Terry Pratchett - The Globe

History is a complex system; its entities are people, its rules of interaction are the complicated ways in which human beings behave towards each other. We don't know enough sociology to write down effective rules at this entity level. But even if we did, the system-level phenomena, and the system-level rules that govern them, would almost certainly be emergent properties. So the rule that propagates the state of the entire system one step into the future is not something we can write down. It is an emergent dynamic.

When the system-level dynamic is emergent, then even the system itself does not 'know' where it is going. The only way to find out is to let the system run and see what happens. You have to allow the system to make up its own future as it goes along. In principle only one future is possible, but there is no short cut that lets you predict what will happen before the system itself gets there and we all find out. This behaviour is typical of complex systems with emergent dynamics. In particular, it is typical of human history and of biological evolution. And cats.

Biologists learned long ago not to trust evolutionary explanations in which the evolving organisms 'knew' what they were trying to achieve. Explanations like 'the elephant evolved a long trunk in order to suck up water without bending down'. The objectionable item here is not the reason why the elephant's trunk is long (though, of course, that can be debated): it is the phrase 'in order to'. This endows elephants with evolutionary prescience, and suggests (wrongly)

that they can somehow choose the direction in which they evolve. All this is obvious nonsense, so it's not sensible to have a theory that attributes purpose to elephant evolution.

Unfortunately, a dynamic looks remarkably like purpose. If elephant evolution follows a dynamic, then it looks as if the end result is predetermined, in which case the system 'knows' in advance what it ought to be doing. The individual elephants need not be conscious of their objective, but the system in some sense has to be. That would be a good argument against a dynamic description if the evolutionary dynamic for elephants was something we could prescribe ahead of time. However, if that dynamic is emergent, then the system itself, along with the elephants, can find out where it's headed only by going there and discovering where it gets to.

The same goes for history. Being able to put a name to a historical period only after it's happened looks remarkably like what you'd observe if there is a historical dynamic, but it is emergent.

This far into the discussion, it may seem that an emergent dynamic is no better than no dynamic at all. Our task now is to convince you that this is not so. The reason is that although an emergent dynamic cannot be deduced, in complete logical detail, from entity-level rules, it is still a dynamic. It has its own patterns and regularities, and it may be possible to work with those directly.

Exactly this is going on when a historian says something like 'Croesus the Unprepared was a rich but weak king who never maintained a sufficiently large army. It was therefore inevitable that his kingdom would be overrun by the neighbouring Pictogoths, and his treasury would be plundered'. This kind of story proposes a system-level rule, a historical pattern, which can sometimes be compelling. We can question how scientific such stories are, because it is always easy to be wise after the event. But in this case the story generalises; rich weak kings are asking to be invaded by mean, poor barbarians. And that's a prediction, wisdom before the event, and as such it is scientifically testable.71

The stories that evolutionary biologists tell are of the same kind, and they become science when they stop being Just-So Stories, justifications after the event, and become general principles that make predictions. These predictions are of a limited kind; 'in these circumstances expect this behaviour'. They are not predictions of the type 'On Tuesday at 7.43pm the first elephant trunk will evolve'. But this is what 'prediction' means in science: saying ahead of time that under certain conditions, certain things will happen. You don't have to predict the timing of the experiment.

An evolutionary example of this kind of pattern can be found in the co-evolution of 'creodonts', big cats like sabretooth tigers, and their 'titanothere' prey -large-hoofed mammals, often with huge horns. When it comes to improving performance for the big cats, the line of least resistance is to develop bigger teeth. Faced with that, the best response for the prey is to develop thicker skins and bigger horns. An evolutionary arms race now becomes pretty much unavoidable: the cats get bigger and bigger teeth, and the prey respond with thicker and thicker skins ... to which the cats' only response is even bigger teeth ... and so it goes. An evolutionary arms race sets in, with both species trapped in a single strategy. The end result is that the cats' teeth get so enormous that the poor animals can hardly move their heads, while the titanotheres' skins, and multiple horns on nose and brow, and associated musculature, get so heavy that they find trouble dragging themselves across the plains. Both species promptly die out.

This creodont-titanothere arms race has happened at least five times in evolutionary history, taking about five million years to run its course on each occasion. It is a striking example of an emergent pattern, and the fact that it plays out in exactly the same way over and over again confirms that there really is an underlying dynamic. In all likelihood it would be happening again, now, except for the arrival of humans, who have clobbered both the big cats and their slow prey.

Notice that we've been calling these system-level patterns 'stories', and so they are. They have a narrative, a consistent internal logic; they have a beginning and an end. They are stories because they cannot be 'reduced' to an entity-level description; that would be more like an interminable soap opera. 'Well, this electron bumped into that electron and the two of them got together and emitted a photon ...' repeated, with slight variations, a truly inconceivable number of times.

One of the central questions about emergent dynamics is: what would happen if we ran the system again, in slightly different circumstances? Would the same patterns emerge, or would we see something completely different? If European history in the early twentieth century was rerun, but without Adolf Hitler, would World War II have happened anyway, by a different route? Or would it all have been sweetness and light? Historically, this is a crucial question. There is no doubting that Hitler was instrumental in starting World War II; the deeper question here is whether he was a product of the politics of the time, and in his absence someone else would have done much the same, or whether it was Hitler who moulded history and created a war when otherwise nothing would have happened.

At risk of being controversial, we are inclined to the view that World War II was a pretty much inevitable consequence of the political situation in the 1930s, with Germany saddled with huge reparations for World War I, the trains not running on time ... and Hitler was merely the medium through which the national will to war was expressed. But it's not the answer that concerns us here: it is the nature of the question. It is a 'what if' question, and it is about historical phase space. It does not ask what happened; it asks what might have happened instead.

This point is well understood on Discworld. In Lords and Ladies we find the following passage: There are indeed such things as parallel universes, although parallel is hardly the right word universes swoop and spiral around one another like some mad weaving machine or a squadron of Yossarians72 with middle-ear trouble.