Review of Memes: the New Replicators

Chapter 11 of The Selfish Gene by Richard Dawkins

One of the themes of Richard Dawkins' work is that Darwinian evolution by natural selection is not a peculiar feature of life on earth, but probably universal to all life or lifelike systems in the universe. It might be argued that the fundamental feature of life is the spontaneous (at least in the first instance, for the atheists among us) arisal of a replicator --- something which reproduces itself, and which is special relative to the rest of brute matter precisely because it replicates itself. Each new copy is by definition itself a replicator and makes copies of itself, thereby tapping in ideal circumstances a phenomenal (to our everyday perceptions) exponential growth in population. Such numbers turn what might be thought of as low-probability events into low-frequency events, and some of the replicators have mutations --- are different in some way than their parent. These mutants will, by consideration of basic probability, probably be worse than their parents. But some will be better at replication, and will increase in proportion of the population. As external selective factors come into play --- the eventual lack of resources or the evolution of predators or parasites to drive an arms race --- complexity and diversity may grow. Experiments in cybernetic ``artificial life" --- our only way of testing lifelike processes somewhat independent of Earth-evolved life --- seem to support the above post-Darwinian armchair philosophizing.

Earthlife may be seen then as the result of some early replicator --- one of a class of RNA molecules, or some particular clay mineral formation --- and the long time (and many generations) since then. Note that if clays were the first replicators, as some biologists seriously speculate, then our chemical form of life based on the replication of RNA and DNA has bootstrapped off of that slower and now obsolete form. Dawkins' specific point in this chapter is that we may be witness to and participants in the arisal of a new replicator, which he calls memes, bootstrapping off of our minds. The first replicators presumably replicated by shaping the materials that randomly came into contact with them. Today's DNA replicates by directing the bodies housing instances of DNA to gather the materials for new copies of the genes and bodies. Dawkins' memes replicate by being imitated in different minds. If my niece sings the theme song of Barney the dinosaur and I start playing that theme in my mind then we can say the meme has spread. A new instance has arisen by virtue of my imitating the expression of the instance in my niece's mind. We can also say that I have been infected by that meme. By expressing the idea of the meme in his book in such a way that many of the people who read it believed the idea and told it to other people Dawkins literally created a new meme which has been spreading with fair success through the population.

There are some interesting consequences of this way of thinking. If we try comparing and contrasting genetic with memetic life we find some significant differences. DNA reproduction is a rather complex process: genes create bodies which eventually create new bodies, with new genes inside them. In the process those bodies may do many things: fly, hunt and eat the bodies manifested by different genes, court members of the opposite sex, or think. And, given cellular machinery designed to reproduce DNA, viruses have arisen which slip in and parasitize the system for their own propagation. Whereas it seems to me that the memetic level, while messy, is not nearly as complex. Memes are reproduced through straightforward communication and imitation, a process analogous to both RNA replication in the primeval soup and to genetic viruses --- the memes exploit complex machinery designed to run things like themselves. I find it difficult to imagine what a fuller memetic ecosystem might be like, but I envision AI lifeforms where the memes and genes have merged --- the next generation is consciously designed, combining replication of memes and bodies --- and where new memes (more broadly, new ideas) can result directly in self-modifications. Or in other words, memotype would determine phenotype but phenotype could influence memotype, creating a much more Lamarckian situation. Imagining a full set of analogues is still difficult --- what is the memetic equivalent of predators? Plagiarists?

A few more complex levels have been elucidated however, particularly meme-complexes and mental immune systems. Memes that benefit each other can form co-evolving meme-complxes, analogous to the sets of genes that form a body or the gene pool of a species. And as a major neo-Darwinian realization was that other genes in a species themselves are part of the environment for determining the success of any gene, so other ideas in a mind or a culture will often play a key part in determining the success of a meme. And mental defense systems can be seen as analogous to biological immune systems, and these defenses may themselves be determined by transmissable memes. Examples of the latter would be ``reject all new ideas" (so simple as to be a simple meme that locks up the infected mind, if it can gain access) or the full spectrum of skepticism and scientific criticism. These defenses create more obstacles for memes to overcome; an idea that contradicts someone's religion or fails scientifically may not take, and some systems may introspectively purge their hosts. These obstacles should act as a selective pressure on other memes. Of course we might expect competition among the defenses themselves, if they have to be taught (which is just a form of memetic reproduction and infection).

But it may be legitimately asked how useful this terminology is, and whether it tells us anything useful. A basic idea in considering simple replicators is that a small replicator will do better than a large one that does the same thing, simply because the smaller one takes fewer resources. There can be more of it in a finite pool, and copies may be made more quickly --- fewer Brownian motions are needed to assemble a smaller RNA molecule, for example. So a straightforward application to memetics might lead to the hypothesis that small memes will do better than large ones. But this can be restated as small ideas do better than big ones because attention spans are limited, and sloganwriters have known for a long time that short snappy slogans do better than cumbersome phrases. And many other statements about memes and infection can be rephrased more conventionally in terms of ideas, communication, and culture. The above analysis might be telling us why short slogans do better, but I do not think it is proven that memetic phrasing will tell us anything that could not be discovered in some other way. In fact, while I am taken with the idea of memes myself I think it likely that there will be nothing that cannot be rephrased in conventional terms.

Which does not mean that Dawkins' idea is useless. Most computer languages are completely equivalent in algorithmic power, but some tasks are much easier in certain languages. Similarly considering ideas as memes may highlight selective forces in a useful and new manner, providing a new angle from which to make discoveries.

One direct consequence for evolutionary psychology of this way of looking at things is, as Dawkins points out , that many behaviors may have been memetically selected for rather than genetically selected. The sociobiologists are themselves conscious of the dichotomy between cultural and evolved behavior but memes allow us to recast the debate in entirely Darwinian terms: we may agree that some behavior has been selected for, but has the level of selection been genetic or memetic? It is possible that attractive genetic explanations of common behavior may be more truthfully explained as a successful meme, perhaps exaggerating some slight genetic knob. Whether even pancultural similarities are sufficient for the evolutionary psychologist is debatable; strong arguments using other primate data are even more useful. (Except that the possibility of cultural transmission in the other apes throws their evidence into doubt as well. Ironically, those animals closest to us genetically and thus best evidence for standard evolutionary psychology arguments are, by that same closeness, those who still share the mental complexity that causes confusion in tracing selected behavior.) And one critique of ``Wilson's ladder" claimed that Wilson's sociobiology depended on selection only happening at the genetic level. Obviously in Dawkins' scheme of things this is false.

Another consequence is that even if Wright's evolutionarty psychology is completely correct the psychology of self-adaptive AIs would not necessarily be easily deducible by applying the same principles. Or rather, those principles could be applied but the conditions will probably be quite different. Sexuality is the most obvious casualty, but I think all analysis based on clearly delineated generations will be inapplicable. Memes are like viruses, and replicate simply by communication --- in fact that is how they replicate now, and not by being instanced in a newly built child. But insofar as to build an AI we will probably need a good understand of what intelligence is the next generation of mindforms could be the result of a vastly complex system of idea trading and cooperation --- far unlike the simple modes of sexual or asexual reproduction biology has today.

One feature Dawkins shares with Robert Wright is a belief in the goodness of altruism. ``We, alone on earth, can rebel against the tyranny of the selfish replicators." While I do not believe in group selection, for all the standard reasons, I think individual humans can be considered a replicative and selective level. Not, I hasten to note, a very good one, as we neither replicate ourselves with any fidelity nor live very long, but there is still the basic selection of whether we survive or not. And the absolute altruism Dawkins and Wright seem to advocate is fundamentally hopeless. We actually may be selfish to an extent because it has been good for our genes to program us in that manner, or for our memes, but even if we could somehow isolate ourselves from our memes I feel we would still have to be selfish. The decision to protect one's life against another trying to take it is the fundamental selfish act --- the definition of self, almost. And those who do not so decide can fall like grass before before those who do. And to a large extent, we may thank (if so inclined) our genes and memes for much non-selfish behavior we show. Immortality is the ultimate selfish desire, short of wishing to personally conquer the entire universe. To strip us of the constraints of the lower levels of replication may not cause great altruism, but great selfishness, altruistic only in the reciprocal sense, as constrained by the threats of other selfish beings.

In conclusion, Dawkins' idea is that ideas can be thought of as mental replicators analogous to genes, and that the principles of variation and natural selection can be applied to them. The most direct consequence of this for evolutionary psychology is that behaviors may be selected not at the genetic level, for the benefits provided to genes, but at the cultural or memetic level, because those behaviors had properties that propagated themselves in the mental environment in which they exist.

Back to Me.