When we predict everything, what if we’re wrong?

“HISTORY, n. An account mostly false, of events mostly unimportant, which are brought about by rulers mostly knaves, and soldiers mostly fools.”

– Ambrose Bierce, The Devil’s Dictionary

Bierce was being funny when he wrote The Devil’s Dictionary, but his definition of history seems pretty well on target. Or so we might think, given the usual portrayal of history as the speeches, battles, and poor-to-middling decisions of kings, beggars, and senators making the same mistakes over and over again.

That “over and over again” is a problem. If history were just about the decisions of individual human beings, we’d expect their actions to look like chaos on all scales. That doesn’t happen. We see plenty of chaotic behavior in normal life, but the further we zoom out, and the larger the time scale we examine, the more regular and repetitive history appears. To explain this regularity, people have proposed plenty of theories of history, ranging from the reasonable to the bizarre. One problem with most of them is that they tend to be qualitative, or concept-based, rather than quantitative, or based on consistent relationships between numerical data. When you’re trying to systematically predict or describe events, a quantitative theory goes a lot further than a qualitative one.

So I was pleasantly surprised several days ago, when I stumbled across this post on the Long Now Foundation‘s blog: Conway’s Game of Life and Three Millennia of Human History. The post briefly describes a remarkable computer simulation of 3,000 years of Eurasian history, recently conducted by ecologist Peter Turchin and his colleagues.

Simulation? History? That means a quantitative model. I was curious.  I dove into Turchin’s report, which you can read here, along with its supporting documents.

Turchin and company created their simulation very simply: they took a map of Africa and Eurasia and chopped it up into “cells” of 100 kilometers square. Each cell was classified as sea or land; land cells were assigned elevation and further classified as desert, steppe, or agricultural land. Every agricultural cell was supplied with a “community” that could possess two types of social traits: military technology and ultrasociality. (Ultrasociality, as the study defines it, is humans’ “ability to live and cooperate in huge groups of genetically unrelated individuals.”) Agricultural cells were randomly populated with ultrasociality traits, while military technology traits were granted initially to cells bordering the steppe, and spread outward from there (a way to simulate the effect of the steppe highway on the transmission and development of military techniques, most notably mounted warfare). The cells were programmed to attack their neighbors. Victorious cells built multi-cell empires, imposing their ultrasociality traits on the vanquished. Victory was more probable over cells with low elevations and fewer ultrasociality and military technology traits than their attackers.