Here is my guess.<p>This is not the first iteration of the computer simulation. They keep tuning the simulation until it produces better and better results. But are you getting better results because the simulation is right, or because it was implicitly fit to the historical data?<p>In machine learning terms, their training set and test set are the same, so odds are that the model is bunk.
I'm so extraordinarily sick of "Computer Simulations <whatever>" being reported with any real credibility as though the Computer Simulation part means accuracy.<p>Sadly, the public can't tell the difference between "Computer Simulations Predict Aerodynamic Qualities of a Porche 911" and "Computer Simulations Show Where Noah's Ark Landed".<p>This article and simulation looks halfway between the two examples above... maybe closer to the Noah's Ark one.
After reading the paper referenced in the article (<a href="http://www.pnas.org/content/early/2013/09/20/1308825110.abstract" rel="nofollow">http://www.pnas.org/content/early/2013/09/20/1308825110.abst...</a>), I can only conclude that this is pretty cool work.<p>The authors actually do point out that their sampling method collects data in a way that is independent of the hypothesis they test, so it is not an example of cherry-picking examples that support their ideas.<p>While I cannot comment on how valid the model is because I'm definitely not an expert in that area, it seems pretty solid; they gave neighboring areas the capability to develop military techniques of certain strengths, the capability to lose it, and saw where civilizations tended to develop the different areas would "fight it out" and transfer military technique back and forth, and the result of their simulation appeared to be quite similar to the map of that time period.<p>There was also some talk in the comments about overfitting, and while as a person who works in chemical simulation I understand those concerns, this work seems to involve simply their taking initial conditions and plugging it into their simple simulation, and obtaining a result which was a remarkably good match to the actual world map of the time.<p>I do think the Ars article, like most scientific reporting, restates the conclusion in a way that is a stronger statement than the actual paper. Unfortunately, the way it was said changes the meaning of what was said in a subtle but important way. But.. that's typical scientific reporting, I guess. Overall the work is pretty cool, showing a computational model for studying the spread of military technology in a field that doesn't tend to frequently use computational models (according to the paper).
Thomas Sowell wrote a spectacular book in the same vein as this. <a href="http://www.amazon.com/Conquests-And-Cultures-International-History/dp/0465014003" rel="nofollow">http://www.amazon.com/Conquests-And-Cultures-International-H...</a><p>Extraordinary amount of detail as to why war drives the rise of new civilizations.
I prefer the beer theory.<p>Settling down and growing things doesn't seem like a military tactic that pays off in the short term. Quite the opposite actually... that is probably a great way to get slaughtered. However it does make sense if your objective is to get wasted.
Are we sure that it was definitively war or merely the availability of resources that gave rise to civilizations? Water, rich soil, minerals (E.G. Iron, Copper, Gold etc...)<p>This sort of modeling is very tricky and, even though I don't want to put down the effort that went into this, I'm not sure if they controlled for other factors. We do know that people started war over these very resources, so it shouldn't be surprising that banding together (I.E. forming societies and by extension, civilizations) was a good way to prevent these from being taken by more powerful adversaries. The technology to defend, seize and improve these resources can also build civilizations so while war may be <i>a</i> factor that is common to it, calling it a cause is a bit of a rush to judgment IMO.
Turchin published a book about his theory of the rise of powerful civilisations, and his theory about their decline, called <i>War and Peace and War</i>.<p>I reviewed it here: <a href="http://chester.id.au/2012/05/14/review-war-and-peace-and-war/" rel="nofollow">http://chester.id.au/2012/05/14/review-war-and-peace-and-war...</a><p>This article covers the creation side. It's worth mentioning that in the book he requires these wars to be between very dissimilar groups. The intense differences create a much nastier competitive dynamic, creating much higher social and cultural pressure. Eventually one group wipes the other(s) out, establishing a new imperial civilisation.
What about trade? Paul krugman (the economist known for his liberal views) wrote a compelling thesis on the role of transport costs in the rise and fall of manufacturing hubs, etc. It might explain the empires unaccounted for in this model.
> if military technology is removed as a factor, the model’s accuracy falls to a mere 16 percent<p>Imho technology drives culture. Drives our history. History focusing on dates and decisions of individuals or relations between groups of people and their character is bogus narrative. It should focus on technology available and how this technology enabled things to happen.<p>It's often claimed that our technological development outpaced our cultural development. I think that's not the case. Imho only new technology can cause culture to change. If not for development of nuclear weapons we would never could form culture that makes us civilized enough not to use it.
The graphic in the article shows that their model predicts civilizations with the highest likelihood along coasts and major waterways, rapidly declining the further inland you get .<p>It would seem like you could get a similarly inaccurate model by simply modelling people preferring to live by the coast, and only displacing inland when density along the coast gets too high.<p>I'm sure war had a substantial role, but their model doesn't seem to predict anything it couldn't have without considering warfare at all.
Intriguing concept and neat approach (tower defense anyone?), but many of us perceive civilization as arising much earlier, around 10,500 BCE.<p>Also, I cringe over this statement: "if military technology is removed as a factor, the model’s accuracy falls to a mere 16 percent."<p>That's like saying, "If you take the beans out of my soup, it doesn't taste like potatoes as much."
I have trouble with this notion because aboriginal tribes have a history of fighting amongst each other. Perhaps the rise of civilization brought on what the authors consider was war?<p>I also don't understand what predict 65% accuracy means? If we are taking about overlap between two binary variables, this is abysmally low.
That war drove the rise of -empires- seems no surprise at all (see e.g. Alexander). But is that <i>civilization</i> ("An advanced state of intellectual, cultural, and material development in human society") ?? I'm having trouble decoding the semantics to understand why this isn't trivial.
Maybe this is legit, but 500-1500 CE seems the worst historical period to be testing such models. Neither civilized areas nor technology (except in China) changed much during that period. Models that were accurate for periods such as 3500-2000 BCE or 1500-1900 CE would be much more interesting.
Interesting article, it disnt mention tha there's also a common anthropological hypothesis popularized by Claude Levi Straus that the basis of these wars was inception of the incest taboo:
<a href="http://m.vice.com/read/the-a-to-z-of-sexual-history-incest" rel="nofollow">http://m.vice.com/read/the-a-to-z-of-sexual-history-incest</a>