Entirely speculating, but it sounds like the fundamental problem is that they had a confused Captain and First Officer who both thought they were the Pilot Flying. Human factors / psychology issue. Sure, on a Boeing you'll feel the other pilot's input, but if you're tired enough, or panicking because the plane's not doing what you expected, you might not get the hint.

Task saturation - followed by loss of situational awareness - is a very real problem in today's airliners, particularly during takeoff and approach/landing. You're trying to do a ton of things in a small amount of time, while dealing with a dynamic situation outside, and in this case after a long, tiring flight over the Atlantic.

There are a great series of (dated but still good) videos from American Airlines about automation dependency, human factors and the simple but very real problems that develop as a result - often leading to panic and disaster when things go wrong. Search YT for "children of the magenta".

In the AF447 case, the lack of feedback/linkage of the Pilot Flying / Nonflying sidesticks was a big contributing factor. It is true that the Pitot tubes froze up. But the Pilot was roused from his sleep, entered the cockpit, and correctly pointed out that the plane was actually in a dive. It's just that the guys flying were making opposite inputs over an extended period, and it was too late to recover, which doomed AF447. Overall, Airbus and its delinked sidesticks was a factor in the accident.

I thought Boeing did not have sidesticks, and the yokes/control columns are linked to give actual feedback (to emulate most airplanes, which do this with mechanical controls). Is that not the case?

The transcripts of AF447 include the pilot telling the right seater to stop. More than once. But he keeps doing inputs.

Airbus 320 pilot told me there is a light that comes on when inputs dont match.

Now reread the transcripts with the assumption that the right seater might not have the best intentions.

What do you mean by best intentions? There are more efficient ways to crash a plane.

But also isn't the right seater the first officier, surely he should obey a repeated direct order from the captain.

The captain was asleep. Maybe there were no clear consensus on chain of command in the cockpit.

By convention the left seater is the captain. In the case of AF447, the captain was asleep, and the FO and a relief pilot were flying the plane.

There were lots of things that went wrong with AF 447 but the side sticks were a minor detail. If anything the stall warning going on and off was a much bigger factor because the one pilot kept pulling back on the stick which is the exact opposite of what he should've done. Had the pilot flying simply flown pitch and power they would've been fine and had plenty of time to calmly troubleshoot the situation.

> I thought Boeing did not have sidesticks, and the

> yokes/control columns are linked to give actual feedback

> (to emulate most airplanes, which do this with mechanical

> controls). Is that not the case?

Boeing's philosophy seems to be to provide physical feedback for most or all of the controls. So, yeah, the control columns are linked (even on fly-by-wire planes like the 777 and 787). But that's not very useful once you're dealing with startled and/or panicked pilots. Take a look at the Atlas Air or Sriwijaya Air crashes.

The problem with Air France has nothing to do with Boeing or Airbus and everything to do with pilot standards and training. At this point Air France is lucky. Look at the Air France incident in 2020 (A318). Or their incident in Bogota in 2017 (A340). Or the 2011, 2012 x2, and 2013 incidents where the Airbus (A320/321) alpha floor protection prevented an Air France crew from crashing on approach to Marseille x2, Hamburg, and Tel Aviv. At some point Air France's luck will run out.

> If anything the stall warning going on and off was a much bigger factor because the one pilot kept pulling back on the stick which is the exact opposite of what he should've done.

Private pilot here.

This is so hard. It’s not an excuse, especially for professional pilots.

But I can share from experience how much your “animal brain” fights you in this situation. Pulling back leverages lift in most normal flight, until you go too slow (stall), and then you suddenly very much need to do the opposite. But the normal behavior becomes second nature, like walking.

My examiner in 2020 said the biggest thing the FAA was stressing about was a serious uptick in incidents with slow flight (and stall) handling during takeoff/landings.

A year ago, we had an experienced local private pilot wreck his plane (lucky no major injuries) succumbing to exactly this. Took off, lift conditions a little worse than normal, started to leave ground effect, felt the plane stall, pulled back, started losing altitude, panicked and pulled back harder, tailed down and fell side wards. Plane totaled. He recounted to me after how he knew what to do. He’s trained multiple times. Done lots of stall recovery. Has done many many short field takeoffs (which is similar, pop the plane off the ground and then stay low to build speed). But was still chagrined at how panick/situation management got the better of him that time.

Being master of your animal brain in all situations is hard. And you never know, because you can never test all situations.

Sure. There are some things in aviation that are counterintuitive: recovering from a stall is one of those. That's why commercial aviation in the United States places such a strong emphasis on regular training (and why commercial aviation in the US is as safe as it is). Unfortunately stall physics are something that flight simulators don't tend to do so well at.

Another counterintuitive thing is that tactile feedback isn't necessarily helpful. There's a lot of whinging about how Airbus doesn't provide any feedback in the sidestick when both pilots are doing conflicting things, same with the thrust levers (or whatever they're called in Airbus land). Pilots have crashed Boeing airliners that give you this "important" feedback.

Anyways, my point was that stall recovery is something that's usually drilled into pilots. Unfortunately for the AF 447 crew, nobody told them that the stall warning would turn off when they flew so far outside the envelope that they were in "undefined behavior" territory. And that led to at least one of the pilots not recognizing that they'd stalled in the first place. But like any crash there are many contributing factors, and if any one of those had been different the plane wouldn't have crashed.

If it were just one or two examples of poor airmanship at Air France then sure, it'd be easy to chalk it up to "flying is hard". And yeah, the pandemic has robbed a lot of pilots of their necessary practice. But the four or five cases I cited where Airbus' "alpha floor protection" saved Air France flights were all pre-pandemic. At some point you've gotta look at what those incidents and accidents have in common (hint: it's not the airplane and not the crew). And that's why my main concern is with Air France's lack of safety culture and by extension their pilot training.

> commercial aviation in the United States places such a strong emphasis on regular training

This in a nutshell is why I never completed my Private Pilot training. As I got close to taking my checkride I began to think about finances going forward. And I realized that I probably couldn't afford to fly often enough to keep my skills sharp and a dull pilot is a dangerous pilot.

25 years later I'm still alive so I think I made the right decision :-)

> Overall, Airbus and its delinked sidesticks was a factor in the accident.

Confused, Bonin exclaimed, "[Expletive] I don't have control of the airplane any more now", and two seconds later, "I don't have control of the airplane at all!"[29]

Robert responded to this by saying, "controls to the left", and took over control of the aircraft.[73][32]

He pushed his side-stick forward to lower the nose and recover from the stall; however, Bonin was still pulling his side-stick back.

The inputs cancelled each other out and triggered an audible "dual input" warning.

--

So in an Airbus this situation is unambiguous - the plane tells you plain and simple: Both pilots are trying to fly at the same time. With a mechanically linked set of controls, you don't necessarily know whether you're fighting the other pilot or the plane.

That does seem to be what was happening here, as both pilots thought the plane was fighting their inputs, but in reality they were fighting each other.

> my airplane

> your airplane [lets go controls]

Those are the words to be exchanged to avoid confusion.

Sure, just make sure that happens exactly that way 1,000,000 times without any failures…

And that’s the root of many issues in aviation. It works 99.99% of the time is not always good enough.

I wonder if this was part of some rare coincidence where their inputs were perfectly synchronized but out of phase in such a way that it made it difficult to consider the obvious. If you pull back but feel the plane and/or stick move laterally, and especially across multiple inputs without a clear pattern, it might more easily occur to you to think the person sitting next to you is also providing inputs. Whereas if you're synchronized across many inputs, that's just something that doesn't typically happen in nature, which tends to always have some stochastic quality; instinctively it's something you might assume would only involve a machine.

"I have the airplane" or "my airplane" is an established practice, but over and above that, the stick/yoke/rudder pedals all actually give you feedback when the other guy is also handling them. That feedback gives the fliers another level of protection.

Agreed, but also given they initially thought the aircraft itself was deviating from a stable approach (or at least the ATC transcription appears to suggest that?), perhaps the pilots may have thought the feedback was due to autopilot and not the person sitting next to them.

Yeah, simple enough. Also simple enough to go-around when you're obviously slow and going to be short of the runway, but look at Asiana 214.

I do wonder if there's some training deficiency or cultural issue that leads to this scenario in Air France flights.

This makes me think of Loic Suberville's videos on language...

> my airplane

> your airplane

> my airplane

Well this is terrifying. Their only job is to pay attention and communicate clearly…especially during takeoff and landings

Most people are not good at their jobs...

Most people don't get to fly airplanes.

As pointed out by a sibling comment, cockpit resource management training solves this very simply by verbally declaring who should have control.

As usual, Star Trek was out in front with CRM on how to verbally switch commands:

1. I'm losing command

2. I've got command

https://www.youtube.com/watch?v=RAvRBDQqSmY

> I thought that Airbus makes the only airliners with no feedback when there are dual inputs applied

This factoid will not die on HN. Conflicting inputs on an Airbus are handled by averaging and an audible "DUAL INPUT" cockpit warning.[1]

In the case of AF447, the co-pilot used his "Priority Takeover" button which meant only his control inputs were in effect.

If you wanted to say there is no physical feedback, then you would be correct.

[1] https://safetyfirst.airbus.com/app/themes/mh_newsdesk/docume...

Yes, that is what I meant. And any other type of feedback would be likelier to be ignored. That's why there's a stick shaker in addition to auditory and other warnings when the plane stalls, even in fly-by-wire planes.

AF447 was (I read the WP article on it) at 38,000 feet and ice seems to be implicated and quite a few other things.

This incident was at 1670 ft. My reading of the article is a P and CP being twats:

"During the go-around on the Air France flight, the captain was making nose-down inputs, while the first officer was pulling back on the controls."

I'm just a simple civvy but I was an air cadet and learned enough about flying in a single engine trainer to know that only one person should have control. You literally say "I have control" - "You have control" - "I have control". These muppets didn't seem to do that:

"It took until 07:52:06, for the captain to become the only person making control inputs."

In the very guarded language of accident review I would say that is a damning indictment of the behaviour of the pilot and co-pilot as being completely reckless and bloody stupid.

That’s all correct, but AF447 was exacerbated by the pilot and copilot (actually 2 copilots, IIRC, the senior pilot was on break) fighting each others controls as the plane flew towards the ocean. The pitot tube was giving bad readings (frozen over), but from what investigators can tell, one control was trying to pull up and even out and the other was still pitching down to gain speed (per the warnings given by the faulty pitot tube).

The tragic thing about that is all they had to do was simply let go of the controls completely and the plane would've been fine. The transcript of that flight still gives me chills: https://tailstrike.com/database/01-june-2009-air-france-447/

For anyone who hasn't read it: Imagine waking up, and a few moments later saying "I can't believe this is happening" right before you reach altitude zero.

That transcript is so sad.

A car, plane or whatever should only have one driver/pilot. Planes have dual controls by default but this sort of nonsense should not be allowed to happen.

This was not AF447 so what on earth happened? I'm guessing inadequacy of some sort and that's on the industry as a whole.

In cases like this it's almost always some variation of normalization of deviance.

If company culture doesn't strictly enforce callouts 100% of the time, it's easy to get into a routine of everything being just fine 100's of times until it isn't.

Really these sorts of things show there should be a failsafe mechanism that provides tactile/haptic feedback to the pilots if the other inputs are "fighting" theirs.

Even an audible alarm isn't enough, as you'll likely get more than a few of those during events where it matters.

Proper CRM of course should make this "impossible" - but in high stress moments any more layers of swiss cheese you can add the better.

Agreed. There should be a stick-shaker or similar to let pilots know the other control is being used. Maybe not the existing stick-shaker, since that's usually for stalls, but something similar.

I have no idea why the pilot-monitoring would have taken his controls. Or why the pilot-flying wouldn't have called him out for it. Possibly a seniority issue - we've seen similar issues before. Hopefully the french aviation authorities can determine the root cause.

Ever see those really long fire trucks?

Two drivers

But the second driver of a tiller truck only steers the rear wheels, more like the pilot of a glider being towed.

> In the very guarded language of accident review I would say that is a damning indictment of the behaviour of the pilot and co-pilot as being completely reckless and bloody stupid.

Well, yes, but these kind of incidents keep happening at AF. Everyone talks about AF447 because it ended in the ocean, but there have been many incidents with AF since AF447, involving serious pilot error which nonetheless ended in a safe landing. So you have to look a bit beyond the "muppets" in the front seats and start to wonder about company culture and training.

I've repeatedly heard stated that after a sufficient amount of force is applied to controls, "the link" between the controls is severed or deactivated, without any details regarding how this is done or what the actual effect is.

Here's an article stating exactly that: https://english.alarabiya.net/News/world/2022/04/28/Pilots-p...

> If opposing forces on the two columns pass a certain limit, the link between them is deactivated or “desynchronized” to prevent accidents in the case of one side getting jammed.

It's a purely mechanical linkage. I've been trying to read the patent (US 5782436A). The USPTO has a new single-page web app. It's awful for finding by patent number, and worse when you want a full screen image of a drawing. It's clear that if one yoke is jammed and the other is moved with enough force, a spring loaded pin is forced out of a notch and the connection between the two yokes releases.

What's not clear is how the linkage still connects to the flight controls. The idea seems to be that the one which is moving wins. What is clear is that, if your side becomes disconnected, your yoke doesn't do anything. No wonder one pilot was frantic. It seems that you get reconnected by moving the disconnected yoke to agree with the one in control. Then the pin falls into the notch and the yokes are connected again. Not sure about that. You can see the pin and notch in the patent drawings.

This is independent of the fly-by-wire system. That's downstream from these mechanical devices.

I wonder if pilot training includes knowledge that yoke disconnection is possible.

They are trained in this. If disconnected, each yoke will control "their" half of the plane. This is mostly intended to maintain some level of controllability if one of the yokes were to get blocked or seized mechanically, this is not intended for human issues

> I wonder if pilot training includes knowledge that yoke disconnection is possible.

Probably, I've heard https://www.youtube.com/c/MentourPilotaviation talk about it as well.

If the linkage breaks, then each pilot gets to control the control surfaces on one side of the aircraft.

In the 777 it's a direct mechanical linkage between the controls, with a torque-driven breakout which temporarily severs the link if there's too much difference in the two pilots' inputs

That still leaves the question of how to interpret opposing inputs for the physical control surfaces, though

On the 757, the two control columns are linked with a roller bearing in a detent (notch). With enough torque applied, the bearing will come out of the notch and the columns can move independently.

Does sound like 447, and a result of the airbus’ dual stick system. Boeing’s yoke system is connected between the pilot and copilot, so much easier to detect opposite inputs. Also the yoke is right in front of each pilot, instead of on the opposite sides as is the case with the dual sticks in airbus planes, which could be a source of confusion/opposite input problems.

the plane involved in a Boeing 777. That's right in the title of the article.

Nowadays planes will yank on the controls on their own initiative, most destructively e.g. in the case of a certain pair of 737MAXs.

They probably assumed that is what was going on. That the other pilot was doing something obviously wrong is not a natural assumption, when there is an alternative.

It seems like Air France has got some procedural drills to work on. One would have hoped that happened already. Somebody should check, this time.

Technicality, but the 737MAX MCAS was rolling a trim wheel, not moving the yoke.

Exerted force on the yoke via the elevator, driven by the trim tab. But they also yank directly on the yoke, for lots of reasons.