The interactor/replicator framework

Soon after Lewontin provided his original formulation of the ‘recipe’ that units of selection must satisfy, Richard Dawkins (1976/2006) published his well-known book The Selfish Gene, where he proposed his gene-centred view of the evolutionary process and suggested a new way of conceiving the units of

selection controversy, which in certain sense specifies Lewontin’s framework.

David Hull (1980, 1981, 1988), and became the standard view among biologists and philosophers of biology for the next two decades (Dawkins 1982a, 1982b; Brandon 1982, 1988; Sober and Lewontin 1982; Sober 1984; Mitchell 1987;

Lloyd 1988, 1992, 2001, 2017a; Sober and Wilson 1994).79 _{Let me put a little bit}

of context here, before I introduce Dawkins and Hull’s framework, since it will be useful to fully comprehend why the authors decide to suggest their specific type

of approach to address the controversy about the units of selection.80

Before Dawkins published The Selfish Gene, there was an agitated debate among evolutionists about which was the entity or entities that natural selection could ‘select’, i.e. which was the level of the biological organisation at

which selection was more effective.81 _{During the 50s and the 60s, it was}

common among biologists—who mainly followed Darwin’s observation about the possible action of natural selection ‘for the good of the group’ (1871, The Descent of Man)—to interpret some types of animal behaviour as adaptations that benefitted the group or species that the animal belonged to. Such general appeal to group benefit was explicitly endorsed by V. Copner Wynne-Edwards in his major work (1962): Animal Dispersion in Relation to Social Behaviour, where he acknowledged the general existence of between-group selection in nature, and showed no objection in postulating it as one of the main causes of evolution and the existence of adaptations. Wynne-Edwards’s views comprised a forceful attack on the Darwinian orthodoxy according to which the individual organism is the main unit of selection, i.e. it is the main unit on which adaptations evolve. Wynne-Edwards views, however, were soon criticized by Maynard-Smith (1964) and, especially, by Williams (1966). Williams’ criticism

was based on the existence of a conceptual distinction between ‘adaptations’,

and ‘fortuitous benefits’: in his view, adaptations look at the past history of the trait, whereas fortuitous benefits look at its future. Wynne-Edwards confuses the

79 _{A notable exception to this trend is developmental systems theory (Oyama 1988; Griffiths and}

Gray 1994, 1997).

80 _{As in most philosophical debates, distinctions and alternative frameworks to conceive the}

same process are usually introduced to overcome one problem or set of problems that philosophers see as ‘urgent’ and in need of a solution. The same is true about the units of selection controversy, so introducing the context is fundamental, and prior, to introducing Dawkins and Hull’s interactor/replicator framework.

81 _{The reconstruction is based on Okasha (2001): ‘Why won’t the group selection controversy go}

away?’, Br J Philos Sci 52: 25-50, and in Gould (2002): The Structure of Evolutionary Theory, pp. 544-556.

two notions and, thus, his arguments about the widespread existence of group selection are just misguided.

Let me explain Williams’ idea more succinctly, because his criticism is fundamental to understand the debate about the units of selection. According to Williams, a trait is said to be an adaptation for doing p if and only if the trait evolved because there was selection for the trait, and there was selection for the trait precisely because having the trait promoted doing p (Sober 1984: 208). Applied to the context of group selection, a trait can be argued to be an adaptation for the group if and only if there was group selection for having the trait (Sober 1993: 85). However, and this is the key argument to be found in Williams’ criticism of Wynne-Edwards, the fact that one trait now benefits the group of entities that bear that trait does not mean that the trait was selected in the first place because it benefitted the group. Notice the causal language that appears here: to say that a trait evolved because it conferred benefits to the group whose individuals bear the trait entails that one is able to tell a causal story of why the trait is there, a causal story that appeals to the benefits that the trait confers to the group. As I said in chapter II, natural selection provides a causal explanation of why a trait exists, so the existence of a causal story that can be told is fundamental to talk about group adaptations.82 _{Williams’ point}

against Wynne-Edwards is, thus, that what he regards as traits that have been selected for the good of the group might benefit the group now, but this benefit: 1) might not be the reason why the trait exists on the first place (i.e. it does not have the right kind of causal history); 2) might be the consequence of kin selection or inclusive fitness theory, which is presented as an alternative to group selection that would explain the existence of the same type of traits that group selection explains, without appealing to a benefit for the group (see Hamilton 1964a, 1964b, for the concept of ‘kin selection’).

82 _{There is a whole debate about what natural selection really explains (Stegmann 2010; Birch}

2012; Díez and Lorenzano 2013; Mogensen 2016), or about whether explanations that appeal to natural selection are causal or rather statistical (Walsh et al. 2002; Brandon and Ramsey 2007; Lewens 2010; Huneman 2012). I will not enter in any of these disputes here, as that is outside the scope of this thesis. I will only appeal to the explanatory nature of the notion of ‘natural selection’ for illustrative purposes.

About 1), Williams made the following point: the individuals that compose a group have generally shorter living times than the groups themselves; if this is so, then the opportunity for between-group selection will systematically be eroded by the opportunity for between-individual selection. Why? Because for natural selection to be an efficient driver of evolution, it is necessary that the degree of variation within the object under consideration is low, so that the possible effects that a specific variation v might have on the fitness of the objects that bear v allow these objects to outcompete other objects in the populatin that do not bear v. However, if each object in the population is constantly varying due to selection at the lower level, then natural selection is not possible. This condition, he believes, is not satisfied among groups, since they experience constant variation and, therefore, the opportunity for group selection is rare, compared to the opportunity of individual selection. Concerning 2), Williams develops the following line of reasoning: According to kin selection, some traits (e.g. offspring caring behaviour) that would in principle seem like group adaptations, are in reality individuals adaptations, because the individuals that evolve those traits will be in a clear fitness advantage over those that do not. Importantly, for an adaptation to be counted as a case of kin selection, it is necessary that the trait evolved to benefit those that are alike, i.e. to benefit the relatives (offspring, or tightly connected relatives). Those cases, detractors of Wynne-Edwards argue, are not to be counted as cases of group selection and, furthermore, most cases of what Wynne-Edwards counts as group selection are indeed cases of kin selection.

Given the context aforementioned, it is now time to introduce Dawkins’

solution to the tension between group selection and kin (or individual) selection, which was just the tension between Wynne-Edwards and Williams. It must be noted, although in passing, that by the moment when Dawkins introduces his solution, this was seen more as a ‘challenge from below’—and Wynne- Edwards’ model was interpreted as a ‘challenge from above’—in the sense that it also seemed to question the role of the organism as a unit of selection. However, this perception was somehow misleading, since one of his points consisted precisely in defending that the organism is also a unit of selection, in at least one of the two different meanings that the concept can adopt (Dawkins

1982a; see Wilkins and Bourrat 2018: section 3). Let me introduce his solution to the tension by quoting Dawkins’ own words in the ‘Introduction to the 30th

Anniversary Edition’ of his opus magnum, where he re-evaluates the importance of The Selfish Gene for the debate about the units of selection:

‘I should perhaps have gone for The Immortal Gene. The Altruistic Vehicle (…). Perhaps it would have been too enigmatic but, at all events, the apparent dispute between the gene and the organism as rival units of natural selection (a dispute that exercised the late Ernst Mayr to the end) is resolved. There are two kinds of unit of natural selection, and there is no dispute between them. The gene is the unit in the sense of replicator. The organism is the unit in the sense of vehicle. Both are important. Neither should be denigrated. They represent two completely distinct kinds of unit and we shall be hopelessly confused unless we recognize the distinction.’ (Dawkins 1976/2006: ix)

This paragraph introduces Dawkins’ popular view about the evolutionary process. The evolutionary process is, for him, the process by which natural selection causes the evolution of two special types of entities: on the one hand, the gene, which is the unit of selection in the sense of being a replicator; on the other, the organism, which is the unit of selection in the sense of being a vehicle. But, what does this jargon exactly mean? And, more importantly, what is the significance of introducing two units in a process which originally involved the existence of only one?

Dawkins believes that we must carefully distinguish between: on the one hand, the units that have the properties that a successful unit of selection must have; and, on the other hand, the larger units that these original units must form to foster their own success. Dawkins refers to the first units as replicators. Replicators are the entities of which copies are made, and that share three basic properties: longevity, fecundity, and copying-fidelity. Why are these three properties necessary to determine the degree of success of the replicator? The answer to this question is simple: if the phenotypic effects of an entity are going

to outcompete the phenotypic effects of another, then the entity must be long- lived enough so that the effects can accumulate, a lesson we learnt from Williams. Longevity, fecundity, and copying-fidelity guarantee that this is so. The unit that fulfils these three properties is the replicator, and the (selfish) gene is

for Dawkins the paradigmatic case of a replicator.83 _{On the other hand, Dawkins}

also recognises the existence of a different type of unit of selection that he calls the vehicle. The importance of introducing the vehicle and its necessity as a unit of selection becomes clear in the following passage taken from Stephen J. Gould:

‘I find a fatal flaw in Dawkins’ attack from below. No matter how much power Dawkins wishes to assign to genes, there is one thing he cannot give them –direct visibility to natural selection–. Selection simply cannot see genes and pick among them directly. It must use bodies as an intermediary.’ (Gould 1977: 24, emphasis added)

Notice the problem that Gould is pointing at. It might well be true that replicators are long-lived, that they have a high-degree of copying-fidelity and that they are fecund. However, they are invisible to natural selection, which can only see and, thus, proximately select, phenotypic differences among individuals. But notice that this is a consequence of the fact that replicators tend to appear combined with each other forming chromosomes, cells, organisms, multispecies communities, etc. (Lewontin’s hierarchy). All these structures are not replicators: they are vehicles that: 1) are directly visible to selection, and 2) their success ultimately determines the rate of success of the replicators that compose them. In Dawkins’ words: ‘Vehicle selection is the differential success of vehicles in propagating the replicators that ride inside them.’ (Dawkins 1982b). At this point, two things become clearer: first, that part of the disagreement about the units of selection turns out to be a semantic question, due to the fact that different researchers mean different concepts by using the

83 _{Notice that the replicator is a conceptual tool, the name to refer to the unit of selection that}

satisfies the three properties, and the gene is the biological entity that satisfies the properties. In Dawkins’ framework, however, other entities, such as memes, can also act as replicators and, what is more interesting, due to the empirical nature of the category, it is perfectly conceivable that other entities apart from genes could also be replicators, provided they satisfy the desired properties.

same expression (‘units of selection’ is polysemic);84 _{second, Dawkins’}

observation in the 30th _{years Edition of The Selfish Gene, where he claims that}

a more adequate title would have been The Immoral Gene. The Altruistic Vehicle.

Let me start from the beginning. In which sense is the dispute about the units of selection semantic? Dawkins gives the following example: ‘The controversy about group selection versus individual selection is a controversy about whether, when we talk about a unit of selection, we ought to mean a vehicle at all, or a replicator.’ (Dawkins 1982b). In Dawkins’ view, Wynne- Edwards, as well as other defenders of group selection, are referring to the vehicles, insofar as groups are in part responsible for the success of the replicators that make up the group. On the other hand, Williams, as well as other detractors of group selection, usually mean the ‘replicator’ when they talk about the unit of selection, and thus strongly oppose Wynne-Edwards’

hypotheses about group selection.85 _{Thus, the dispute between group selection}

and individual selection is partially a semantic dispute, as the authors mean different concepts by the expression ‘unit of selection’. However, the dispute is only partially semantic, since sometimes they also disagree about which entity is the vehicle of selection, and there is a genuine empirical debate, as Dawkins reminds us: ‘the organism and the group of organisms are true rivals for the vehicle role in the story, but neither of them is even a candidate for the replicator role. The controversy between “individual selection” and “group selection” is a real controversy between alternative vehicles’ (Dawkins 1976/2006: 254-255).

Second, Dawkins believes that a more appropriate title for his book would have been The Immoral Gene. The Altruistic Vehicle. Why? Because in Dawkins’ view of the evolutionary process, the real beneficiary of evolution, the

84 _{A semantic question, but not a verbal dispute. There is a genuine semantic disagreement}

about the meaning of ‘unit of selection’, the disagreement is not merely about what to call the different units of selection.

85 _{It is not strange that Dawkins explicitly acknowledges that the properties that he attributes to}

replicators, and that he had polemically expressed in The Selfish Gene by demanding the immortality’ of replicators (1976/2006: chapter 3; see also 1982b), derive from the properties that Williams attributes to the units of selection.

real unit that has the right properties and that pursues its own interests is the replicator—which is empirically the ‘gene’, conceived as ‘a piece of chromosome which is sufficiently short for it to last, potentially, for long enough for it to function as a significant unit of natural selection’ (Dawkins 1976/2006: 45)— whereas vehicles are there altruistically, just to serve the interests of the

replicators that make them up. Importantly, in Dawkins’ view, the replicators

make their vehicles, and the replicators that survive longer are those that make better vehicles, as the vehicles are the entities that compete which each other to pass on their replicators. Using the conventional distinction between proximate and ultimate causation (Mayr 1961), it could be argued that the replicators are the ultimate beneficiaries of the evolutionary process, whereas the vehicles are only the proximate causes that determine the success of the former.86

Dawkins’ view of the evolutionary process is very controversial, and has been subjected to a lot of criticism, that I will not examine here. But one important criticism that I will examine came from David Hull, who made a fundamental contribution by clearing up the notions of replicator and vehicle that Dawkins had originally introduced (Godfrey-Smith 2000: 404-407). For Hull

(1980, 1981, 1988)87_{, the process of natural selection encompasses two}

different types of entities: on the one hand, an entity whose structure persists throughout evolutionary time; on the other, an entity that has the capacity of interacting with others and, thus, causes differential replication. According to Hull, these two types of entities had not been clearly defined by Dawkins, and even if he had introduced the notion of replicator and vehicle, the distinction was not clear enough to make sense of these two roles. The two entities introduced by Hull came to be known, respectively, as the replicator and the interactor. To quote:

86 _{Lisa Lloyd would disagree with referring to the replicator as the beneficiary of the evolutionary}

process, because in her view of the units of selection, there are not two, but four different questions (Lloyd 1992, 2001, 2017a). I think she is right that the question about which entity is the replicator and the question about which entity is the beneficiary of the evolutionary process are not necessary the same question. However, this does not refute the claim that Dawkins mixes the two categories in one, and that’s the reason why I introduce Dawkins’ replicator by calling it the ‘ultimate beneficiary’.

87 _{It is conventionally accepted to use the definition of the replicator that Hull uses in his (1980).}

However, I think the definitions that he gives in his excellent (1988), is far more precise, as it eliminates the weird requirement of ‘directedness’ from the definition. I will thus take (1988) as my referent.

‘[I]n an effort to reduce conceptual confusion, I suggest the following definitions:

replicator – an entity that passes on its structure largely intact in successive replications

interactor – an entity that interacts as a cohesive whole with its environment in such a way that this interaction causes replication to be differential

With the aid of these two technical terms, selection can be characterized succinctly as follows:

selection – a process in which the differential extinction and proliferation of interactors cause the differential perpetuation of the relevant replicators.’ (Hull 1988: 408-409).

I now need to explain more clearly what Hull was getting at with this distinction. To do so, let us go back to Gould’s criticism of Dawkins. According to Gould, the main problem with Dawkins’ replicator-centred account was that genes (or replicators, see fn. 83) were not directly visible to the process of natural selection. And this was a non-salvable obstacle for the replicator account. Remember again a point I made in chapter II, and that has been repeated several times: natural selection is a causal process. Evolutionary change can be driven by many different types of processes; those different processes generate a sorting among the interacting entities. Among those processes, selection is the causal process. Relegating natural selection to the persistence of replicators simply misses the main point. Hull’s vision of natural selection, on the contrary,