Defeating IEDs with Data

In our Afghan counterinsurgency, some of our best weapons are exactly those traits of our society that are least comprehensible by the insurgents—and most alien to their culture. The popular image of the Americans’ big bombs and drones fighting IED-planting guys in flip-flops and pajamas who live in mud-brick houses isn’t false. But it’s perhaps the least interesting part of the picture. A combination of social and mathematical analysis is likely to be a better weapon in the long run. And these abilities simply don’t develop in closed societies where critical thinking is not valued—like the jihadi circles of the Middle East and Afghanistan.

IEDs—improvised explosive devices, for those who have just awakened from a sleep of several years—are the leading killer of our military in Iraq and Afghanistan. They are particularly critical in Afghanistan, where terrain much more rugged than that of Iraq means that often there is only one way in and out of an area by road. Since 2007, a huge, well-funded cross-disciplinary military group called JIEDDO (Joint Improvised Explosive Device Defeat Organization) has been working on ways to eliminate the IED threat.

The obvious one—up-armoring IEDs to make them bombproof—is in some ways the least productive. It takes a lot longer to get a new up-armored vehicle through production and testing than it takes an Afghan bomb-making cell to adjust its improvised explosive devices to destroy a vehicle and its occupants. It’s also not clear that bigger is better for troop transports; as paratroopers in the 82nd Airborne told me in Zabul Province, Afghanistan last November, bringing in the enormous Stryker vehicles to the province this past summer “just made the IEDs bigger.” (The Strykers were sent to Helmand this winter.) And another obvious effort, training troops to avoid and detect IEDs, is necessary, but it doesn’t solve the root causes of the problem.

JIEDDO’s third and most intellectually fascinating goal is to map and eliminate the insurgent networks that build and plant IEDs and discover the cause-and-effect of creating explosives.

“The enemy can adapt its tactics faster than we can field new equipment, so we are focusing on analytics,” said Major Sam Huddleston, an instructor at West Point and a former brigade planner in Iraq, who sat down with me on February 12 to describe his efforts to help senior Army commanders defeat IEDs. The topic could not have resonated with me more than at this very moment: I was at West Point to attend the funeral of Captain Dan Whitten, a 28-year old paratrooper of great promise who was killed by an IED in Zabul Province, Afghanistan on February 2.

Huddleston is preparing a book for internal Army use explaining how a commander can use trend models, forecasting, crime hot-spot maps, spatial choice models, risk assessment, and network analysis to defeat IEDs in an area of operation.

One of the most promising fields is the relatively new discipline of social network studies. “We’ve got some of the best network analysts in the world floating in and out of JIEDDO,” Huddleston said. (He later e-mailed me Chris Wilson’s terrific series on Slate about how the heavily mathematical theory was used to find Saddam.)

Moving deeper, to the whys of IEDs, involves studying the Iraqi or Afghan population. Do poorer areas breed IED-makers? Are they associated with certain tribes? As an embedded reporter in Afghanistan, I’ve seen that this work is complicated by the absence of government-collected data. Estimates of the population vary by as much as 10 million people.

In an op-ed I wrote last December in The Wall Street Journal
, I lamented the unsophisticated data analysis I saw at the battalion and company levels in the course of five embeds with the American Army in Afghanistan.

I charged that while we are fighting a population-centric war, we are graphing a terrain-centric war. I’d never seen a PowerPoint slide with an “x” and “y” axis establishing a correlation between sets of data. Most of the Army’s massive production of slides revolves around maps.

Colonel Tim Trainer, the head of West Point’s systems engineering department, very graciously addressed some of my concerns. He explained that this sophisticated data analysis goes on at a higher level than I had seen [INS: during] my embeds. I’d been spending time on the battalion level—a group of about 600-700 troops commanded by a lieutenant colonel. The analysis I wanted to see takes place on the brigade level (3,000 troops commanded by a colonel) or higher. I subsequently heard from a number of ORSAs (Operations Research and Systems Analyst—basically, Ph.D.-level analysts) deployed in Afghanistan.

Huddleston explained that there was a mathematical reason it made more sense to analyze some data at the brigade level. “At a company level, this analysis does no good. It is some help at the battalion level, and a lot of help at the brigade level. If you look at the IED frequency in a battalion’s area of operations, it may have a lot of noise, spiking up and down. But if you look at the whole country, you may see a clear trend.”

He compared the situation to the “noise” in a high-beta (high variance) individual stock’s performance compared to the more predictable variations in the stock market as a whole. The uncertainty of results of statistical analysis falls as you go up from company to battalion to brigade level.

The problem, from the standpoint of the Army’s ORSAs, is that everyone who becomes a general spent intellectually formative years at the data-noisy company command level. So many commanders who could gain from ordering data analysis are skeptical of it. For instance, a commander without much background in data analysis may look at a map of IED sites and think, “Bad place to go.” But with more sophistication, he might add, “Good place to set an ambush for the enemy…especially at such and such a time on a Friday.”

Also, commanders are crazily busy in the field. And the Army’s higher education program for field-grade and general officer levels—when they have time to work on their skills and analyze their experience—do not require exposure to sophisticated quantitative techniques. Every West Point graduate must take two years of math and one course in statistics, but that’s often forgotten 20 years later. Officers who obtained their commissions through ROTC or OCS may have much weaker backgrounds in math.

And even as quantitative sophistication becomes more important in nearly every phase of life, West Point students—like American-born college students generally—are veering away from quantitative majors. (West Point provides about 20 percent of the officer corps, with ROTC providing another 40 percent, and the Army’s Officer Candidate School the remaining 40 percent.)

As Huddleston pointed out to me, the percentage of U.S. Military Academy graduates choosing to major in math, science or engineering has fallen almost steadily over the last ten years. The numbers are significant, given that the West Point class is around 1,000 a year: from 64 percent in 1999 to a low of 46.7 percent in 2007 and back up to 55.6 percent in 2009. (I had thought the explanation might be that during the Iraq war, more cadets chose to major in Arabic and other foreign languages. This is true, but, Major Joe Sowers, a public affairs officer at West Point, notes that the increase in Arabic majors (and more generally foreign language majors) is not sufficient to account for the decrease in quantitative majors.)

But as the efforts of JIEDDO show, we have a massive and culturally fascinating advantage over our current—and probably our future—enemies. It takes certain cultural habits to abstract from individual instances to generalizations, and to criticize and break with old patterns of thought to solve new problems. We may eventually defeat IEDs, and whatever replaces them, with the strengths of our culture, not our armaments.

OG Image: