A bouquet of flowers lay at the site of the plane crash that killed two children, ages seven and nine. Beside the flowers lay the note: "Mama loves you and always will."
"My kids were about that age at the time," reflected Dr. Barry Strauch, chief of the National Transportation Safety Board's Human Performance Division. Part of his job is to sift through the wreckage left by crashes like this one in Georgia, which killed not just the two unnamed children but also Senator John Tower of Texas in 1991, in hopes of piecing together the events leading up to the crashes to find out if human error was involved. The Georgia crash was caused by a defective propeller.
But figuring out who to blame when human error is involved isn't really the point of the investigations any more, not since the establishment in 1983 of the NTSB's Human Performance Division, he said,
"Before, it was often assumed that people made errors because they didn't follow procedures. Now the assumption is that people are always capable of error," he said in a presentation sponsored by the Program in Science, Technology and Society on February 2. His talk detailed how changes in the NTSB view of human error have encouraged investigators to focus on how and why errors occur, rather than just point the finger.
"Accidents are often caused by a whole bunch of little errors -- one right after another," said Dr. Strauch, a psychologist who specializes in the study of human factors in machine/human interactions. The cause of the errors can be personal, such as fatigue or stress, but more often they have to do with human-machine interface problems or task-related factors, such as heavy workload or poor skills.
Dr. Strauch noted that stressful life events, like divorce or the death of a family member, are rarely found to play a role in airline accidents. "It's very, very unusual to come across pilots in accidents who've been through these things recently," he said. "On occasion there's a skills problem. We'll find that a pilot had 'busted' a check ride."
He then moved from the theoretical to the actual to describe the series of pilot errors leading up to the December 20, 1995 crash of American Airlines flight 965 in Cali, Columbia. The Boeing 757, en route from Miami, crashed into a mountain just outside the Cali airport, killing all eight crew members and all but four of the 155 passengers. The airport lies in a valley shadowed by 12,000-foot peaks on either side. The air traffic controllers work without radar, which was destroyed by terrorists several years before the accident.
When the air traffic controllers offered the Flight 965 crew a last-minute change in runway that would shorten their flight time by a few minutes, the crew readily accepted. They entered the new landing plan into the electronic Flight Management System (FMS). But because of another, earlier input that led to the deletion of a critical navigation fix, the pilots essentially lost track of their position. Rather than reverting to a manual navigation check, one of the pilots punched in what he thought was the code for the original runway, but what turned out to be the code for a navigation fix near Bogotï¿½ï¿½ï¿½ï¿½ï¿½ï¿½ï¿½ï¿½ï¿½ï¿½ï¿½ï¿½ï¿½ï¿½ï¿½ï¿½ï¿½ï¿½. The plane immediately went into a turn to reverse course. It crashed into the mountain just one minute later.
The air traffic controllers might have helped the pilots understand their position, but "in Latin America it's difficult for a controller to question a pilot, even if the pilot was obviously confused. In the US it's very different," said Dr. Strauch. He added that "once the pilots' workload gets too great, they may have difficulty evaluating their options."
A mere five minutes elapsed between the shift in landing plans and the crash.
One result of the Flight 965 investigation was to change the on-line appearance of the navigation data provided by the FMS. Up to that point, the industry had failed to address the obvious differences between the appearance of the electronic information and printed charts.
"If you have a problem and find yourself really hustling to catch up, this discrepancy takes an extraordinary amount of time to sort out," he said.
John Carroll, professor of behavioral and policy sciences at the Sloan School, pointed out that while that change will be good, the focus is still on improving technology rather than human performance. He suggested that airlines might be wise to encourage more conservative decision-making in cases like these, much as the nuclear power industry has done.
"It could take the form of 'Don't change your landing plan in mountainous terrain unless you have X amount of time,'" he said.
The NTSB is an independent investigative branch reporting directly to Congress and the president. It cannot set policy nor regulate airlines, but can make recommendations. Dr. Strauch, who is a pilot by avocation, is certain that the recommendations made by the NTSB make a difference. He describes his work as "intense and dynamic. There aren't that many chances in this world to do things where you can say, 'I've done this and now things are better.'"
A version of this article appeared in MIT Tech Talk on February 11, 1998.