Why is automation dependency encouraged in modern aviation ?

[FullWings]

I remember doing a brief for an approach to somewhere - can’t remember exactly where but it was a decent 1st World airport. Started talking about the offset VOR and how we were going to fly it and the guy in the RHS said “I don’t do NPAs”. “Hahaha!” went I, then I realised he was being serious. I asked him how we were going to get in, given the ILS was U/S in the AIS and the weather was definitely not good enough for a visual. OK, I’ll do it in RNAV he said; fine, I replied, but as it’s not in the database and offset, how are you going to do that using our SOPs? Cue look of horror and confusion.

Look, I said, let’s rewind back to the beginning and forget you said you didn’t do VOR approaches. We went through the entire process: FMC set up, expected modes and FMAs, contingencies, visual segment, blah blah. Didn’t take that long and he flew a perfectly acceptable raw data VOR that I was able to land off, which seemed to amaze him. A few beers later, he confessed that he had become very under confident with raw data, A/P in or out, and had been dreading the next sim where, by all accounts, you ended up having to do something like the above. Not an issue, I said, it happens to us all but the way back is practice, not avoidance. Ask and ye shall receive...

[Judd]

Those guys are a timing bomb for the next air disaster but let’s keep denying it.

I can well believe it. Friend of mine flying in the Asia region crewed with a local cadet just out of flight school and who had just completed his type rating. They were flying in Cirrus cloud, calm conditions, when the cadet first officer nervously said to the captain "I don't like flying in cloud..." This from the legal second in command of a jet transport. Image the chaos that would occur if the captain became suddenly incapacitated. In fact it does't even bear thinking about.

[safetypee]

Recent posts, HPSOV L, Vessbot, Judd, #77-79, consider the same problem from different aspects - situation awareness.
Perception, change, understanding, time, are critical components of awareness - how we make sense of things.
Reluctance to 'click click', 'my plane'; the difficulty is in identifying the need to act, that some situations require a change of action.
Pilots must consider the 'AP's perception', requires working knowledge, limitations, awareness.

Viewing accidents from human viewpoint usually concludes human error - more training. Alternatively, considering a joint human - automation system, what might be concluded as a perfectly serviceable aircraft, becomes an inadequate combination of man and machine in a particular situation (Man Machine Situation).
Not more training, but improved system alerting that auto-thrust is not available in some circumstances, that attempting to engage AP outside limits gives a 'uh uh' alert, cueing a change in awareness, an awareness which is assumed by others to include every circumstance, but circumstances are rarely as assumed.

First check assumptions, reconsider circumstances - avoid.
Improve the machine to aid the pilot - joint system.
Last, reluctantly consider the very difficult task of changing human behaviour which can never be assured or as assumed - not all situations are foreseeable.

Slowly deteriorating situations are difficult to identify, cf 'Boiling Frog'. There is a natural reluctance to act because action is already in hand, except that as situations change so must actions, which requires understanding, awareness, mental skills - which degrade faster than manual skills.

The industry requires a combination of improvements for man and machine, but the myth that human aspects are more effective come from management thinking. The issue is not about manual flight, its knowing when to fly manually, and adapting normal skills for abnormal situations.

https://nescacademy.nasa.gov/video/a...118f7585baa81d 'Man-Machine' >'Slides', 'Download'

http://hfs.sagepub.com/content/56/8/1506.full.pdf 'Manual flight skills'.

https://en.wikipedia.org/wiki/Boiling_frog

[PilotLZ]

The airline policies which tend to take the most heavy-handed approach against manual flying are mostly concentrated in a certain part of the globe where there are lots and lots of expats who often complain about the locals' level of training being substandard. That's where you can expect to be told to consider autoland as the normal way of landing day in and day out. For whatever weird reason, many of said countries don't even bother teaching their local pilots the basics of flying. Many are never taught what they should be looking for during landing, many don't even look outside. All they have been taught is "at X feet set Y degrees of pitch, pull the thrust levers all the way to idle and wait". To make matters even worse, FOs hardly get any PF time. The Captain is PF by default. Many companies even explicitly forbid the FOs to handle the aircraft unless there's a TRI in the left-hand seat. What sort of Captain can such a FO become, I fear to think. By the time they get into the LHS, they likely still have less sectors as PF than a junior FO fresh out of line training in other parts of the world.

Why did I bother writing that? Because, very sadly, "policies" like that make automation dependency a status quo in many carriers. A pilot with such a company is trained that way from day one, right until he/she becomes a Captain or a TRI and begins hammering the same way of doing things into the heads of those joining after him/her.

What can be done about it? Tightening the regulatory requirements for manual flying proficiency sounds tempting right until you realise that it can also backfire. For example, if you require a pilot to perform a raw-data approach every X days, how can you be certain that, during a period of low flying activity, his/her only flight won't be in conditions where flying raw-data is not appropriate? Hence, I don't think that hard-and-fast rules are the best way forward. The best strategy would be involvement of the authority, the aircraft manufacturer and national and international safety bodies into stimulating airlines to rethink their training programmes and objectives. Airbus are already making some steps towards this. Let's hope that others will follow suit.

[vilas]

​But we have crashes like Indian 605 where in a perfect plane on a clear day, they fell into mode confusion on the approach and couldn’t bring themselves to just fly the airplane, instead trying to untangle the modes all the way into the crash. Or, Flash 604 where in IMC the AP failed to engage on departure, and instead of flying the plane they maintained a panicked and sustained effort to engage it, again all the way into the crash. Or, Asiana 214, with again a perfect plane on a perfect day where, unlike the other 2 examples and to their credit, they clicked the red button and decided to fly the airplane. ​​​​​​

Vessbot I disagree on 2 of the three accidents you quoted. Indian airlines and Asiana both were visual approaches manually flown. There was no mode confusion because even to get confused you need to know what's going wrong. They were flying an approach without ever looking at the speed or ROD. The IAS in Indian case was 26kts below Vapp and in Asiana case was 31 kts below Vapp. They would have crashed even in a Dakota(actually more easily). The tragedy is both were under command check and not copilot release check.

[Check Airman]

Originally Posted by pineteam

Of course pilots should be able to hand fly in any conditions. I don’t expect my colleagues to fly as accurately as the AP but he should feel confortable at any time to disconnect the AP If required and fly within acceptable limits. But sadly many aren’t. Once a friend of mine was observing my flight; that’s part of the upgrade process. That day I was flying with a very weak first officer; one of the weakest actually: one of those guys who never ever hand fly. That day I was doing a raw data approach ( no A/THR, no bird) and to show my friend how nervous this guy was even in a perfect CAVOK day. I asked the FO to take control pretending I needed to adjust my seat. He had nothing to do really, only keep the same pitch and leave the leavers on Idle. We were still passing 6000 feet descending to 3000. As soon as he took control, I looked at him he was completely overwhelmed, nervous as hell and shaking. Not even 20 seconds later he told me very nervously: “Let me know when you ready to take control!” I then took control and told him that he needs to relax and at anytime he could practice raw data with me. He is a nice guy actually tho; I actually feel sad for him; but yeah you get the idea and this kind of weak pilots is not an isolated case. Some are even on the left seat and I bet you in your outfit too. Those guys are a timing bomb for the next air disaster but let’s keep denying it... SMH

Interesting story. My inner instructor would’ve come out, and I’d have let him hand fly for a bit to build confidence. Still, very sad that we’re discussing this though.

[Check Airman]

Originally Posted by vilas

Vessbot I disagree on 2 of the three accidents you quoted. Indian airlines and Asiana both were visual approaches manually flown. There was no mode confusion because even to get confused you need to know what's going wrong. They were flying an approach without ever looking at the speed or ROD. The IAS in Indian case was 26kts below Vapp and in Asiana case was 31 kts below Vapp. They would have crashed even in a Dakota(actually more easily). The tragedy is both were under command check and not copilot release check.

That makes it all the worse. They were in VMC. Their scans and skills had deteriorated to the point where a visual approach could not be executed, let alone an approach in IMC.

[Check Airman]

Originally Posted by FullWings

Not an issue, I said, it happens to us all but the way back is practice, not avoidance. Ask and ye shall receive...

Thank you for being a Captain (uppercase C).

[KayPam]

Originally Posted by PilotLZ

The airline policies which tend to take the most heavy-handed approach against manual flying are mostly concentrated in a certain part of the globe where there are lots and lots of expats who often complain about the locals' level of training being substandard. That's where you can expect to be told to consider autoland as the normal way of landing day in and day out. For whatever weird reason, many of said countries don't even bother teaching their local pilots the basics of flying. Many are never taught what they should be looking for during landing, many don't even look outside. All they have been taught is "at X feet set Y degrees of pitch, pull the thrust levers all the way to idle and wait". To make matters even worse, FOs hardly get any PF time. The Captain is PF by default. Many companies even explicitly forbid the FOs to handle the aircraft unless there's a TRI in the left-hand seat. What sort of Captain can such a FO become, I fear to think. By the time they get into the LHS, they likely still have less sectors as PF than a junior FO fresh out of line training in other parts of the world.

Why did I bother writing that? Because, very sadly, "policies" like that make automation dependency a status quo in many carriers. A pilot with such a company is trained that way from day one, right until he/she becomes a Captain or a TRI and begins hammering the same way of doing things into the heads of those joining after him/her.

What can be done about it? Tightening the regulatory requirements for manual flying proficiency sounds tempting right until you realise that it can also backfire. For example, if you require a pilot to perform a raw-data approach every X days, how can you be certain that, during a period of low flying activity, his/her only flight won't be in conditions where flying raw-data is not appropriate? Hence, I don't think that hard-and-fast rules are the best way forward. The best strategy would be involvement of the authority, the aircraft manufacturer and national and international safety bodies into stimulating airlines to rethink their training programmes and objectives. Airbus are already making some steps towards this. Let's hope that others will follow suit.

This is completely incredible.
My airline is lightyears away from this and I still think they don't encourage raw data enough.

The most efficient solution for this type of airline could be simulator requirement, couldn't it ?
Requirement to be able to fly a complete raw data 2D approach of the TRE's choice. It does sound completely reasonnable to be able to do this, right ?
Plus, there are clear boundaries for what is acceptable. Did the pilot go below the safety altitude ? Did he descend at more than the authorized VOR/ADF deviation ? Did he go further than 0.5nm away from the DME arc ? All of this "without a go around or immediate correction".

Then each airline will decide how to reach that goal.

[FlightDetent]

Originally Posted by KayPam

This is completely incredible.

Also mostly made up to suit the narrative.

[Vessbot]

Originally Posted by vilas

Vessbot I disagree on 2 of the three accidents you quoted. Indian airlines and Asiana both were visual approaches manually flown. There was no mode confusion because even to get confused you need to know what's going wrong. They were flying an approach without ever looking at the speed or ROD. The IAS in Indian case was 26kts below Vapp and in Asiana case was 31 kts below Vapp. They would have crashed even in a Dakota(actually more easily). The tragedy is both were under command check and not copilot release check.

I don't follow. If you know what's wrong, then you're not confused, you're the opposite: you accurately know the situation. Confusion is if you think the automation setup will result in something different than what it's actually resulting in. And it can come in 2 flavors: first, not even knowing anything is wrong and merrily blundering toward CFIT; and second, knowing something is wrong but not understanding exactly what or why, and usually accompanied by an attempt to fix by frantically pushing buttons and twisting knobs.

Unless you think that the Asiana pilots intended to be 31 knots slow, or the Indian pilots intended to be 26 knots slow, (and intended to level at a selected altitude 2300 feet underground) then how can you say that they were not confused about the modes, or what the modes should have resulted in?

Their failure to monitor their airspeed didn't come from a generic laziness toward monitoring (although it was part of it), it came very specifically from a conditioned expectation that the automation mode (AT only, yes AP was off) would have it taken care of. I don't see how this would happen in a Dakota.

---

Actually we disagree with all 3 and not 2, because by process of elimination, it appears that you think the 3rd one (Flash 604) was mode confusion, while I think it's not. They accurately understood that the AP wouldn't engage (or would quickly disengage after trying), but they did not accept that this was forcing them to fly the plane, and instead placed all their mental energies into continuing to try to engage it as it completed a wingover into the ocean.

[Vessbot]

Originally Posted by safetypee

Reluctance to 'click click', 'my plane'; the difficulty is in identifying the need to act, that some situations require a change of action.Pilots must consider the 'AP's perception', requires working knowledge, limitations, awareness.

There are two things, and I’m not catching on if we agree or disagree. First, the identification of a need to act, and second, the ready base of skill to fly the plane, if the need happens to be for that. And I think the reluctance comes from an interplay of both. The skills have to be there, hard-baked to a point of casual ease; as I said in another post, if you’re unwilling to use the skills, it’s kind of the same as not having them. So the yearly sim (or maybe several-times-per-yearly, easy VMC into the quiet outstation) does not count. I think of it as an analogy of two “channels” of mental habit/action. In most automation cases, the starting point is with it engaged and in the automation channel, and for all of us, that is the easier way, and the place you spend most of the flight.

Then if something creates a need to fix a situation, it may be of low intensity, and the most appropriate action would be to probably stay in the automation. Let’s say you accidentally double-push a button instead of single push, so it goes in the wrong mode, and you realize it immediately. Just push it again, verify the right mode is up, done. No need for AP off.

Or the case may be severe. Like Flash 604, We’re low to the ground, the AP won’t engage, and after say 2 or 3 tries and bank angle going through 45 degrees, it’s time to fly the plane. It’s an obvious need to shift to the manual mental channel. And to overcome “the reluctance” to climb the ridge from one to the other.

But, (and I think this is an important part) the realization doesn’t come in a single shocking unambiguous announcement (like a FIRE warning), rather it builds up gradually. At the first failed try (and maybe wings are still level) is there a need to shift channels? Probably not, like my double-push example above. Just try again. OK second try, and it’s banking through 20 degrees? Maybe, maybe not, “you’d have to be there to really know.” 3rd try, and banking through 45 degrees? Hopefully all of us are going click-click “my plane” by now. 5th try? People are writing about us on forums decades later. But it’s a sliding scale of time-increasing realization, or the boiling frog like you said. And, when the automation mental channel is worn far deeper and more comfortable, with this sliding increase in need, the light bulb may never go off, as we emotionally grasp harder and harder at doing anything we can within the automation channel, and the panicked tunnel vision narrows down only on that view. The sliding-scale nature of the development allows us to be in denial and reinterpret the severity as a lower level until it’s too late.

So the only way to help that (and this is where the analogy kind of fails) is to have the manual channel also worn as deep and comfortable as we can get it, which would lower the barrier between the two and make the jump feasible at an earlier stage of the development.

Not more training, but improved system alerting that auto-thrust is not available in some circumstances, that attempting to engage AP outside limits gives a 'uh uh' alert, cueing a change in awareness, an awareness which is assumed by others to include every circumstance, but circumstances are rarely as assumed.

Well they already had alerting of what mode was active at the top of the PFD. Sure, even better alerting could be designed after careful study of the exact sequence of cascading events of the accident in question, and programming the plane to recognize that the pilot’s intent may be mismatching their inputs, but that only addresses this accident. (Though it should still be done).

What about all the future ways that pilots will dream up to fall into other compound cascading confusions that weren’t thought of? It’s just patching one hole in a dynamic pilot-aircraft-environment system where countless others might spring new leaks. It’s just adding more automation patchwork, and misses the fundamental failure that is outside the automation itself, but rather is in the pilot. The failure to recognize the difference between the intended state and the actual state (of the raw data, physical results), and if such exists (more than obviously a quick transient) click-click/ “my plane.” This must never stop being our job until the stick has been replaced by a mouse and we no longer have an interface with the control surfaces.

[HPSOV L]

A good discussion. I think Safetypee made some thoughtful observations and had some interesting links. The Dekker report is also good reading:


https://www.onderzoeksraad.nl/nl/med...t_s_dekker.pdf

[safetypee]

Vessbot, how do you recognise the need to disconnect ?
Are you able to describe the process which you use, which will always apply in every situation. If this starts with understanding the situation, then how do you understand.
To you it might be obvious, to others no so depending on the situation, experience, training - even of this can be trained at all.

How can we be sure that the situation as we understand it is correct, and even if correct that we will choose an appropriate course of action; errors in understanding or errors in acting (not acting).
http://www.pacdeff.com/pdfs/Errors%2...n%20Making.pdf

http://perigeantechnologies.com/publ...rspectives.pdf
Note the myths:
Not just joining up the dots, but knowing what forms a dot in the first instance, where the meaning of a dot varies with context.

[Judd]

What can be done about it? Tightening the regulatory requirements for manual flying proficiency sounds tempting right until you realise that it can also backfire

The only problem with tightening the regulatory requirements is that the regulatory chiefs themselves are mostly retired airline pilots who are former children of the magenta line. For them it is easier to look the other way. Unfortunately it looks like we are stuck with the fact there will always be occasional crashes caused by incompetents in the left hand seat who simply can't fly. The cost of doing business?

[Chris Wheeler]

I wouldn’t worry too much about exact wording. You understand the company wants you to be proficient in automation as well as manual flying. Try to follow your companies FOM, as far as when you can manually fly an approach. After that, only you know what you need to work on. Then do what’s necessary to be proficient.
I also think a lot of what a company prints has to do with covering their butts.

[Vessbot]

Originally Posted by safetypee

Vessbot, how do you recognise the need to disconnect ?
Are you able to describe the process which you use, which will always apply in every situation.

Yes. You compare what you would like the airplane to be doing with what it is actually doing, and if those 2 things mismatch (and especially if you feel behind in any way in recognition, or proposed solution), you disconnect. (The old cutesy chestnut, "What's it doing now?")

An exception would be if the result is a common sequence of events in type, that you recognize immediately as well as the automation solution, (you never felt mentally behind) then it's OK to just do the automations solution. ("It's doing that thing again"). Such as in the R&N thread about the 777 stall warning out of JFK, people are talking about a common speed loss due to too-early altitude capture, and a double-push of the altitude button to deal with that)

If you're not sure which one you're in, you're in the first one.

If this starts with understanding the situation, then how do you understand.
To you it might be obvious, to others no so depending on the situation, experience, training - even of this can be trained at all.

How can we be sure that the situation as we understand it is correct, and even if correct that we will choose an appropriate course of action; errors in understanding or errors in acting (not acting).

This doesn't sound to me like a question about comparison of the understanding in your mind vs. the understanding in the airplane's computers, but just a question about the understanding in your mind. How do you understand? You take all your mental habits and abilities you've built up since flight lesson 1, and bring them to bear on all of the information you're receiving about the current situation. Could that be faulty? Yes. Could the airplane have a more accurate SA than the pilot, therefore the automation should be trusted? Yes, and there have been accidents due to that, like the Sukhoi business jet on a sales flight with a customer and a test pilot or something like that, where everything was functioning correctly and they got a GPWS but they blew it off and CFIT'ed.

We spend a lifetime/career assembling all of our experiences into the ability to build accurate SA and judgement... from mentors, hangar stories, accident reports, ground school, books by old sages, official manuals, etc. And hopefully become better at it than this crew. But at the end of the day in the cockpit, we're still the final arbiter of which SA picture is the most accurate, and must act on that; there's no logical way around that. If the computer has a better picture than you, then you still must decide that, and by virtue of that, that has become your own picture as well, to replace your previous one . There's no second version of you, outside of the situation, that judges which picture is better, and decides which one to pick. No, it's only the first you (until they make automation that can override you. So far it can make strong suggestions, but it doesn't override you.)

Evaluating that you're too confused for now, and making the willful choice to trust the automation for the time being, may also be valid, but in the end it's still an example of the same thing. Once you've made that evaluation then you, as the final arbiter, are making the choice (good or bad, result not known yet) which way to go.

---

I haven't had time to look at the links, but hopefully I will tomorrow.

[FullWings]

Vessbot,

That’s a couple of very cogent posts, dense with some really good observations.

For me, having had a fairly long career in aviation but coming from an engineering/science background, one of the (many) issues that gets me is the user interface that a modern commercial transport presents to its operator. In a word: suboptimal. They are an unholy mix of a century’s worth of ideas and technology, keeping the bad as well as the good for some kind of continuity. IMO many EFIS+FMC presentations are actually worse for SA in some areas than an equivalent steam-driven setup: un-annunciated hidden modes, weird logic and a plethora of known but too-expensive-to-fix bugs in prehistoric firmware/hardware.

As far as manual flying goes, I am mostly with those who say that it’s worthless to just follow the FD as you’re just inserting a monkey between the AFDS and the control surfaces. No need for any kind of instrument scan, just keep those needles crossed in the box (or whatever your aircraft requires) and you’re golden. I’m a little bit of a bully in that respect as if someone asks if they can fly manually, I say sure, but the FDs are going off with the AP! Make ‘em sweat.

The idea of giving airline pilots a few hours every now-and-then in a light aircraft to get the old scan and handling ability up again often comes around and is generally pooh-poohed as being impractical, expensive, etc. Considering the cost of most FFS at £thousands a session and the limited exposure in the real thing to situations where you don’t have to compromise safety and regulations too much (non-RVSM, not at the end of a 12hr sector into bad weather, definitely not 3am on your body after a dodgy curry the night before, etc.) it would actually be more cost effective and 100% of it would be manual flying, IF and/or VFR. An explanation that sort of holds water is that this kind of activity looks suspiciously like “fun” and therefore what right-minded company would encourage such a thing, even if it led to increased safety and reduced wear-and-tear on airframes. Or, even pilots who were confident enough to do a visual circuit instead of an 12-mile ILS, saving £££ and getting there quicker?

[vilas]

I don't follow. If you know what's wrong, then you're not confused, you're the opposite: you accurately know the situation. Confusion is if you think the automation setup will result in something different than what it's actually resulting in. And it can come in 2 flavors: first, not even knowing anything is wrong and merrily blundering toward CFIT; and second, knowing something is wrong but not understanding exactly what or why, and usually accompanied by an attempt to fix by frantically pushing buttons and twisting knobs.

Again no! I have said it before with AP on or off pilot's scan remains same only difference is with AP on, the AP makes changes and in manual pilot makes changes. With or without AP It's the loss of scan that causes the problem. Mode or no mode, right or wrong mode, you simply cannot make an approch without a periodic glance at the speed. In the cases we are discussing the crew never even expressed a surprise or utter a word regarding falling speed because they never looked at the speed and worse not even the mode. They assumed a mode that wasn't there, they assumed a speed that wasn't being maintained. In both the cases even the go around was not because they noticed speed or the ROD but because they were not reaching the runway otherwise they would have executed a 3 or 4G Landing. In case of Indian airlines they lost their lives because they were nine seconds too late. The engines were spooling up. Somebody needs maintain the parameters either the AP does it, the mode does it or the pilot does it. Nobody is not acceptable.

[FullWings]

I have said it before with AP on or off pilot's scan remains same only difference is with AP on, the AP makes changes and in manual pilot makes changes.

I’m not too sure the scans are the same, as with AP in the aircraft is being controlled through the MCP/FMC, which necessarily takes attention away from instrument monitoring. Most SOPs are when the PF is hand-flying, the PM will look after all the selections so the PF can concentrate on the instruments?

With or without AP It's the loss of scan that causes the problem. Mode or no mode, right or wrong mode, you simply cannot make an approch without a periodic glance at the speed.

Agree with that.

Maybe there’s a third type of confusion: unconscious, where you don’t even know you don’t know what you’re doing? (Going Rumsfeldian here...)