Automation Paradox

Sunday, September 8th, 2013

Aircrew call it the automation paradox:

A growing body of evidence indicates that while cockpit automation may be relieving pilots of mundane chores when their workload is actually low (ie, while climbing to altitude and cruising), it is causing bigger headaches than ever when the workload is particularly high (ie, during take-off, descent, approach and landing).

This has taken its toll on their “stick-and-rudder” skills:

Instead of flying their planes, flight crew now spend most of their time in the air programming and monitoring various pieces of equipment (a typical airliner has around 90 automated systems on board), inputting data and checking that everything is working correctly.

Many of today’s younger pilots (especially in the rapidly expanding markets of Asia and the Middle East) have had little opportunity to hone their airmanship in air forces, general aviation or local flying clubs, allowing them to amass long hours of hand-flying various aircraft in all sorts of weather conditions and emergencies. As a result, the tendency among pilots today is to trust a plane’s automation more than their own skill and judgment.

I’ve mentioned Korean pilots before.

Masterful pilots are slowly disappearing:

Pilots with Chesley “Sully” Sullenberger III’s background are becoming the exception rather than the rule. Captain Sullenberger, readers will recall, carefully ditched his Airbus A320 close to a jetty in the Hudson River in January 2009 without loss of a single life, after the plane had been disabled by a flock of geese while climbing out of LaGuardia airport. It should be pointed out that Sully learned to fly at the age of 16, flew F-4 Phantoms in the air force, and had 40 years and 20,000 hours of mostly hands-on experience when he performed his heroics on the Hudson.

The problem today is that aircrew may log thousands of hours on the flight decks of modern airliners, but their actual hands-on flying experience may amount to mere minutes per flight. When things get frantic — whether through a mistaken input or a sudden runway change by air-traffic control during descent — aircrew can be so preoccupied punching fresh instructions into the flight-management computer that they may fail to notice their airspeed and altitude are falling precipitously.

This reduction in situation awareness, along with the degradation of basic piloting skills and a huge increase in cognitive workload on flight crew are all part of the unintended consequences of cockpit automation. Combined, such human factors can quickly lead to disaster.

America’s two recent fatal air crashes — the Asiana Boeing 777 passenger jet on final approach into San Francisco international airport on July 6th and the United Parcel Service Airbus A300 freighter coming into land at Birmingham airport in Alabama on August 14th — are cases in point. Though investigations have barely begun, both situations point to distractions the pilots faced while trying to take control of the aircraft. In both instances, the pilots seem to have been unaware, until the last few minutes, of their proximity to the ground and of how slowly their planes were flying. Both finished up crashing short of the runway.

In both instances, federal investigators have found little evidence of equipment failure before the crash.

(Hat tip to Jonathan Jeckell.)


  1. David Foster says:

    This airline pilot suggests that up to 50% of the 1500 hours now required for airline pilots should be allowed to be accomplished in gliders.

  2. Isegoria says:

    What we want are skilled pilots, and we shouldn’t care exactly how they got skilled, but we don’t have a convenient way to measure piloting skill (apparently).

    So, instead of measuring output, we measure input — but not all flying hours are the same. Really, it’s surprising that flying on autopilot counts at all. Anyway, if the old metric is 1500 hours of “flying”, perhaps 500 hours of really flying — taking off, landing, otherwise maneuvering — should qualify.

    Allowing some hours outside a jet-liner to count seems perfectly reasonable.

  3. David Foster says:

    There are indeed measurements of piloting skill — for example, confront the individual with a series of problems, either in a real airplane or a simulator, and see how he does — but they are imperfect.

    I think one issue is basic airmanship skills. Another is deep understanding of the systems, and this second point relates to the way the systems are designed, from a user-interface standpoint, in the first place. In his book Taming Hal, which I’ll be talking about in a future post, Asaf Degani reviews a whole set up disturbing situations — in aviation, in medical devices, in sea transportation — in which what the crew thought the control system had been programmed to do and what they had actually programmed it to do were quite different.

  4. Isegoria says:

    I’m reminded of The Logic of Failure.

Leave a Reply