Haidt’s group selection


Jason Collins


September 28, 2012

Having given my thoughts on Haidt’s generally excellent The Righteous Mind in my last post, I want to turn to Haidt’s use of group selection in the last third of his book. The central themes of his book don’t rest on group selection (in my opinion), but Haidt is at the centre of the reemergence of group selection in the social sciences and his points are worth discussing.

Haidt uses the more modern phrase “multilevel selection” in addition to “group selection” through the book. Multilevel selection refers to a method of accounting for selection at the different levels (e.g. the gene, individual, group etc.), while the older concept of group selection usually refers to natural selection between groups and the evolution of group-level adaptations. Multilevel selection also involves what are called “trait groups”, which may briefly form and break up, compared to the rigid reproducing groups of the old group selection.

It seems that Haidt has a reasonable grasp on these distinctions but his use of the term multilevel selection is often confusing. For example, he keeps using the phrase “product of multilevel selection”. But multilevel selection is, as the name suggests, selection at different levels. You can look at a scenario through a multilevel selection framework and conclude that all the selection occurs at the level of the gene or individual. Multilevel selection and inclusive fitness are just different accounting methods (or languages). It is not a case that one happens and the other doesn’t. What Haidt is implying, and the way he should state it, is that selection is occurring predominantly at the level of the group. Based on some passages of the book, it is clear that Haidt understands this, but at other times his language is loose.

When Haidt argues for the importance of selection at the level of the group (I’ll refer to it as group selection for rest of this post), he offers four lines of evidence: the role of group selection in the major evolutionary transitions; the shared intentionality of humans; gene-culture evolution; and the potential for fast evolution.

The major evolutionary transitions, such as the emergence of eukaryotic cells from the combination of bacteria, are one of the few areas where many evolutionary biologists will agree that group selection occurred. Haidt characterises the major evolutionary transitions as times where methods to control freeriding evolved at one level, allowing superorganisms to arise at the next. Haidt then follows in the footsteps of biologists such as David Sloan Wilson and E. O. Wilson and argues that the evolution of “ultrasociality” in humans is a similar transition.

I don’t want to rehash this argument in-depth, but it goes back to the classic group selection debate. In the evolutionary transitions of the past, a reproductive bottleneck was present. Once two bacteria are combined in a cell, the only way they can reproduce is if the “group” reproduces. But that bottleneck does not exist in human groups, so there is opportunity for freeriding. We then get to the old debate about the level of freeriding and whether the level of group extinction and the degree of gene flow between groups allows group selection to outweigh this freeriding.

While Haidt follows in others’ footsteps in referring to the major evolutionary transitions, his other arguments are more his own. On shared intentionality, Haidt argues that the ability to share intentions between people allows collaboration, the division of labour and shared norms. While Haidt claims this is group selection, this is a case where the multilevel selection framework should be properly applied. How much benefit does one get as an individual from understanding what someone else is thinking, versus the benefit you get from pairing with someone who also has that ability and working together to succeed as a group? While having more people in your group who are able to share intentions will help you defeat other groups, shared intentionality is clearly beneficial to an individual. Being in a group of mind readers when you have no idea what is going on is suboptimal. Which level the selection predominantly operates at needs to be analysed (and will depend upon assumptions about what makes up a group). This is the type of scenario that I have argued before is simpler to analyse in an inclusive fitness framework.

Haidt’s third line of evidence is a somewhat confusing take on gene-culture evolution. Haidt argues that cultural group selection supported “prototribalism”, which led to an environment that then supported genetic evolution. However, Haidt’s examples do not sound like group selection. For example, Haidt writes:

[I]ndividuals who found it harder to play along, to restrain their antisocial impulses, and to conform to the most important collective norms would not have been anyone’s top choice when it came time to choose partners for hunting, foraging, or mating. In particular, people who were violent would have been shunned, punished, or in extreme cases killed.

This sounds like individual level selection against violent, non-conformist individuals. I am not sure why Haidt was so keen to covert Boyd and Richerson’s arguments on cultural group selection into genetic group selection, but try he did.

Haidt’s biggest reach, however, comes with his argument that the potential for fast evolution supports group selection. Haidt notes that gene-culture evolution reached fever pitch in the last 12,000 years, and that is an assessment I would agree with. He refers to the group selection experiments conducted by William Muir, in which Muir rapidly improved egg laying by selecting groups of successful chickens (achieved, of course, through the effective creation of a reproductive bottleneck in the experimental design). Haidt then pushes the rapid evolution argument to the limits when he seeks to implicate group selection in the emergence of religion. As large-scale religion only emerged since the dawn of agriculture, Haidt suggests the rapid recent evolution identified by the likes of John Hawks, Greg Cochrane and Henry Harpending provides scope for recent group selection. He writes:

[G]roup selection can work very quickly (as in the case of those group-selected hens that became more peaceful in just a few generations). Ten thousand years is plenty of time for gene-culture coevolution, including some genetic changes, to have occurred. And 50,000 years is more than plenty of time for genes, brains, groups, and religions to have coevolved into a very tight embrace.

The problem is that group extinction and reproduction generally occurs more slowly than individual level selection. At the individual level, we see large differences in fertility every generation. For many people, it is the end of the genetic line. To the extent heritable traits underlie this variation, we can see rapid changes in genotype. In contrast, studies of rates of group extinction suggest it is slow. Further, groups tend not to be simply wiped out, but the “loser” groups tend to merge into the victor, bringing their genes with them.

Having said all this, we might be able to build a multilevel selection model in which we allow temporary religious or other “trait groups” to form and break up in short periods and divide the degree of selection between the various levels. However, I still doubt we will see significant selection at the group level for most of these examples and I don’t feel that this was the sort of group selection Haidt was interested in. Further, I expect the inclusive fitness framework would give a clearer picture. If this trait group approach could have provided a stronger argument, Haidt might have used it.