Inscrutable Technology

Cockpit automation is a source of suspicion, even mistrust, among many pilots. Cockpit systems tend to be overly complicated and their functioning inscrutable. A technical memorandum on “Cognitive Engineering … of the Vertical Navigation (VNAV) Function”, a central component of the drive toward increasing automation, by a team of researchers found:

“The typical VNAV function automatically chooses the active altitude target from a possible list of sixteen, and chooses the active speed target from a possible list of twenty-six … [and] Pilots are ‘surprised’ by the behavior of the VNAV function when the aircraft trajectory or the thrust indicators do not match their expectations.”

The researchers concluded: “VNAV function behavior is not intuitive in the descent and approach phase of the flight plan.”

This problem was amply demonstrated in the July 2013 crash of Asiana Airlines Flight 214 at San Francisco. The National Transportation Safety Board (NTSB) concluded, “The crew over-relied on automated systems without fully understanding how they interacted.”

Thousands of hours of flying experience, hundreds of hours in classrooms and simulators, yet the crew was unaware that the engines were in idle thrust during the critical landing phase. Automation did not help to prevent this accident — rather, it contributed to the catastrophe.

During the course of its exhaustive investigation into the crash, then-NTSB Chairman Deborah Hersman observed that “mode confusion or mode awareness is an issue … Because it’s not just in this accident … It’s over and over again.”

She asked Stephen Boyd, of the FAA’s transport aircraft directorate, a basic question, “What’s the purpose of automation?”

Boyd: “Are you asking from a regulatory perspective?”

Hersman: “I’m asking for a one-sentence answer.”

Boyd: “Madam Chairman, it’s very difficult to answer because there is such a wide range of what automation means … it’s hard to say what the purpose is, and I would have to say it depends on what you’re automating.”

Hersman: “Okay. So what’s the purpose of autothrottles? Is it to reduce workload? Is it to increase safety? I mean, I’m just looking for a really big picture understanding of why you would implement that rather than manually fly all the time, be responsible for throttles all the time?”

Boyd: “Well, I would say that’s fundamentally a question for the designers to answer.”

Captain John Cashman, formerly Boeing’s chief pilot for the B777 program, did little better, saying the customer wanted automation to “improve the efficiency of aircraft.”

Unsaid in these unsatisfying exchanges is that automation has been increased because of the gains in calculating capability and the reduced size and power demands of computers. Automation has increased because the technology enabled it.


Cockpit automation has increased because technology enabled it, not because it met the pilots' needs

Cockpit automation has increased because technology enabled it,
not because it met the pilots’ needs

Hersman never did get an answer to her basic question.

Let us try a basic definition:

“Automation is applied to relieve the pilots’ routine workload while increasing their control of the airplane and their awareness of the airplane’s safe and unsafe location and path in three-dimensional airspace; automation calls to the crew’s attention any deviation from normal operating parameters with plain and compelling language, all while facilitating the pilots’ ‘heads-up’ view out the windscreen.”

Or, the short version:

“To increase pilots’ control of the aircraft and their awareness of systems functioning.”

In other words, attending to the task of entering values on a key pad, to the detriment of looking through the windscreen, would be a negative according to this definition. So would engines at idle thrust in the moments before touchdown during landing. So would attempted takeoff without deployment of trailing edge flaps. So would all fuel dribbling out through a leaky pipeline, leading to an unusually low fuel state too soon in the flight. There is a long list of problems where existing automation has failed — either designers did not account for a seemingly improbable failure (technological hubris), or explicit alerts to the crew were not provided (design complacency).

Under the definition above, sixteen altitude targets, twenty-six speed targets, and the computer selecting between this array — to the frequent mystification of pilots — would be impermissible. The automatic selection of target altitudes and speed is not in accordance with strengthening the crew’s control of the airplane.

Increasing the automation because it was possible to do so has only served to increase pilot surprise. According to the technical memorandum referred to at the outset of this discussion, more than half of the pilots surveyed reported being surprised by VNAV behavior (e.g., unexplained altitude errors, unpredictable speed targets during approach). One group of pilots reported the VNAV function as “the most disliked feature of automated cockpit systems.”

Hersman asked the salient question — one that neither the FAA nor designers of modern jetliners have asked themselves. Yet inscrutable technology, its secrets buried in lines of computer code, is piled on top of existing protocols that are, at best, only partially understood by pilots.

“A Cognitive Engineering Analysis of the Vertical Naviation (VNAV) Function” by Lance Sherry et. al., National Aeronautics and Space Administration, Ames Research Center, TM-2001-210915