"Historians distrust large-scale models"

[Bottlestein]

To everyone who is so keen on a "large-scale model" of history:

What precise field do you believe this will be based on? What field do you imagine the change of state equations in the model to come from? Evolutionary Psych? Econ? Sociology and Econ? Do you imagine these fields all have models compatible with each other? Do you believe they match-up so perfectly that the output of one model coincides perfectly with the input of the "next" model?

Forget the laughable assumptions that we know all the historical initial conditions, and the change of state equations will be anything as deterministic or simplistic as those in fluid dynamics. Do you believe you can come up with one governing equation that you can say will "predict" one single decision correctly for all Homo sapiens that ever lived, from prehistory to 2010, and can you provide any reasonable method for determining the expected failure rate for this equation?

Before any smug "well there's always Game Theory" nonsense - read some of Prof. Colin Camerer's work. He is a behavioral decision theorist / experimental economist at Caltech - one of the better schools in Game Theory. Read how many effects he identified that prevents a Game Theory based Economics model from accurately predicting the results of one game in one localized sample. Now you have a better understanding of the task of generalizing even one result from one subject to everyone who ever lived.

[Steel]

To everyone who is so keen on a "large-scale model" of history:

What precise field do you believe this will be based on? What field do you imagine the change of state equations in the model to come from? Evolutionary Psych? Econ? Sociology and Econ? Do you imagine these fields all have models compatible with each other? Do you believe they match-up so perfectly that the output of one model coincides perfectly with the input of the "next" model?

Forget the laughable assumptions that we know all the historical initial conditions, and the change of state equations will be anything as deterministic or simplistic as those in fluid dynamics. Do you believe you can come up with one governing equation that you can say will "predict" one single decision correctly for all Homo sapiens that ever lived, from prehistory to 2010, and can you provide any reasonable method for determining the expected failure rate for this equation?

Before any smug "well there's always Game Theory" nonsense - read some of Prof. Colin Camerer's work. He is a behavioral decision theorist / experimental economist at Caltech - one of the better schools in Game Theory. Read how many effects he identified that prevents a Game Theory based Economics model from accurately predicting the results of one game in one localized sample. Now you have a better understanding of the task of generalizing even one result from one subject to everyone who ever lived.

A "large-scale model" should be expected to make "large-scale predictions". You cannot and should not assess a "large scale model" on its capacity to predict specifics.

Lets model a crowd reacting to a fire in a confined space. We'll never know exactly who is going to be in a given room, and their whole life history when a fire breaks out, so why bother. You say that because we don't know little Timmy at the front burnt himself with matches when he was 3 we can say NOTHING about how the crowd will move? We cannot think about ways to efficiently position fire exits because people are all precious, unique flowers?

If you do find a first model is not sufficient to describe all you want in reality you then think about ways to augment the existing model or alternative modelling strategies. You don't declare everything is impossible and just stop all effort.

[xt828]

My understanding was that one of the problems with the larger-scale models is that, in essence, history doesn't repeat itself, so it becomes difficult to make generalisations. You mention that Hitler may not have come to power in every iteration - but we can't know either way, because we have no way of observing. Essentially, we're trying to create models from a sample of one. Correct me if I'm wrong here.

[Formless]

That's just a saying, in a sense its simply untrue. History repeats itself in many ways: for example, humans on multiple continents without ever interacting with each other developed civilization independently at different times. They developed agriculture in every instance, too. They developed warfare, government, and other institutions, all on their own without contact with one another. As its sometimes said, history doesn't repeat itself, but it sure does rhyme.

[xt828]

The grand scale stuff works, to an extent, but the details which differentiate don't so much - the lack of wheeled transport in the Americas, for example, or the near total lack of development beyond hunter-gatherer stage in Australia. On top of that, there are the cultural differences, like the different paths of foreign interaction pursued by China, the Middle East and India, and Europe, and the personal or institutional elements like the sudden rise of Islam or the death of Charles the Bold of Burgundy.

[Simon_Jester]

Models can only really deal with impersonal forces, and so Diamond's GG&S model can predict the lack of steel weapons amongst Atahuallpa's army, but it could not predict that the Inca would undergo a succession crisis shortly before Pizarro arrived.

Agreed. That said, while random crises are not predictable, the fact that such crises will occur is predictable: all monarchies have succession crises, some on a regular basis. It might be possible to predict (at least in theory) that because of a (predictable) technological inferiority, the Incas would collapse at the first instant a crisis arose. Given Spanish presence in the New World, even if Pizarro had been defeated it might not have made much difference, because future conquistadors would come, and sooner or later one of them would find a weakness to exploit.

But your general point is well taken: modeling the interaction of civilizations is very difficult if not impossible, especially when so much hinges on personal decisions.

A "large-scale model" should be expected to make "large-scale predictions". You cannot and should not assess a "large scale model" on its capacity to predict specifics.

Lets model a crowd reacting to a fire in a confined space. We'll never know exactly who is going to be in a given room, and their whole life history when a fire breaks out, so why bother. You say that because we don't know little Timmy at the front burnt himself with matches when he was 3 we can say NOTHING about how the crowd will move? We cannot think about ways to efficiently position fire exits because people are all precious, unique flowers?

If you do find a first model is not sufficient to describe all you want in reality you then think about ways to augment the existing model or alternative modelling strategies. You don't declare everything is impossible and just stop all effort.

There's still a problem, though.

Most really good and flexible models of the behavior of large groups start from making reasonable predictions about the behavior of the members that are true on average. Look at areas like fluid dynamics and statistical mechanics for examples of this in action.

If you can't model the behavior of a generic man in a crowd, you can't model the behavior of the crowd itself, because the individual behaviors have collective consequences. We can construct working models for crowd dynamics or highway traffic because people moving in a confined space act a lot like particles: they follow fairly predictable trajectories and react predictably to the presence of others nearby.

Modeling history would be far, far more difficult, though I'm not going to make any statement on whether or not it's possible.

[Channel72]

It's probably possible to model human history. The problem is that we currently have a very small, badly documented sample size. Formless' earlier comparison to evolution is interesting; we're able to model evolution because we have a fossil record which spans billions of years. Human history, on the other hand, only began ~7,000 years ago really, and most of it is undocumented or exaggerated. Over greater time scales, the variations would probably average out.

Talking about how unpredictable history is by pointing out how some King only rose to power because his rival happened to come down with diarrhea or something is missing the point; if you were to focus on some pivotal moment in evolutionary history, you'd probably discover that many species only came to exist under similarly random or unpredictable circumstances. And yet we still have a functioning model for biological evolution.