Reply to Peter Godfrey-Smith on selfish memes

Here's Peter Godfrey-Smith criticizing selfish genes and memes:

Two kinds of agential narrative have a special psychological potency. The first is a paternalist schema. Here we posit a benevolent agent, often a large one, who intends that all is for the best. This category includes various Gods, the Hegelian "World Spirit" in philosophy, and stronger forms of the "Gaia" hypothesis, according to which the whole earth is a living organism. The second schema is a paranoid one. Now we posit hidden agents, often small, pursuing agendas that cross-cut or oppose our own interests. Examples include demonic possession narratives, the sub-personal creatures of Freud's psychology (superego, ego, id), and selfish genes and memes. And while it is true that sometimes there are large and kind agents or small and nefarious ones at work, the psychological appeal of these ideas means that we tend to take up such stories too readily and run with them too far. The account of evolution in terms of "selfish genes" (Dawkins 1976) is a paranoid narrative of this kind. It relegates other entities in evolution, such as whole organisms, to the role of mere "vehicles."

This is a situation where a communicative device or heuristic has been allowed to take on too substantial a role; it becomes a kind of foundational description. Instead, the way to think of gene-level evolutionary processes is like this. Any collection of entities which vary, inherit characteristics in reproduction, and differ in how much they reproduce will evolve by natural selection. These include entities bigger than us, like social groups, entities smaller than us, like cells and genes, and organisms like us. As long as they satisfy the requirements of variation, heredity, and fitness differences, they will behave in a Darwinian way. The recognition that genes have the necessary features – they vary, inherit features in replication, and differing how much they are replicated – is the recognition of one of Darwinian population among others. It is not true that when we find small things doing this, inside us or underneath us, we're finding what it's all about, what it all means, the agents whose plots and programs are behind everything else we see.

I like this passage because it is eloquently put. However, I think we have to label it as misleading.

Richard Dawkins famously argued against individual and group selection because individuals and groups didn't "replicate" - and instead recombined. While there's clearly something to this, we have subsequently learned that this isn't really a convincing argument against all group selection. Wade's flour beetles convincingly showed that group selection and gene selection could peacefully coexist.

Godfrey-Smith is making much the same argument as Dawkins - but the other way around. Like Dawkins he's arguing that group selection and gene selection are incompatible. Peter likes the idea of selection at multiple levels, but thinks that this invalidates gene-level and meme-level selectionism. Dawkins apparently [1] thought that gene-level and meme-level selection invalidated group selection. The truth is more like: selection on memes and genes is one way of looking at things and selection at multiple levels is another. Individual and group-level selection don't disprove gene-level selection - they just show that it isn't the only way of looking at things. However, it was never claimed to be the only way of looking at things in the first place.

Compare memetics with atomism. Ballistics has different principles from atomic physics - and it is useful in cases where atomic physics is not. However, Ballistics doesn't disprove atomic physics. It's just working on a different scale. No violations of atomic physics are involved.

Stories of selfish genes and memes don't deserve to be described as "paranoid narratives". That really is what's going on. Sure: it's not the only way of looking at the world, but it is a valid way of looking at it. Indeed, it's the most complete and comprehensive way of looking at things - though the associated models are not always tractable. If you work at higher levels you miss out on the details.

[1] I realize that Dawkins was probably attempting to verbalise a valid point about how too much recombination often messes up the possibility of adaptations at higher levels.

Update: 2013-09-29: I notice that David Queller has an excellent Godfrey-Smith-smackdown on this topic here. Queller is right.

References

• “Selection pressures” are metaphors. So are the “laws of physics.”