## Counterinsurgency doesn't compute

When I was in graduate school studying political science, I encountered a lot of quantitative theory, especially in peace research. That basically meant a bunch of professors coming up with mathematical models for predicting when wars would occur. Problem was, every time some professor came up with a theory, some other professor would shoot it down on the grounds that it explained why World War II erupted but not the Russo-Japanese of 1904.

Personally, I’ve always thought quantitative social science is bullshit. But for the U.S. military, it’s damn important. Wouldn’t it be great if the Pentagon could create a computer model of Afghanistan that would predict how the populace would react to a given strategy, or allow commanders to test out whether a population-centric approach would work better than aggressive military operations? Better to test this on a computer than on the battlefield.

But it’s not gonna happen. As I noted in Wired’s Danger Room and an earlier article in Training & Simulation Journal, military experts are grim about the prospects.. One problem is that they’re practically starting from scratch. The Pentagon built lots of computer models and simulations during the Cold War. Very geeky models of physics and firepower, as in X number of American tanks firing at Y number of Russian tanks at Z meters will destroy N number of targets. If you read that one F-22 can defeat six enemy fighters, or that a missile defense system can shoot down 90 percent of a North Korean missile strike, remember that this numbers mostly come from computer models and theoretical guesswork, because there haven’t been (thankfully) many real-life opportunities to dogfight with Chinese jets or intercept Nodong missiles.

Even if these statistics are sometimes cooked to make the Pentagon’s newest wonder weapon look good, at least the underlying physics behind them are fairly straightforward. But how do you mathematically quantify the reaction of Afghan villagers to the presence of a U.S. outpost, and how do you do it in a reliable enough fashion that General McCrystal can go, “hmmm, maybe it’s better to put the outpost here instead of there?” How do you simulate something as big and diverse as the Iraqi economy, when all those expensive models built by Wall Street geniuses told us that derivatives were a great investment?

As one military modeling expert told me, if social scientists can’t even agree on theories of mass behavior, then how can the military create  models? A computer simulation is only as good as the theoretical algorithms behind it. It’s like the movie “Go Tell the Spartans,” where Burt Lancaster plays an Army advisor battles to save a Vietnamese village from the Viet Cong while a computer in Saigon assures him the enemy won’t attack.

The Pentagon will have to create something, because planners need some kind of tool to assess various options. The question is whether those tools will get it right, or tell us that our strategy will lead to victory even as the Taliban march overrun Kabul.