Christopher X J. Jensen
Professor, Pratt Institute

Steven Pinker makes it clear that he is not a “group selectionist”

Posted 26 Jun 2012 / 5

Frequently I feel like I am listening to an early 2000’s George W. Bush speech when the ‘opponents of group selection’ step up to the podium. Seemingly, you are either “with us or against us” when it comes to considering selection acting at a level above the individual. As someone who is open to thinking about selection acting in a variety of ways and on multiple levels, this is uncomfortable. Am I a “group selectionist”? No, I am not, any more than I am a predationist or a parasitismist: group selection is just another mechanism by which evolutionary change might happen, and I am willing to consider it as a possible hypothesis when conditions warrant this consideration.

This kind of pluralism is not displayed by Steven Pinker in his latest post on The Edge entitled “The False Allure Of Group Selection“. In this rather rambling piece Pinker lays out why he thinks we should banish the idea of group selection from evolutionary theory. While he points out a lot of real weaknesses in the way group selection has been applied to culture and human behavior, most of his arguments are more rhetorical than substantial.

One of the favorite rhetorical devices of defenders of evolutionary orthodoxy is to ‘define down’ both natural selection (in general) and group selection (as a proposed specific form) until there can be no overlap between the two; this renders ‘group selection’ outside of their definition of ‘natural selection’ and therefore makes it evolutionarily irrelevant. Natural selection has to be about selecting genes, but the orthodox see only one narrow way in which gene frequencies can change: relative propagation of “vehicles” carrying genes. They correctly argue that individual organisms are the model of this vehicle, but then take a crucial and erroneous additional step. The evolutionarily orthodox argue that since most evolution occurs because of the relative survival and reproductive output of individuals, natural selection is by definition ‘budding off’ of some ‘vehicle’ carrying genes. If your vehicle does not behave like an individual organism, it cannot be part of the evolutionary process. Pinker uses this device in this piece, and you can also see it applied by Richard Dawkins here.

This is silly. If group selection changes the frequencies of genes in a population by a mechanism that is different than that ‘selecting’ for individual fitness — for instance by causing sudden collapse of a group — that does not change the fact that evolution is occurring. If Richard Dawkins has popularized one valuable idea, it is that evolutionary processes ultimately produce gene frequency changes. But how those changes in frequency occur need not be constrained to the most common mechanism, and multiple mechanisms may be acting on gene frequencies simultaneously.

It is hard to believe that experts like Pinker and Dawkins do not realize that they are accounting for the actions of selection by counting changes in gene frequencies rather than what they claim, which is that selection acts on genes. Evolution has mechanisms: a lioness catches a gazelle, taking that gazelle’s genes out of the future gene pool and making it more likely that the lioness will have additional offspring, proliferating her genes. The genes are never “selected”: they influence the outcome of ecological processes and as a result of what happens ecologically, gene frequencies shift. Saying that genes are the target of selection is a bit like describing a soccer match by strictly explaining where the ball travelled throughout the game: you will find out who won the game, but you won’t have much of a sense of how or why the ball ended up in the net once on one side and two times on the other.

One of the ideas that Pinker seems to really struggle with in this piece is the idea of stability. On the one hand he suggests that “laws and constants of entire universes” might evolve by natural selection, but on the other hand he is loathe to consider that culture might itself be selected for. A lot of this struggle with the idea of stability is encapsulated in this pair of sentences:

Sure, some things last longer or do better in competition than others because they have traits that help them last longer or compete more effectively. But unless the traits arose from multiple iterations of copying of random errors in a finite pool of replicators, the theory of natural selection adds nothing to ordinary cause and effect.

I do understand what Pinker is saying here: there is a lot of imprecise language being used that leads to people discussing “the evolution of a building” or the “fitness of a car design”, but I think he needs to be careful in these cultural realms. Different elements of culture are more stable than others, and they do propagate. We just cannot identify the actual analogs of genes that allow culture to evolve. Clearly it is not the car or the building itself  that evolves (any more than it is the organism itself that evolves), but somehow if that car proves to be valuable its design propagates. We should not reject a Darwinian version of cultural change just because we have not fully come to understand cultural evolution.

Another one of Pinker’s favorite ‘define down’ ideas is that evolution has to be the result random processes (i.e. mutation) and unintentional design. If there is any design intent involved in a process, according to Pinker it is not evolution. This is a fascinating perspective in light of coevolution and domestication, and if Pinker is right a lot of evolution textbooks will need to be purged of the idea of artificial selection. Are domesticated dogs a species? Clearly they are distinct from wolves in many ways, so avoiding the hair-splitting issue of what exactly constitutes a species, let us say that domesticated dogs split from wolves as a result of some human intent (i.e. artificial selection). Does this mean that somehow the trajectory of dogs is no longer an evolutionary trajectory simply because some human intention was inserted into the process? Put most simply, humans and their intentions are just another component of the environment. While culture may evolve very differently than hydro-thermal vent bacteria because human intentions apply to culture, that does not mean that culture cannot evolve by a finess-maximizing process. Of course this does not mean that culture necessarily changes due to some fitness-maximizing process, but we should be open to the possibility that it could.

Pinker ignores a lot of gene-culture coevolution work in this piece. He does not seem to question that human norms are cultural inventions that influence the fate of human societies, but never really makes the connection that willingness to accept cultural norms is a biological (i.e. genetic) characteristic. Where did this characteristic come from, especially if it might be advantageous to buck the trend and disobey the norm? Can selection on individuals alone really explain a biological propensity towards obeying norms?

Pinker loves the pithy observation that a ‘fleet herd of deer is really just a herd of fleet deer’, a nice truism borrowed from George C. Williams. But for Pinker, the idea that individual characteristics scale up to group characteristics is all you need in order to understand the evolutionary process. He never seems to provide an explanation for situations in which what is good for the individual within the group is bad for the survival of the group as a whole and nonetheless what is good for the group still prevails. This seems to happen, particularly in humans!

The whole piece is a rather obvious reaction to ‘The Wilsons’ (among others), who have been getting a lot of press lately for their books espousing group selection. David Sloan Wilson and E.O. Wilson have been pounding away at the idea that evolution can be used to explain all manner of human phenomena, and interestingly those who associate with orthodox evolutionary biology do not very much appreciate this idea. In the case of Pinker this is curious because he has spent a lot of his time defending the idea that biological realities shape what we call “human nature”, but somehow when it comes to the cultural and historical elements of humanity, Pinker wants that decoupled from any evolutionary process. This is curious.

Pinker often has some really smart things to say, so I am trying to salvage something from this piece, which I mostly think is about establishing orthodoxy in evolutionary biology: patrolling the parapets of the castle of evobio as it were. One thing that I think that he is right about here is the issue of ‘currency’ in models of social selection. What does it mean for a religion to have high fitness, and how do we separate indirect measures of fitness (like winning a war or erecting a new church) from the actual propagating unit of fitness? In biology this is easy: if gene frequencies do not change, there is no evolution. For culture this is hard: no one has identified what constitutes a “meme”, although clearly there are some units of cultural propagation that must ultimately decide the fate of different cultures. That we do not know these suggests that we should be cautious, but not dismissive, of cultural evolution models.

I also think that Pinker is right to question E.O. Wilson’s analogy of bees and humans. The term “eusocial” has a nice, precise meaning when applied to a variety of colonial organisms, but I am not sure that it works for humans. We need our own category of special sociality. I will get to this topic more once I finish reading E.O. Wilson’s book.

Pinker rightly points out that coercion plays a critical role in the success of groups such as whole countries. He calls these “organized” groups, suggesting that they need to work together but that altruism is not part of the equation. I think that this is right, and we need to have a better understanding of how coercion plays into the dynamics of groups.

The question that group or multilevel selection asks is simple: are gene frequencies changing due to processes occurring at more than one level of biological organization? Since the fundamental level of selection is generally the individual organism, what we wonder when we take a multilevel selection perspective is whether mechanisms determining changes in gene frequencies occur solely at the individual level. Often they do: in a beaker full of medium, individual bacteria are dying or surviving/reproducing based on their individual ability to compete for food. But sometimes emergent phenomena at the level of groups of bacteria matter (as when biofilms form), and then we have to ask the question: are the same gene frequency changes occurring at this higher level as were occurring at the individual level, or is the ‘selective pressure’ pushing in a different direction? If the selection at the level of the individual and the group pushes gene frequencies in the same direction, noting that group phenomena matter is unremarkable. It is in those (admittedly rare, but important) cases in which group selection pushes one way and individual selection pushes another that it is important to understand that “selection on genes” is just the accounting metaphor of the population geneticist and not an actual evolutionary mechanism.

A Major Post, Adaptation, Coevolution, Cultural Evolution, Group Selection, Kin Selection, Memetic Fitness, Multilevel Selection, Natural Selection, Punishment, Web

5 Comments to "Steven Pinker makes it clear that he is not a “group selectionist”"

Artem Kaznatcheev 27th June 2012 at 3:30 pm

I am a little confused about this debate as a whole. To me the difference between gene-selection and group-selection really seems to be a question of best level of description. If I want to describe a gas under pressure pushing a piston, then I could try to describe in terms of the motion of the individual particles or in terms of thermodynamics. The two views are incompatible in the sense that certain terms defined at one level are undefined at another or have a different meaning. The only important thing that I keep in the back of my mind is that there is statistical mechanics that explains thermodynamics as a limit of certain kinds of single-particles processes. I can sleep soundly at night knowing that I didn’t introduce any ‘magic’ by thinking at the thermo level.

For me the same applies to group-selection. It seems like just another level of description. Sure, I could go down to the level of individuals but then my model becomes intractable. On the other hand, I can observe a big symmetry (like that most individuals within a group I am modeling are interchangeable, at least in a statistical sense) and move up a level of description. Of course, this thinking requires that I actually plan to model something. It seems that neither Pinker nor many of the defender of group-selection actually want to make a connection to rigorous statements and modeling. So the debate becomes one of semantics. How should we ‘use’ these words if we never bother to say anything rigorous with them? Well, I frankly don’t care.

Artem Kaznatcheev 27th June 2012 at 3:34 pm

But in terms of being careful with words. I am not sure if Pinker argued against evolution in the context you described (buildings, cars, dogs) as much as he argued against evolution by natural selection. How can you make evolution un-natural? Well, goal directed mutation is a great way to do so. If you basic mutation operator is goal-biased then I think it doesn’t make sense to call something as evolution by natural selection, and instead you might as well call it optimization; especially if the environment you are optimizing in is static-ish on the timescale of interest. In this sense the evolution of cars is not natural, but dogs are (we are not effecting the mutations, only the selective pressure).

Sorry for the two separate comments, but a minor bug report on your theme. When writing a long comment (in chrome on windows) the window expand to eventually push the ‘post comment’ button under the banner so that it can’t be clicked.

Chris Jensen 27th June 2012 at 4:03 pm

Thanks for the comment Artem! It is an honor to have you comment, and I encourage other visitors to check out your site and your published work.

It is extremely critical to tease out where points of disagreement about “kin- versus group-selection” or “individual- versus multi-level selection” are and are not a matter of semantics.

I believe that most people who think about these issues have come to acknowledge that talking about ‘selection on genes’ is extremely metaphorical: there is no mechanism in nature that actually selects for genes. Most selection “for genes” comes indirectly through the differential survival and reproduction of individual organisms. This is why I refer to population genetics as a practice of accounting: after a whole lot of ecological interactions have occurred between individuals, you can ask the question “how have gene frequencies changed?”, and this can tell you which genes were most successful (or if you want to confuse others, which genes are “most selfish”). The debate here is whether or not ecological mechanisms influencing survival and reproduction that only emerge at higher levels of organization (kin, larger social groups, subpopulations, and so on) can push gene frequencies in a different direction than mechanisms operating solely on individuals.

David Sloan Wilson (among others) has argued that kin selection and group selection are compatible, with all forms of group selection relying on increasing relatedness and all forms of kin selection relying on the ability of individuals to limit their altruism to a group of related individuals. I am still working through my own opinion on this issue, which is troubling because I purport to teach about kin selection and group selection in a number of my courses. If it is the case that advocates of kin and group selection are simply arguing from ‘opposite sides of Mount Evolution’ (to paraphrase Wilson), this debate is all heat and no light.

But I do not think that semantic disagreements alone explain this debate. As demonstrated by Pinker’s piece, there are a fair number of people who believe that evolutionary change is only caused by mechanisms that operate on the individual (notwithstanding the ones who still think that genes are subject to mechanisms of evolutionary change). They claim that you can understand all evolved phenomena simply by considering how ‘forces of natural selection’ (I chafe at the clunky metaphor of ‘natural selection’ more and more by the day) acting on individuals scale up. This is a provocative claim. You can see it in Pinker’s piece, as he tries to reduce everything we observe in human interactions to what happens between individuals. The mantra for the individual-selection-only folks is one of direct reciprocity and kin selection (even though the “individual” nature of kin selection is highly questionable). If they are right, the world is a very different place than if emergent phenomena (i.e. groups) are also subject to mechanisms of evolutionary change.

There is much at stake in this debate, and all differences are not semantic.

As to your second comment, I think that there is a real danger in defining down natural selection to exclude the possibility that culture evolves via a fitness-maximizing process (particularly on the technicality that humans are not ‘natural’, which is a pretty out-dated idea). Humans create things with intention, but that does not mean that intention leads to success. There is a big difference between cultural and biological evolution in that cultural evolution (once it gets going) can ‘compete’ alternative designs that were created with intention against each other; this is why culture evolves so rapidly. But plenty of designers have the intention to create the ‘next big thing’ and see their ideas go extinct: there is still an environment out there that allows for the propagation of some ideas more than others. I have always interpreted the “natural” in “natural selection” to refer to those elements of the biotic and abiotic environment that affect survival and reproduction, so you will have to do some pretty serious convincing to make me exclude humans from this environment.

To throw away the value of a fitness-based approach to culture seems premature: while we have not yet come to understand what a “meme” might be, no one has proposed an alternative theory of cultural change with any broad explanatory power. We need to wait and see if natural selection acts on ideas.

Sorry about the comment glitch you experienced — self-publishing is imperfect!

Artem Kaznatcheev 13th July 2012 at 1:39 am

Thank you for the kind words. Sorry for the slow response., I think I forgot to click the ‘notify me by email’ button; I will not make that mistake again.

I don’t have much to say on your first point, but I think we misunderstood each other on the second point. I was not trying to suggest that humans are unnatural or somehow separate from nature. What I was trying to say, is that we have to put bounds on what we call “evolution by natural selection” or otherwise the words become pointless. One of the boundaries I proposed was that the mutation operator has to be random, and not goal directed. I got the impression that this is what Pinker meant. However, to avoid my unauthoritative words defining something that more qualified people should define, I will instead call this “evolution by natural selection of random mutations (NSRM)” to avoid.

Artem Kaznatcheev 13th July 2012 at 1:42 am

My point was that the evolution of cars, or buildings is not NSRM. My claim is that the mutations in this sort of evolution are better modeled as randomized directed mutations. When a new car design is made, we have undergone a mutation, but this mutation was goal directed. The car designer had some fitness-function in mind, of course her function was only an approximation and only slightly correlated with the real fitness function on cars. However, I would argue that if we average over all the car-designers (i.e. if we average over all mutation operators) there is in general some non-zero correlation between what they believe the fitness function is and what the true fitness function on cars is.

Thus, when we average over all mutations, there is a tendency for the possible mutations to increase fitness. They are randomized, but goal-directed. I would not call such mutations random, but instead random directed mutations. So the evolution in this case is not NSRM, but a selection on randomized directed mutation. My claim was that this sort of selection should not be called ‘natural’, however this is a matter of terminology. The key point is that the two types of evolution are different. One has mutations that don’t take into account fitness and just happen, the other has a mutation operator that has partial information about the fitness landscape.

Leave a Reply