The thread title is from the video “Children of the Magenta Line”. The lowest level of automation is instruments and visual.
Full Transcript
Okay, automation dependency—as I mentioned earlier today—the reason this segment is in this course is because, as we look at accident history, we find that in 68% of these accidents, automation dependency plays a significant part in leading these crews to either a critical flight attitude or the requirement to extract maximum performance from their planes.
Automation-dependent pilots allowed their airplanes to get much closer to the edge of the envelope than they should have. As we started to study this issue, we decided to take a new tack. So, you’re going to hear a lot more at American Airlines about what we’re going to call levels of automation and technology judgment.
What are levels of automation? Well, the academic types will tell you there are seven levels of automation or something like that. It gives me a headache. I’m a simple guy. I’m going to suggest to you that there are basically three levels of automation in our planes.
The lowest level, you might call manual, is when you’re hand-flying the airplane—the throttles and the controls—and by the inputs and cues you’re getting from your instrumentation and outside visual cues, you’re determining the vertical and lateral path of the plane by flying it.
The next level is when you have the autopilot engaged, and you’re using the flight guidance system to tell the autopilot what to do for short periods of time—flight level change, vertical speed, altitude hold, heading select.
The highest level is when the autopilot is engaged and you’re using the flight management computer (FMC) to tell the autopilot what to do for hours and hours.
So: low, medium, high.
Now the question becomes: which is the appropriate level for the task at hand?
You could say all this automation was put in two-crew airplanes to reduce workload, and that going up a level reduces workload. That would be true in many scenarios—like crossing the North Atlantic.
But you could also say that going down a level can reduce workload in certain scenarios.
What we find in the accident history is that in the vast majority of cases, the crew has lost situational awareness. How did that happen? They became task-saturated. And in most of those accidents, they became task-saturated because they were trying to operate at too high a level of automation in a rapidly changing flight path requirement.
What they needed to do was drop down a level in order to maintain situational awareness and reduce workload.
As an example, suppose you’re coming in on a STAR in an FMC airplane. You’ve probably had that STAR loaded in the computer for an hour, everything going as planned, you’re in LNAV and VNAV and stroking down the arrival.
But then, around 20,000 feet, ATC calls and says, “American so-and-so, change your arrival, turn left heading X, descend and maintain Y, intercept radial Z for the new arrival.”
At this point, is it time to go up a level or drop down a level?
It’s time to drop down a level.
Why? Because he said: turn, descend, intercept. That means: heading select, flight level change. Get the airplane going where he asked, and your head’s up. Then he said intercept a radial: tune it, identify it, present it. You must present raw data per American Airlines procedures. That’s required work.
Later, if you have time, you can type it all into the FMC and hook it up—but that should not be your initial move. Why? Because:
- The airplane isn’t going where you want yet.
- One person typing means the other must monitor. And both must verify the new flight path is correct. FMCs are not error-resistant.
You can input something wrong, and it will do it. You can input something right, and it might do something wrong due to a database anomaly. It takes two people to verify it—and that takes time. That path-saturates the crew. Drop down a level. Reduce workload. Stay safe.
Another example: Suppose you’re flying into Dallas on the Bowie Arrival on a perfect, clear day. You’ve had it programmed since Colorado Springs. Everything’s synced up and going great.
Then ATC says, “Change your runway to 13R.”
You’re 26 miles out. You see the airport. You can see the pavement. What do you do?
From highest level—LNAV/VNAV—you go straight to the lowest: hand-flying. You line yourself up with the pavement.
Meanwhile, your copilot is trying to help. You’re dealing with traffic, spacing, maintaining visual awareness. Is it a good time to be typing?
No.
Let me ask: what could the computer possibly bring you that you don’t already have? We’ve become, as I call it, children of the magenta. We think we need the magenta line and magenta V-bar, or the plane won’t fly.
Another case: A Fokker aircraft was cleared for 17L, then told to sidestep to 17R. Visual conditions, 10–12 miles out. They saw the airport, but instead of flying it visually, they tried to reprogram—and ended up landing without a clearance.
You’ve got to pick the appropriate level of automation for the task at hand.
To that end, we need to change the culture that tells us to operate at the highest level at all times. We created that culture. The industry did. But it needs to change.
That’s why we have a new course: Human Factors and Safety Training. It will focus on levels of automation and technology judgment.
The top bullet on the slide says: “Be the ball.” That means: stay connected. Don’t disengage just because the autopilot is on. Especially in patterns, configuration changes, and maneuvering—remain mentally and tactilely connected to the aircraft.
Quick example: Bucharest. An A310 with an autothrottle malfunction, nose-high attitude, no visible horizon. One throttle jammed, the other went to idle. The airplane banked over and rolled nearly inverted. If the copilot had his hand on the throttles, 128 people would still be alive.
Autopilots have limitations. Worse, automation-dependent crews often lack confidence in their ability to fly. So they turn to the autopilot in deteriorating situations—and instead of saving them, it kills them.
We’ve seen it in mode confusion, terrain warnings, upsets, and even mid-air conflicts.
Let me tell a story about a 757/767 sim ride. I had a captain and FO, very sharp, nailed their check rides. On the last day, I gave the captain a scenario from Orange County to San Jose, with an MD-80 sim target on a collision course at 5,000 feet.
As the captain coasted toward the beach, he hooked up his autopilot. Then, seeing the MD-80: “Give me vertical speed. Give me flight level change.”
He was trying to avoid a mid-air with the autopilot. That’s not what we trained him for. But it was understandable—he’d been typing since day one.
What should he have done?
Click-click, hand-fly. The pilot with the controls and throttles will do what it takes not to hit a plane he can see.
The autopilot cannot leave altitude in less than 0.23 NM. It’s in the algorithms. It doesn’t understand urgency.
Disconnect. Hold the controls firmly. Test pilot Dale Rantz says: the autopilot may have started correcting—don’t disconnect with a weak grip and have the yoke jump out of your hands. Start there and take control.
The red box on the slide—this one isn’t in your manual. This is from us to you. We must return to the principle of fly the plane first.
We used to say it. We used to do it. But automation crept in. We were told to become “automation managers.” But the accident history says automation managers are plugging themselves into the ground all over this planet.
We are not automation managers.
We are captains and pilots. We will use automation as a tool, but when it doesn’t fly the path we need, we will turn it off and fly.
Example: If you’re on a non-precision approach, in vertical speed, and the plane will descend through MDA—do you push “altitude hold”?
No. You disconnect. You can stop it. The autopilot can’t.
What’s the most asked question in the 757/767 cockpit?
“What’s it doing now?”
If that’s asked in a low-altitude environment, the pilot flying should disconnect. Maintain the intended flight path. Stabilize. Then ask why.
Maintain your flying skills.
If you turn off the autopilot but leave the autothrottles on, are you really flying?
No—you’re just guiding the airplane to the V-bars.
To maintain piloting skills, we must both manipulate the controls and manage the energy. That requires a full cross-check and engagement.
Cecil won’t order you to do this. He’s a leader. But he is saying: as often as you can, when appropriate, practice. Turn off the autopilot and autothrottles. Fly your airplane. Maintain your skills.
So that when something happens, you’ll have the proficiency and confidence to take over and fly any situation successfully.
We haven’t changed procedures. What we are saying is:
When you intend to hand fly to maintain skills, turn off both the autopilot and the autothrottles.
We are pilots. Captains. But also cockpit managers.
We must manage automation—but not be ruled by it.
Increasing automation reduces workload most of the time.
But in some cases, it increases task saturation, reduces situational awareness, and leads to accidents.
Automation cannot generate flexible responses to unanticipated changes.
In those moments—drop down a level.
And fly the plane.
One scenario in the video is given a late change by ATC (Air Traffic Control) of the assigned runway. The question is should the crew reprogram the FMC (Flight Management Computer) or disengage the auto pilot “And fly the plane.” (using “instruments and visual”)
Not perfectly analogous to marine navigation but a similar situation might be approaching the pilot station when the pilot cancels and the ship told to anchor.
Should the bridge team reprogram the ECDIS or just do a round turn and navigate to the anchorage using the instruments and visual?
It doesn’t make sense to ignore the ECDIS, radar and other instruments altogether. It’s about workload, not being some kind of purist.
