The Autothrottle, the Asiana Two Fourteen Approach at San Francisco, and What Mode Confusion Tells Us About the Future of Cockpit Automation

On July 6, 2013, Asiana Airlines Flight 214, a Boeing 777 arriving from Seoul, struck the seawall short of Runway 28 Left at San Francisco International Airport in clear, unlimited-visibility conditions. The aircraft was airworthy. The autothrottle worked exactly as designed. Three people died and 187 were injured because the crew’s mental model of what the automation was doing did not match what it was actually doing.

What the Flight Data Recorder Revealed

Flight 214 carried 291 passengers and 16 crew members on a visual approach to SFO. The ILS glideslope for Runway 28 Left was out of service that day - a valid NOTAM the crew had received. They flew the approach using the Precision Approach Path Indicator (PAPI) lights, which is completely standard procedure at major airports worldwide.

During the descent, a sequence of autopilot mode changes caused the autothrottle to enter hold mode. In hold mode, the thrust levers freeze at their current position. The system is technically still active - it is simply not doing anything.

Speed began to decay. At 500 feet above the ground, the aircraft was already below its target approach speed with no automatic correction underway. At 200 feet, the crew recognized the problem and called for thrust. The GE90-series high-bypass turbofan engines require three to five seconds to spool up to meaningful power output. In a decelerating aircraft already below approach speed at 200 feet, three seconds is not recoverable.

At 112 feet, the stick shaker activated - the mechanical stall warning. The tail struck the seawall at 106 mph. The main landing gear caught the embankment. The aircraft rotated, skidded to rest on the runway, with both engines separated from the wings and the tail section largely destroyed.

The NTSB published its final report in June 2014 (report number AAR-14-01), describing it as one of the most comprehensive investigations into automation-related accidents the agency has ever conducted.

How Autothrottle Modes Work

An autothrottle is a motorized system connected to the thrust levers that moves them automatically to maintain either a target airspeed or a target thrust setting. The concept dates to the 1960s, when Boeing and Douglas introduced production systems on early jet airliners. On a modern transport-category aircraft like the 777, the autothrottle integrates with the autopilot, flight management system, thrust management computer, and envelope protection software - and it has a substantial number of distinct operating modes.

Speed mode actively maintains a target airspeed. Speed drops, power is added. Speed climbs, power is reduced. This is the mode most pilots visualize when they picture autothrottle engaged.

Thrust mode (N1 mode on Boeing aircraft) maintains a specific thrust setting, not a specific airspeed. The engines hold a fixed output. If the aircraft slows, the autothrottle does nothing about it - airspeed management becomes the pilot’s responsibility.

Hold mode freezes the thrust levers at their current position. The autothrottle is still technically engaged but is neither adding nor reducing power, and it will not respond to airspeed changes.

Retard mode is specific to the landing flare. At a defined radio altitude, the autothrottle commands the levers to idle to ensure consistent, clean touchdowns.

On a normal approach, this sequence is seamless: speed mode all the way down the glidepath, retard mode in the flare. The problem occurs when the system transitions to a different mode mid-approach in a way that isn’t apparent to the crew.

The System Didn’t Fail - The Interface Did

This is the central finding of the investigation: the Boeing 777’s autothrottle operated exactly as designed. Every mode transition fell within the aircraft’s certified operating envelope. There was no mechanical failure of any kind.

The failure was cognitive. The NTSB cited crew monitoring failures and inadequate cross-checking as contributing causes. The captain was relatively new to the 777, building hours under the supervision of a check airman whose intervention came too late. The report is thorough and fair in how it distributes contributing factors.

But the underlying interface reality cannot be argued away. On Boeing aircraft, the autothrottle physically moves the thrust levers when in active mode - you can see them move, feel them move if your hand is resting on them. That is real, useful feedback. In hold mode, however, the levers don’t move. They sit frozen at their current position. A frozen lever in hold mode looks physically identical to a frozen lever in speed mode correctly holding target speed. The lever position tells you nothing about whether the system is working or stopped.

Human factors researchers at NASA and the FAA have been documenting this phenomenon - called automation surprise - since the early 1980s. Four decades of research later, a reliable design solution remains elusive.

Boeing vs. Airbus: Two Philosophies, One Unsolved Problem

Airbus took a fundamentally different approach. On Airbus autothrust systems - found in the A320 family and all later models - the thrust levers do not normally back-drive. They sit in fixed detent positions, and the flight guidance computers manage thrust within the approved envelope. Mode status is displayed prominently on the Flight Mode Annunciator (FMA) on the primary flight display.

The Boeing approach provides physical feedback that disappears in exactly the modes where it would be most useful. The Airbus approach places mode awareness entirely on reading a display. Neither philosophy has eliminated the problem. Both Boeing and Airbus systems have been involved in automation-related accidents. This is a genuinely hard design problem the aviation industry has not solved.

NASA’s Flight Deck Human Factors Research Program has documented automation complacency as a predictable cognitive response: when a system reliably works, the brain reallocates attention away from monitoring it. That is efficient human cognition. It is also the mechanism exploited by the unusual case - the mode transition the pilot didn’t catch.

Why This Matters for Pilots Flying GA Today

Autothrottle is no longer exclusive to the airlines, and this is the part every non-airline pilot flying automation needs to understand. Innovative Solutions and Support (ISSS) achieved FAA certification for a retrofit autothrottle for the Beechcraft King Air called ThrustSense - the first certified retrofit autothrottle for a production turboprop. It manages engine torque automatically through climb, cruise, and descent, and the King Air community has embraced it for legitimate reasons: turboprop torque management in a twin is demanding, and automation frees meaningful mental bandwidth in single-pilot operations.

Garmin is actively developing autothrottle integration for the G3000 and G5000 avionics suites. The trajectory is clear: within this decade, autothrottle will be a standard option in high-performance piston and turboprop aircraft.

What cannot be assumed is that mode confusion is an airline problem. The cognitive challenge is identical regardless of aircraft size. Engine spoolup physics are identical. The gap between what a pilot thinks the automation is doing and what it is actually doing does not respect aircraft category.

What the Industry Has Done - and What Remains Unsolved

FAA Advisory Circular AC 120-71B addresses automation mode awareness directly, mandating pilot monitoring callouts, mode awareness checks, and cross-verification procedures. Airlines that implement it rigorously show improved outcomes. It is sound policy.

Garmin’s newer avionics feature more prominent mode displays than the Boeing glass cockpit of 2013. That is genuine progress. But clearer annunciation only addresses the moments when a pilot is looking at the display - not the moments when workload is high and attention is elsewhere.

The overall accident rate in commercial aviation has dropped dramatically across decades of increasing automation. The technology works, and the data supports it as a safety tool. But the accidents that still occur in highly automated environments consistently share a common thread: not mechanical failure, but a mismatch between crew expectations and system state. That is simultaneously a training problem and a design problem.

The long-term answer requires automation that makes its own state changes legible to the humans sharing the cockpit - better tactile feedback, clearer mode annunciation, and interfaces designed with the assumption that the human operator is occupied, not idle, at the exact moment a mode change occurs. That work is active and ongoing.

Key Takeaways

• The Asiana Flight 214 crash on July 6, 2013 killed 3 and injured 187 in clear VFR conditions - caused by mode confusion, not a mechanical failure
• Autothrottle hold mode freezes thrust levers with no visual distinction from active speed mode on Boeing aircraft, making undetected transitions possible
• Engine spoolup lag of 3–5 seconds on high-bypass turbofans makes a late power call non-survivable at low altitude in a decelerating aircraft
• Autothrottle is already in GA turboprops via ThrustSense and is expanding into Garmin G3000/G5000 avionics - mode awareness is now a relevant skill outside the airlines
• FAA AC 120-71B has improved training standards, but the underlying interface design challenge remains genuinely unsolved across both Boeing and Airbus philosophies