Wednesday, November 27, 2013

Did chariots & cavalry drive social complexity for 3,000 yrs.? (letter to PNAS)

Battle scene decoration on the chariot of Thutmose III
(click to enlarge)
(It's holiday time, so it's probably a good time for non-risk, non-InfoSec posts.  Time for war chariots!)

tl;dr:  I wrote a letter to PNAS disputing the claims made by a famous author in a recent article.  My beef was with their modeling methods. We'll see if it gets published and if the author(s) respond.

As you may know, I do quite a bit of Agent-based Modeling (ABM) as part of my PhD work. I've also studied the origins of social complexity -- institutions and political order -- including the role of the war chariot in Late Bronze Age Near East. Thus, I was quite interested in a recent article in the prestigious journal Proceedings of the National Academies of Science (PNAS):
Turchin, P., Currie, T. E., Turner, E. A. L., & Gavrilets, S. (2013). War, space, and the evolution of Old World complex societies. Proceedings of the National Academy of Sciences 110(41): 16384–16389. doi:10.1073/pnas.1308825110 (full text - free)

They use ABM to test this hypothesis over a 3,000 year period:
". . . the diffusion of horse-related military technologies from the steppe-sown interface to the rest of Afroeurasia results in a characteristic spatiotemporal pattern of spread of intense forms of warfare, leading to macrostate forms of political organization."
In simple language, they are claiming that military innovations such as the war chariot, composite bow, stirrups, and heavy cavalry were the primary causal factors in driving the timing and location of great states and empires such as New Kingdom Egypt, the Han Dynasty in China, the Roman Empire, and more, from 1500 BCE to 1500 AD.

That's a bold hypothesis!

Their ABM is quite simple (basically three variables), and their agents operate over a landscape that aimed to duplicate "Afroeurasia", at least in a crude way.

They ran their simulations a bunch of times and found that their three-variable model explained 42% of the variance the empirical data, with an R-squared of 0.65. For a simulation of such a broad geography and such a long duration, they believe that these are compelling results and that their hypothesis is supported. Here are maps of the empirical data compared to their simulated data (click to enlarge):

Looks pretty good, doesn't it?

But I saw some weaknesses in their method, both in their design choices for the ABM and in their experiments, and this led me to write the letter to PNAS.  Basically, I think their results provide weak support for the hypothesis that "horse-related military technology" is the primary driver for social complexity over this long time period.  Below, you can read the full letter (only 500 words).


The lead author, Peter Turchin, is a fairly major figure in social science (more here, too).  He specializes in population biology and "cliodynamics" — mathematical modeling and statistical analysis of the dynamics of historical societies.  He basically invented the field of "cliodynamics" and is founder and editor-in-chief of the journal of the same name.  PNAS is a high prestige journal, so getting a paper published there is a big deal for his career and the prestige of the field he founded.  Putting this all together, he and his co-authors will probably be pretty defensive about my criticisms, and they probably won't back away from any of their claims, either.

 (Coincidence: he and I are the same age.)

While I'm not a huge fan of cliodynamics (it's both too broad and too idealized for my tastes), I can sympathize with their ambitions and generally support their style of research.

The Letter

For those who want more details, here's my letter.  There's a strict 500 word limit so I couldn't go into much detail or explanation for each these points.

Does diffusion of horse-related military technologies explain spatiotemporal patterns of social complexity 1500 BCE – 1500 AD?

Russell C. Thomas, Dept. of Computational Social Science, George Mason University
Submitted to Proceedings of the National Academy of Sciences of the United States of America

Turchin et. al. [2] use an Agent-based Model (ABM) to evaluate the hypothesis “that the diffusion of horse-related military technologies from the steppe-sown interface to the rest of Afroeurasia results in a characteristic spatiotemporal pattern of spread of intense forms of warfare, leading to macrostate forms of political organization.” Because of design and simulation choices, it appears that their ABM provides less support for this hypothesis than the authors claim.

First, the core features are underspecified—too simple and too abstract—and therefore do not adequately model the labeled phenomena. “Ultrasocial traits” could just as well be interpreted as any assets, resources, knowledge, or capabilities that endow a region with power or influence, and not just ultrasocial traits. “MilTech traits” could just as well be interpreted as any factors that promote socio-cultural hegemony, not just military traits, much less horse-related military traits. The process labeled “warfare” could just as well be labeled as “accretion”—i.e. any combination of expansion, annexation, colonization, migration or warfare. In the Supplement, the authors do acknowledge the simplicity and abstraction, and propose various ways to make the ABM more specific.

Second, other than parsimony, there seems little justification for the design choice to model MilTech as a single innovation diffusion process lasting 3,000 years. A more realistic choice (not complicated) would be to model waves of innovation originating from the steppe interface, each with shorter duration.

Third, elevation appears to be a poor surrogate for defensibility. Using the same sources, it should be possible to code each region on a scale of ruggedness, e.g.  (max − min) / mean. Substituting ruggedness for elevation might significantly change the results, especially in high plateaus.

Fourth, random seeding of MilTech across the entire space is not a good alternate treatment. A better alternative treatment would be to seed MilTech (as “socio-cultural hegemony traits”) in the lowland civilizations that had already achieved a high level of social complexity in 1500 BCE: Egypt, Mesopotamia, Indus Valley, and Northern China. I suspect that the simulation results would be identical. If so, this would considerably weaken support for the hypothesis.

Fifth, running the simulation over several shorter periods and smaller geographic regions would be more compelling than the author’s choice of 3,000 years because sharper comparisons could be made between empirical data and simulations. Initial conditions would be more accurate, dark ages could be excluded, etc. It would be very compelling if the same model and parameters accurately reproduced empirical data from different periods and regions.

Finally, the results would be much more compelling if an alternate causal pathway were tested. The authors propose: “spread of military technologies→intensification of warfare→evolution of ultrasocial traits→rise of large-scale societies”. However, at least for the war chariot in the Late Bronze Age Near East, there is strong evidence that innovation and diffusion was driven by prestige and elite emulation [1] and only later took on major significance in warfare. Therefore, a good alternative causal pathway to test would be: evolution of ultrasocial traits (incl. prestige objects)→rise of large-scale societies→spread of military technologies→intensification of warfare.


  1. I read the paper and have to agree with your assessment. The use of random distribution to seed the model seems a bit like stacking the odds in their favour.

    Good luck getting your letter published!

  2. Update: My letter and the authors' reply has finally been published. Here's my blog post with summary of both plus my commentary.