The essence of complex life is cooperation. Genes link together to form chromosomes, cells clump together to form organisms, organisms group together to form societies. You and I are possible is because individual genes, chromosomes, and cells made a pact to work together instead of going it alone. We are also a social species. Our ancestors hung together in groups rather than pushing forth in competitive isolation.

Cooperative agreements among independent entities are so entwined with the nature of life that discovering the terms of such associations- from amoebae to ant colonies and beyond- is central to understanding how life exists in the first place. These are weighty issues, and I do not think it exaggeration to place the study of cooperation at the heart of any Grand Unifying Theory of biology.

I mention the importance of cooperation in biology because you might think that scientists who study cooperation ought show signs of being good at it themselves. But you’d be mistaken. The recent Nature paper by Nowak, Tarnita and Wilson is the latest salvo in a bitter and, in my opinion, largely pointless divide.

Biologists who study the evolution of cooperation fall into two tribes. Those who think cooperation arises primarily because it facilitates the reproduction of kin (the kin selectionists), and those who think cooperation arises because groups that cooperate do better than groups that don’t (the group selectionists). Not what you’d consider the stuff of lurid vendettas, unless of course you are already familiar with the sort of minutiae that causes academics to work themselves into a huff.

The kin selection tribe is older and better established. They’ve been productive for decades and tend to carry on with their lines of research without paying group selectionists much heed, except to issue condescending comments now and again when provoked.

The group selectionists find being ignored absolutely infuriating. They are pretty sure they’ve revolutionized Evolutionary Science, but no one notices and they have grown hoarse shouting into the void. So they write their own screeds that more or less pointedly tell the kin selectionists where they can shove their theory. Nowak et al being the latest installment.

I am giving the debate a tribal frame for a reason. When I sit down with the actual papers and models, the level of actual, quantitative disagreement falls far short of the bitter rhetoric that surrounds it. In my humble opinion, the partisans have become more interested in discrediting the other side than in advancing mutual understanding. Small statements are taken out of context and destroyed in straw-man arguments, studies are cherry-picked for rhetorical effect, quotes are mined, and the result is a downwardly tribalistic spiral as frustration grows and everyone starts to hate everyone else. A fine pickle for cooperation research if you ask me.

The reality is both groups are interested in slightly different types of questions, yet often fail to recognize this. Instead, they see the other as attempting to answer the same questions, but doing it wrong, and that ends up in confusion for everyone.

Now that I’m done placing a pox on both houses, here are a few thoughts.

Let me pick on Nowak et al for four things. First, their triumphal tone and thinly-veiled insults against colleagues throughout the paper are unprofessional, unjustified, and likely to damage their case in the long run. Framing their paper in such an adversarial tone was a mistake, as at this point they are unlikely to convince their opponents of anything. Which is a real shame, because on some counts they are absolutely correct.

Second, they pigeonholed kin selection too narrowly, making some questionable assumptions about weak selection and about what the social benefits & costs in Hamilton’s model (rB>C) really mean. This has downstream consequences not for the efficacy of their modeling of groups but for their arguments about the failure of kin selection.

Third, their use of the term “standard natural selection” is unorthodox, to say the least. What they mean is, “group dynamics modeled without a parameter tracking relatedness”.

Fourth- and this is what really kills me- they fail to understand the distinction between proximate hypotheses that explain mechanistic processes, and ultimate hypotheses that explain the more abstract population genetic conditions under which social alleles may spread. The two classes of explanations are not mutually exclusive. Kin selection is not mechanistic. It is neutral about what the social alleles actually do, merely describing whether they spread. As such, kin selection is compatible with an enormous range of mechanistic explanations about gene regulation, ecological conditions, etc. When Nowak et al propose an “alternate theory” involving nesting behavior and group formation, it merely shows that they don’t understand kin selection as an ultimate hypothesis, and hence are not in a position to critique it.

I can sympathize with this last point for historical reasons, though. Kin selection must have proven endlessly frustrating to scientists interested in detailed mechanisms of the transition of solitary to social. For example, questions about gene regulation, and about ecological context. This is because kin selection was accepted so quickly early on that it rose to a sort of undeserved hegemony, and others had to face an uphill battle against a perception that kin selection theory had already “solved” the eusociality problem. It hadn’t. It had merely provided some underlying population genetic conditions under which a new social mutation could spread. Kin selection sucked some of the air out of other potentially interesting avenues of research, creating an ambiance where resentment could fester.

Now, on to the kin selectionists. In my opinion they are logically correct but mired in reductio ad absurdum. If I want to build a model of the evolution of human populations, I don’t generally need to include terms in the model of how every individual cell in the bodies of each person relates to the other cells. One can certainly build an accurate model- even down to tracking individual atoms- but why go through all that if it is only tangential to the dynamic in question? We abstract away all the reduction as being unnecessarily redundant. Why, then, must kin selectionists insist that every model of group evolution include terms tracking the internal elements of those groups?

They hold it is because relatedness within groups drives the interactions among groups. Which it can do, some of the time. But kin selection is sometimes argued as a first principle rather than an empirically verifiable hypothesis. There are instances where such models are necessary, but there are others where they are superfluous, and it’s myopic on their part to assume that other researchers are interested in the same sorts of questions that they themselves are. Nowak et al’s model demonstrates this point well. They did, in fact, evolve theoretical eusociality without accounting for relatedness. Yet because of all the other baggage in the paper I fear this most important aspect will be ignored.

A difficult aspect for we empiricists to grasp is the extent to which the debate is about the abstractions of the best framework for building models. At least 2/3 of Nowak et al is an exposition of how their game theory math is better suited to grasping the dynamics of groups than the population genetics math of the kin selectionists, rather than an argument about what is really happening to real animals out in the real world. Think of it like programmers arguing whether Python or Perl is better language for a particular problem. It would be helpful for partisans to be able to separate arguments about the fact of what happens in the world with the opinions over which construct is preferable. Failure to understand this basic distinction is behind at least some of the miscommunication that happens among camps. David Sloan Wilson summarizes this point nicely.

I had thought that the recent papers on multilevel selection had turned the debate down a path to reconciliation and mutually agreeable recognition that not everyone was working on the same sorts of questions, and that different math might be better suited to these different questions. But apparently that view hasn’t yet caught on.