’t discriminate either and therefore accepts 6 women and 54 men. Considering both departments together, 100 men and 100 women applied, but only 33 women were accepted, compared with 57 men. A bias exists in the two departments combined, despite the fact that it does not exist in any single department, because the departments contribute unequally to the total number of applicants who are accepted. (25)
This is how the counterintuitive statistical phenomenon known as Simpson’s Paradox is explained by philosopher Elliott Sober and biologist David Sloan Wilson in their 1998 book Unto Others: The Evolution and Psychology of Unselfish Behavior, in which they argue that the same principle can apply to the relative proliferation of organisms in groups with varying percentages of altruists and selfish actors. In this case, the benefit to the group of having more altruists is analogous to the higher acceptance rates for grad school departments which tend to receive a disproportionate number of applications from men. And the counterintuitive outcome is that, in an aggregated population of groups, altruists have an advantage over selfish actors—even though within each of those groups selfish actors outcompete altruists.
Sober and Wilson caution that this assessment is based on certain critical assumptions about the population in question. “This model,” they write, “requires groups to be isolated as far as the benefits of altruism are concerned but nevertheless to compete in the formation of new groups” (29). It also requires that altruists and nonaltruists somehow “become concentrated in different groups” (26) so the benefits of altruism can accrue to one while the costs of selfishness accrue to the other. One type of group that follows this pattern is a family, whose members resemble each other in terms of their traits—including a propensity for altruism—because they share many of the same genes. In humans, families tend to be based on pair bonds established for the purpose of siring and raising children, forming a unit that remains stable long enough for the benefits of altruism to be of immense importance. As the children reach adulthood, though, they disperse to form their own family groups. Therefore, assuming families live in a population with other families, group selection ought to lead to the evolution of altruism.
|(pg 24) Darker area represents altruists and shrinks in|
both groups--but notice the right circle gets bigger.
As we have seen, however, kin selection is a special case of a more general theory—a point that Hamilton was among the first to appreciate. In his own words, “it obviously makes no difference if altruists settle with altruists because they are related… or because they recognize fellow altruists as such, or settle together because of some pleiotropic effect of the gene on habitat preference.” We therefore need to evaluate human social behavior in terms of the general theory of multilevel selection, not the special case of kin selection. When we do this, we may discover that humans, bees, and corals are all group-selected, but for different reasons. (134)
A general proclivity toward altruism based on section at the level of family groups may look somewhat different from kin-selected altruism targeted solely at those who are recognized as close relatives. For obvious reasons, the possibility of group selection becomes even more important when it comes to explaining the evolution of altruism among unrelated individuals.
We have to bear in mind that Dawkins’s selfish genes are only selfish with regard to concerning themselves with nothing but ensuring their own continued existence—by calling them selfish he never meant to imply they must always be associated with selfishness as a trait of the bodies they provide the blueprints for. Selfish genes, in other words, can sometimes code for altruistic behavior, as in the case of kin selection. So the question of what level selection operates on is much more complicated than it would be if the gene-focused approach predicted selfishness while the multilevel approach predicted altruism. But many strict gene selection advocates argue that because selfish gene theory can account for altruism in myriad ways there’s simply no need to resort to group selection. Evolution is, after all, changes over time in gene frequencies. So why should we look to higher levels?
|David Sloan Wilson|
Sober and Wilson demonstrate that if you focus on individuals in their simple model of predominantly altruistic groups competing against predominantly selfish groups you will conclude that altruism is adaptive because it happens to be the trait that ends up proliferating. You may add the qualifier that it’s adaptive in the specified context, but the upshot is that from the perspective of individual selection altruism outcompetes selfishness. The problem is that this is the same reasoning underlying the misguided accusations against Berkley; for any individual in that aggregate population, it was advantageous to be a male—but there was never any individual selection pressure against females. Sober and Wilson write,
The averaging approach makes “individual selection” a synonym for “natural selection.” The existence of more than one group and fitness differences between the groups have been folded into the definition of individual selection, defining group selection out of existence. Group selection is no longer a process that can occur in theory, so its existence in nature is settled a priori. Group selection simply has no place in this semantic framework. (32)
Thus, a strict focus on individuals, though it may appear to fully account for the outcome, necessarily obscures a crucial process that went into producing it. The same logic might be applicable to any analysis based on gene-level accounting. Sober and Wilson write that
if the point is to understand the processes at work, the resultant is not enough. Simpson’s paradox shows how confusing it can be to focus only on net outcomes without keeping track of the component causal factors. This confusion is carried into evolutionary biology when the separate effects of selection within and between groups are expressed in terms of a single quantity. (33)
They go on to label this approach “the averaging fallacy.” Acknowledging that nobody explicitly insists that group selection is somehow impossible by definition, they still find countless instances in which it is defined out of existence in practice. They write,
Even though the averaging fallacy is not endorsed in its general form, it frequently occurs in specific cases. In fact, we will make the bold claim that the controversy over group selection and altruism in biology can be largely resolved simply by avoiding the averaging fallacy. (34)
Unfortunately, this warning about the averaging fallacy continues to go unheeded by advocates of strict gene selection theories. Even intellectual heavyweights of the caliber of Steven Pinker fall into the trap. In a severely disappointing essay published just last month at Edge.org called “The False Allure of Group Selection,” Pinker writes
If a person has innate traits that encourage him to contribute to the group’s welfare and as a result contribute to his own welfare, group selection is unnecessary; individual selection in the context of group living is adequate. Individual human traits evolved in an environment that includes other humans, just as they evolved in environments that include day-night cycles, predators, pathogens, and fruiting trees.
Multilevel selectionists wouldn’t disagree with this point; they would readily explain traits that benefit everyone in the group at no cost to the individuals possessing them as arising through individual selection. But Pinker here shows his readiness to fold the process of group competition into some generic “context.” The important element of the debate, of course, centers on traits that benefit the group at the expense of the individual. Pinker writes,
Except in the theoretically possible but empirically unlikely circumstance in which groups bud off new groups faster than their members have babies, any genetic tendency to risk life and limb that results in a net decrease in individual inclusive fitness will be relentlessly selected against. A new mutation with this effect would not come to predominate in the population, and even if it did, it would be driven out by any immigrant or mutant that favored itself at the expense of the group.
But, as Sober and Wilson demonstrate, those self-sacrificial traits wouldn’t necessarily be selected against in the population. In fact, self-sacrifice would be selected for if that population is an aggregation of competing groups. Pinker fails to even consider this possibility because he’s determined to stick with the definition of natural selection as occurring at the level of genes.
Indeed, the centerpiece of Pinker’s argument against group selection in this essay is his definition of natural selection. Channeling Dawkins, he writes that evolution is best understood as competition between “replicators” to continue replicating. The implication is that groups, and even individuals, can’t be the units of selection because they don’t replicate themselves. He writes,
The theory of natural selection applies most readily to genes because they have the right stuff to drive selection, namely making high-fidelity copies of themselves. Granted, it's often convenient to speak about selection at the level of individuals, because it’s the fate of individuals (and their kin) in the world of cause and effect which determines the fate of their genes. Nonetheless, it’s the genes themselves that are replicated over generations and are thus the targets of selection and the ultimate beneficiaries of adaptations.
The underlying assumption is that, because genes rely on individuals as “vehicles” to replicate themselves, individuals can sometimes be used as shorthand for genes when discussing natural selection. Since gene competition within an individual would be to the detriment of all the genes that individual carries and strives to pass on, the genes collaborate to suppress conflicts amongst themselves. The further assumption underlying Pinker’s and Dawkins’s reasoning is that groups make for poor vehicles because suppressing within group conflict would be too difficult. But, as Sober and Wilson write,
This argument does not evaluate group selection on a trait-by-trait basis. In addition, it begs the question of how individuals became such good vehicles of selection in the first place. The mechanisms that currently limit within-individual selection are not a happy coincidence but are themselves adaptions that evolved by natural selection. Genomes that managed to limit internal conflict presumably were more fit than other genomes, so these mechanisms evolve by between-genome selection. Being a good vehicle as Dawkins defines it is not a requirement for individual selection—it’s a product of individual selection. Similarly, groups do not have to be elaborately organized “superorganisms” to qualify as a unit of selection with respect to particular traits. (97)
The idea of a “trait-group” is exemplified by the simple altruistic group versus selfish group model they used to demonstrate the potential confusion arising from Simpson’s paradox. As long as individuals with the altruism trait interact with enough regularity for the benefits to be felt, they can be defined as a group with regard to that trait.
Pinker makes several other dubious points in his essay, most of them based on the reasoning that group selection isn’t “necessary” to explain this or that trait, only justifying his prejudice in favor of gene selection with reference to the selfish gene definition of evolution. Of course, it may be possible to imagine gene-level explanations to behaviors humans engage in predictably, like punishing cheaters in economic interactions even when doing so means the punisher incurs some cost to him or herself. But Pinker is so caught up with replicators he overlooks the potential of this type of punishment to transform groups into functional vehicles. As Sober and Wilson demonstrate, group competition can lead to the evolution of altruism on its own. But once altruism reaches a certain threshold group selection can become even more powerful because the altruistic group members will, by definition, be better at behaving as a group. And one of the mechanisms we might expect to evolve through an ongoing process of group selection would operate to curtail within group conflict and exploitation. The costly punishment Pinker dismisses as possibly explicable through gene selection is much more likely to havearisen through group selection. Sober and Wilson delight in the irony that, “The entire language of social interactions among individuals in groups has been burrowed to describe genetic interactions within individuals; ‘outlaw’ genes, ‘sheriff’ genes, ‘parliaments’ of genes, and so on” (147).
Unto Others makes such a powerful case against strict gene-level explanations and for the potentially crucial role of group selection that anyone who undertakes to argue that the appeal of multilevel selection theory is somehow false without even mentioning it risks serious embarrassment. Published fourteen years ago, it still contains a remarkably effective rebuttal to Pinker’s essay:
In short, the concept of genes as replicators, widely regarded as a decisive argument against group selection, is in fact totally irrelevant to the subject. Selfish gene theory does not invoke any processes that are different from the ones described in multilevel selection theory, but merely looks at the same processes in a different way. Those benighted group selectionists might be right in every detail; group selection could have evolved altruists that sacrifice themselves for the benefit of others, animals that regulate their numbers to avoid overexploiting their resources, and so on. Selfish gene theory calls the genes responsible for these behaviors “selfish” for the simple reason that they evolved and therefore replicated more successfully than other genes. Multilevel selection theory, on the other hand, is devoted to showing how these behaviors evolve. Fitness differences must exist somewhere in the biological hierarchy—between individuals within groups, between groups in the global population, and so on. Selfish gene theory can’t even begin to explore these questions on the basis of the replicator concept alone. The vehicle concept is its way of groping toward the very issues that multilevel selection theory was developed to explain. (88)
Sober and Wilson, in opening the field of evolutionary studies to forces beyond gene competition, went a long way toward vindicating Stephen Jay Gould, who throughout his career held that selfish gene theory was too reductionist—he even incorporated their arguments into his final book. But Sober and Wilson are still working primarily in the abstract realm of evolutionary modeling, although in the second half of Unto Others they cite multiple psychological and anthropological sources. A theorist even more after Gould’s own heart, one who synthesizes both models and evidence from multiple fields, from paleontology to primatology to ethnography, into a hypothetical account of the natural history of human evolution, from the ancestor we share with the great apes to modern nomadic foragers and beyond, is the anthropologist Christopher Boehm, whose work we’ll be exploring in part 3.Read Part 1 of A Crash Course in Multilevel Selection Theory: The Goundwork Laid by Dawkins and Gould
And Part 3: The People Who Evolved Our Genes for Us: Christopher Boehm on Moral Origins.