The evolutionary algorithm
The American philosopher Daniel Dennett (1995) has described the whole evolutionary process as an algorithm, that is, a mindless procedure which, when followed, must produce an outcome. Nowadays we are used to the idea of algorithms, although Darwin, Wallace and other early evolutionists would not have been. Many of the things we do are based on algorithms, whether it is adding up sums, dialling a telephone number or even making a cup of tea. Our interactions with machines are particularly algorithmic and the prevalence of machines makes it easier for us to think this way – take a cup, put it under the spout, choose the drink, put in the right amount of money, press the button, take the cup out – if you do the right steps in the right order then the result is a cup of cappuccino, do it wrong and you have a mess on the floor. The computer programs that hold our medical records or run the graphics in our computer games are all algorithms, as are the ways we interface with word processors and financial packages.
Algorithms are ‘substrate–neutral’, meaning they can run on a variety of different materials. A human with a pencil and paper, a hand–cranked adding machine, and a digital computer can all follow the same algorithm for some mathematical procedure and come to the same answer. The substrate does not matter – only the logic of the procedure does. In the case of Darwin’s own argument the substrate was living creatures and a biological environment, but as Dennett points out his logic would apply equally to any system in which there was heredity, variation, and selection. This, again, is the idea of Universal Darwinism.
Algorithms are also completely mindless. If a system is set up so that it follows a given procedure then it does not also need a little mind, or extra–something, inside to make it work. It just must mindlessly happen. This is why Dennett describes Darwin’s theory as ‘a scheme for creating Design out of Chaos without the aid of Mind’ (1995, p. 50). The design simply must come about when millions of creatures, over millions of
years, produce more offspring than can survive. The ones that live do so because they are better adapted to the environment in which they find themselves. They then pass on their characteristics to their offspring and so it goes on. The environment itself is constantly changing because of all these developments, and so the process is never static.
Algorithms must always produce the same result if they start from the same point. This seems to suggest that, if evolution follows an algorithm, its results must be predetermined and predictable. This is not the case, and chaos theory explains why not. There are many simple processes, like dripping taps or moving gases, or the path drawn out by a swinging pendulum, which are chaotic. They follow simple and mindless algorithms but their end results are complex, chaotic and unpredictable. Beautiful shapes and patterns can emerge, but although the
kind
of pattern may be repeatable, the detail cannot be predicted without running the procedure right through. And since chaotic systems can be highly sensitive to initial starting conditions, a tiny difference at the beginning may lead to an entirely different outcome. Evolution is like this.
The complexity theorist Stuart Kauffman also likens the evolution of life to an incompressible computer algorithm. We cannot predict exactly how it will all unfold and can only ‘stand back and watch the pageant’. We can, however, ‘find deep and beautiful laws governing that unpredictable flow’ (Kauffman 1995, p. 23).
We can now see that even if evolution is only following a simple algorithm, it is a chaotic system and its outcome can be incredibly complex. Moreover, the results cannot be predicted without running it –and it is only being run once. We can do experiments to test predictions of the theory, but we cannot rerun the evolution of life on earth to see whether it might go a different way next time. There is no next time. Until we find life on other planets there is only this once.
Many interesting arguments remain: such as just how much pattern and order inevitably springs up in the universe even without selection; the role of historical accidents in shaping the path of life, and whether evolution will always tend to produce certain kinds of thing, such as wormlike creatures with a mouth at the front, symmetrical animals with pairs of legs, or eyes or sex. Their resolution will help our understanding of evolution enormously but none of this really matters for grasping the basic principle of the evolutionary algorithm. When this algorithm gets going the inevitable result is that design is created out of nowhere – but we cannot predict exactly what sort of design it will be. Evolution emphatically did not have to end up with us. It had to end up with
something more than it started with – and that something just happens to be this world with us in it.
Is there progress in evolution? Gould (1996
a
) famously argues there is not, but I think he has a concept of progress that I do not share. He is right to rule out progress
towards
anything. This is the whole point of Darwin’s inspiration – and what makes his theory so beautiful – there is no master plan, no end point, and no designer. But of course there is progress in the sense that we now live in a complex world full of creatures of all kinds and a few billion years ago there was only a primeval soup. Although there is no generally accepted measure for this complexity, there is no doubt that the variety of organisms, the number of genes in individual organisms, and their structural and behavioral complexity have all increased (Maynard Smith and Szathmáry 1995). Evolution uses its own products to climb upon.
Dawkins (1996
a
) describes this as ‘Climbing Mount Improbable’ – as time goes on natural selection inches up the gentle slopes to reach the heights of ever more improbable creations, and when there are strong selection pressures, progress may be maintained for many generations. Dennett describes the progress as ‘lifting in Design Space’, the crane or wedge of natural selection very slowly, and by tiny steps, finds and accumulates good design tricks by building on the efforts of all the earlier climbing. In this sense, then, there is progress.
This progress is not necessarily steady or always increasing. There are long periods of stasis between periods of rapid change. Also, some animals, like crocodiles, stay the same for long periods, while others change rapidly. And sometimes millions of years of accumulated design are suddenly wiped out, as when the dinosaurs became extinct. Some people believe that we humans are in the process of obliterating as much biodiversity as was lost in that previous extinction. If we do, then the evolutionary algorithm will start its creative work again on whatever is left.
All this creativity depends on replicator power. The selfish replicators get copied, and they do this willy–nilly so long as they have the machinery and building blocks they need for that copying. They have no foresight, they do not look ahead or have plans or schemes in mind. They just get copied. In the process some do better than others – some obliterate others – and in this way evolutionary design comes about.
These, then, are some of the general principles that apply to any theory of evolution. If memes are really replicators and can sustain an evolutionary process then all these principles must apply and we should be able to build a theory of memetics on this basis. So are they? We can now ask
two important questions – What are the criteria for being a replicator? Does the meme fulfil those criteria?
Memes as replicators
For something to count as a replicator it must sustain the evolutionary algorithm based on variation, selection and retention (or heredity). Memes certainly come with
variation
– stories are rarely told exactly the same way twice, no two buildings are absolutely identical, and every conversation is unique – and when memes are passed on, the copying is not always perfect. As the psychologist, Sir Frederic Bartlett (1932) showed in the 1930s, a story gets a bit embellished or details are forgotten every time it is passed on. There is memetic
selection –
some memes grab the attention, are faithfully remembered and passed on to other people, while others fail to get copied at all. Then, when memes are passed on there is
retention
of some of the ideas or behaviours in that meme -something of the original meme must be retained for us to call it imitation or copying or learning by example. The meme therefore fits perfectly into Dawkins’s idea of a replicator and into Dennett’s evolutionary algorithm.
Let us consider the example of a simple story. Have you heard the one about the poodle in the microwave? An American lady, so the story goes, used to wash her poodle and dry it in the oven. When she acquired a brand new microwave oven, she did the same thing, bringing the poor dog to a painful and untimely death. Then she sued the manufacturers for not providing a warning ‘Do not dry your poodle in this oven’ – and won!
This story has spread so widely that millions of people in Britain have heard it – but they might have heard another version, like the ‘cat in the microwave’ version, or the ‘Chihuahua in the microwave’. Perhaps Americans have an equivalent version in which the woman is from New York or Kansas City. This is an example of an ‘urban myth’, a story that takes on a life of its own regardless of its truth, value or importance. This story is probably untrue but truth is not a necessary criterion for a successful meme. If a meme can spread, it will.
Stories like this are clearly inherited – millions of people cannot have suddenly made up the same story by chance, and the way the changes creep in can be used to demonstrate where a story originated and how it spread. There is clearly variation – not everyone has heard the same version even though the basic story is recognisable. Finally, there is
selection – millions of people tell millions of stories every day but most are completely forgotten and only very few achieve urban–myth status.
Where do new memes come from? They come about through variation and combination of old ones – either inside one person’s mind, or when memes are passed from person to person. So, for example, the poodle story is concocted out of language that people already know and ideas they already have, put together in new ways. They then remember it and pass it on, and variations occur in the process. And the same is true of inventions, songs, works of art, and scientific theories. The human mind is a rich source of variation. In our thinking we mix up ideas and turn them over to produce new combinations. In our dreams we mix them up even more, with bizarre – and occasionally creative – consequences. Human creativity is a process of variation and recombination.
In thinking about thinking we should remember that not all thoughts are memes. In principle, our immediate perceptions and emotions are not memes because they are ours alone, and we may never pass them on. We may imagine a beautiful scene from memory, or fantasise about sex or food, without using ideas that have been copied from someone else. We may even, in principle, think up a completely new way of doing something without using any memes from anyone else. However, in practice, because we use memes so much, most of our thinking is coloured by them in one way or another. Memes have become the tools with which we think.
Human thinking (indeed all thinking) may itself depend on other Darwinian processes. There have been many attempts to treat learning as a Darwinian process (e.g. Ashby 1960; Young 1965) or the brain as a ‘Darwin machine’ (Calvin 1987, 1996; Edelman 1989). And the idea that creativity and individual learning are selection processes is far from new (Campbell 1960; Skinner 1953). However, all these ideas concern processes entirely within one brain, while the meme is a replicator that jumps from one brain to another. Darwinian principles may apply to many aspects of brain function and development, and understanding them will be very important, but this book is just about memetics.
There are many reasons why some memes succeed and others fail. These reasons fall roughly into two categories. First, there is the nature of human beings as imitators and selectors. From the memetic point of view the human being (with its clever thinking brain) acts both as the replicating machinery, and as the selective environment for the memes. Psychology can help us understand why and how this operates. There are the properties of our sensory systems that make some memes obvious and others not, the mechanisms of attention that allow some memes to grab
the available processing capacity, the nature of human memory that determines which memes will be successfully remembered, and the limitations of our capacity to imitate. We can, and will, apply this to understanding the fate of memes but it is more properly the domain of psychology and physiology than memetics.
The other kinds of reasons concern the nature of the memes themselves, the tricks they exploit, the ways they group together and the general processes of memetic evolution that favour some memes over others. These have not previously been studied by psychology and are an important aspect of memetics.
Putting together all these reasons we may be able to see why some memes succeed and others fail; why certain stories take off while others are told once and never again. Other examples include recipes, clothes fashions and interior designs; trends in architecture, rules of political correctness, or the habit of recycling glass bottles. All of these are copied from one person to another and spread by imitation. They vary slightly in the copying and some of them are more frequently copied than others. That is how we get useless popular crazes, and good ideas that never seem to get off the ground. I think there can be no doubt that memes count as replicators. This means that memetic evolution is inevitable. It is time we began to understand it.
Memes and genes are not the same
A word of caution is needed here. I have explained that the meme is a replicator and in this sense it is equivalent to the gene. However, we must not fall into the trap of thinking that memes can only work if they are like genes in other ways. This is simply not so. The science of genetics has blossomed in recent decades to the point where we can identify particular genes, map the entire human genome, and even undertake genetic engineering. Some of the insights gained from all this understanding may help us to understand memes, but alternatively some of it may just mislead us.