No one knows how many millions of land mines are buried in Iraq, one of the world's most heavily mined countries, but the U.S. Army is fielding the most advanced, highly engineered system yet developed for detecting them:
The U.S. soldier.
 |  |
Jim Staszewski, CMU cognitive psychologist, with a mockup of an M14 anti-personnel mine. (Lake Fong, Post-Gazette) | |
Some mine sweepers are equipped with a new handheld tool, introduced in just the past couple of months, that combines a traditional metal detector with a ground-penetrating radar. But the more significant advance in demining, a practice that must continue for years after the war ends, may be the revamped training that U.S. soldiers have received since last spring.
Devised by Jim Staszewski, a cognitive psychologist at Carnegie Mellon University, the training program teaches soldiers to use the thought patterns and techniques honed by an expert, a 30-year mine-detection veteran. It builds on the work of the late CMU scientists Herbert Simon and Allen Newell, pioneers in the study of the nature of human expertise.
Staszewski's approach, called cognitive engineering, has significantly boosted the ability of soldiers to find all types of mines, including hard-to-find plastic anti-personnel mines, even when using an older-generation detector.
"We were expecting to see maybe a doubling of the detection rate," recalled Alan Davison, chief of the Army Research Laboratory field element at the Maneuver Support Center at Fort Leonard Wood in southern Missouri. But some tests showed the techniques more than quadrupled the detection rate, convincing Army officials to scrap their old training methods in favor of Staszewski's.
Only a couple of years ago, the Army was despairing at soldiers' inability to find plastic mines, which began appearing two decades ago and contain only tiny amounts of metal. Tests showed that soldiers using a handheld metal detector called the AN/PSS-12 were missing anywhere from 70 to 96 percent of the low-metal mines.
"We just thought our equipment wasn't any good," Davison said, prompting development of the new metal detector/radar device, called the handheld standoff mine detection system, or HSTAMIDS. But it turned out that human factors were the biggest shortcoming. "We were teaching techniques that just weren't effective."
Using the new techniques, however, tests showed detection rates approaching 100 percent.
Measuring success
Just how the training might be working in Iraq is not yet known, said Maj. Charles May of the Army's Countermine/Counter Booby Trap Center at Fort Leonard Wood. Though the expert program has been widely disseminated within the Army in both publications and on a military Web site, it's not known how many of the soldiers in Iraq have had the formal training and how many have taught themselves the techniques.
Another complicating factor is that Iraq has a lot of laterite soil -- reddish and rich in iron -- that makes locating mines with metal detectors more difficult.
In Iraq, most of the land mines date to the 1980s war with Iran and are laid along the Iran-Iraq border. Many also have been used in northern Iraq, where Saddam Hussein's regime fought the Kurds, and near the southern border with Kuwait, where the Iraqi military planted mines during the first Gulf War.
An estimated 30 casualties from mines and unexploded ordnance occurred monthly in northern Iraq in 2001, according to Human Rights Watch, and at least 21 people were killed or injured elsewhere in Iraq that year. In the first Gulf War, land mines killed or injured 81 people, representing 6 percent of U.S. casualties.
In the current conflict, a land mine killed a BBC cameraman and mines have caused injuries and resulted in limb amputations for a number of soldiers.
Though Iraq's soil harbors a wide variety of mines, an alarming proportion of those encountered during Desert Storm were of the M14 type, the plastic anti-personnel mines that began appearing in the early 80s, Staszewski said. Since then, succeeding generations of the M14 have been made with less and less metal, making them increasingly hard to detect.
The AN/PSS-12 and most other mine detectors, however, were designed to find metal. The AN/PSS-12 is mounted on a long pole, much like the metal detectors used by optimistic fortune seekers prowling public beaches. The detector is held near the ground and, when it passes over something metallic, sends an audible signal.
Despite the military's growing unease through the 1990s with the device's poor ability to detect low-metal mines, it also was apparent that some mine sweepers working with humanitarian organizations were using it with great success, Davison said.
"These are people who had lifted thousands of mines and still had their hands and legs," he added.
What was it that made them experts? How do experts differ from everybody else? These were the questions being asked by the Army and the same sort of questions that mesmerized CMU's Newell and Simon.
The CMU researchers were popularly known for creating the first thinking machine, launching the field of artificial intelligence, in the mid-1950s. But this quest to make computers think like humans went hand-in-hand with their efforts to understand how humans think. As a result, the pair had as much impact on the field of psychology as on computer science.
"Herb Simon and Allan Newell pretty much got cognitive science off the ground," Staszewski said. "I'd like to think this work is a direct descendant of them" and other CMU psychologists, including the late William Chase.
How to be an expert
One of the things they learned is that you don't have to be a genius to be an expert.
In his 1991 book, "Models of My Life," Simon wrote: "Experts, human and computer, do much of their problem solving not by searching selectively, but simply by recognizing relevant cues in situations similar to those they have experienced before."
Staszewski, who has worked in the cognitive science field since 1980, tracked down an expert mine sweeper, Floyd "Rocky" Rockwell of Rio Oso, Calif., an Army vet who worked with a humanitarian demining organization. He collected 36 hours of video and audio tape of Rockwell in action, while prompting him to think out loud as much as possible.
"It sounds like something that would be really simple, but it's not," said Davison, who met Staszewski in 1999. "Jim is incredibly good. He gets into their hearts and minds and learns what makes them tick."
Staszewski spent six months analyzing his interviews and tapes of Rockwell, who died two years ago in auto accident.
"These guys are responding to a pattern, not just a sound," he concluded. "You're looking for something, rather than just reacting to a sound" from the detector. "They see patterns in their head, not on the ground."
The experts also have learned to alter their technique to compensate for the device's shortcomings. In searching for low-metal mines, for instance, the "sweep rate" of 1 meter per second advocated in the device's manual is too fast; one foot a second should be the upper limit.
Also, in practice, many soldiers sweep the head of the detector more than two inches above the ground. At more than two inches, however, a detector could sweep right over a low-metal mine without emitting a warning tone.
Out of this, Staszewski was able to distill some insights and design a series of training drills and techniques that would enable others to think and act like the expert. Working with Herman Herman, a robotic vision expert at the National Robotic Engineering Consortium, he devised a training course with a robotic video tracking system that would show every bit of ground that a soldier had swept with the detector and identify any gaps. He also devised simulated anti-personnel mines, consisting of a small bit of wire embedded in a hockey puck (direct from House of Hockey in Aspinwall).
A former high school football coach, Staszewski was adept at working with young soldiers. And his dedication was such that he did much of the work for free. "There were times we didn't have the means to pay him and he did it anyway," Davison said.
Tests in the fall of 1999 at Fort Leonard Wood showed that a group of conventionally trained soldiers could detect about 30 percent of low-metal mines, while those who underwent 15 hours of "expert" training were able to find almost 90 percent of them.
"We were shocked," Davison recalled.
Notably, he said, "everybody did well." Researchers had suspected that some soldiers would have a greater aptitude than others. But the tests suggested it didn't seem to make much difference.
Had the improvement been more modest, Army officials might have been quicker to adopt the methods, Davison suggested, somewhat facetiously. So they tried the experiment again a year later, at Aberdeen Proving Ground in Maryland. After an hour of standard Army training, soldiers were able to find 15 percent of the low-metal mines; after an hour of expert training, another group found 75 percent.
The Army still wasn't convinced. But the following summer, when the technique was used again at Fort Polk, Louisiana, the detection rates went up further. What's more, the detection rates went up when soldiers trained fellow soldiers, indicating the instructors weren't the deciding factor.
Finally persuaded, Army officials decided in February 2002 to use the new training system from then on. Though developed with the old detector, the same techniques have since been adapted to the more capable -- and complex -- HSTAMIDS and used at training centers established in Germany, Afghanistan and Kuwait.
Staszewski also has begun working with the Pentagon's humanitarian demining office, spending time last fall in Angola and, last month, in Azerbaijan to train indigenous people to operate automated demining machines.
"There is a realization that the human operator is an integral part of the system," Staszewski said. "Just as you want robotics, electrical engineering and mechanical engineering people working on the physical system, we have enough cognitive science to work on better engineering of the operators."
Byron Spice can be reached at bspice@post-gazette.com or 412-263-1578.
Correction/Clarification (Published April 15, 2003): The first name of the late Allen Newell, a pioneer in cognitive psychology at Carnegie Mellon University, was misspelled in the April 14, 2003 version of this story about improved mine sweeping techniques.
