# Are more fatalities in commercial aircraft accidents caused by pilot error than by automated features?

This claim struck me:

Overall these automated features are an overall benefit to everybody. They are preventing some of the more classic aviation disasters that pilots can induce into these aircraft.

### Question

Is it correct that pilot error causes more fatalities in commercial airline accidents than those caused by automated features, like MCAS?

The reasons it struck me are two-fold:

1. There have been very prominent accidents involving automated features in recent years
2. I don't know of many cases of pilot error causing commercial airline accidents (although I don't follow closely).

### What I know so far

I ordered List of aircraft accidents and incidents resulting in at least 50 fatalities by date descending and examined the last 6 years' of crashes. It's easy to spot two caused by automated features: Ethiopian Airlines Flight 302, and Lion Air Flight 610, but I lack the knowledge to tell which others were caused by automated features, nor which were by pilot error. Note: I only chose 6 years because that's what fit on screen, a larger sample (perhaps 25 years) would be much more rigorous.

• Hi Steve ,welcome to ASE. Your question is hard to answer, as one reason alone very seldom leads to an accident. Coincidentally, a model often used to investigate reasons leading to accidents is some times called "Reason model" by it's formal propounder James T. Reason (go figure...), more descriptive name is the "Swiss cheese model". Accidents generally are a instances where a hazard slips through a number of layers of defense, which for one reason or another are not functioning properly. Typical slices in aviation are the pilot, redundancy of the systems themselves, backup systems etc... Feb 7 at 14:15
• This may be somewhat unanswerable due to the prevention paradox: if automation was capable of mitigating all pilot errors, this would mean that all remaining accidents are due to failing automation. Feb 7 at 14:43
• One problem with this question is that accidents (not only in aviation) typically have multiple causes. Take, for example, the seemingly clear-cut automation issue with MCAS. Yes, the design of MCAS and the missing documentation were flawed; but there was also (1) Bad maintenance (sub-par replacement parts); (2) Insufficient pilot training (3) Pilot confusion and failure to react timely and correctly. This will likely be the case with most failures short of bombs and missiles. Feb 8 at 13:30
• Note that most "automated failures" are actually human errors because humans designed and manufactured the automated systems. Of course, unlike the pilots, those humans aren't typically in the aircraft when they crash. Feb 9 at 4:56
• If you're interested in this topic, I'd recommend taking the time to watch the classic lecture about automation dependency Children of the Magenta Line (more about autopilots than newer systems like MCAS) by Captain Warren Vanderburgh of American Airlines in 1997. It's not really either pro nor anti-automation, but more of a really meaningful discussion about the importance of being able to select and use the right mode and level of automation for a situation and having the proficiency to drop down to lower levels of automation when necessary. Feb 9 at 8:27

"Automation failure" tend to be lumped in with all other "Mechanical failure" in any statistics. Those statistics are fairly easy to find but absolutely back up the claim that pilot error causes far more accidents than mechanical failure

If we have a look at the data from the 2010s (which is the most recent in that link above, but has remained fairly consistent thoughout the decades)

Pilot error Mechanical Weather Sabotage Other
57% 21% 10% 8% 4%

So pilot error accidents causing fatalities accounted for more than the rest put together in the 2010's.

This begs the question of why have humans in charge if all the do is make mistakes! Which is a logical fallacy, what you can't see from the statistics is all the times the human up front made a decision which prevented an accident from occurring. It's still very difficult (if not impossible) to program a computer to make decisions in a rapidly changing/dynamic environment.

• Those stats are interesting but difficult to rely on, as the database doesn’t show the cause for each accident. Also, there are often multiple causes in an accident. Take AF477 for instance. Is that an equipment failure (the pitot tubes froze and started giving incorrect information) or a pilot error (spatial disorientation and wrong reaction)? Also I counts accidents only, not fatalities. Feb 8 at 8:50
• Something to add to that last paragraph is that a lot of accidents where Human error was the cause, can also be traced back to a particular automated system malfunctioning, forcing the humans to do actions manually but are no longer used to. It's a paradox: more automation means less human involvement required, thus less pilots gaining experience manually flying the plane, thus when something does go wrong, they are less prepared to take over the task. IIRC Fuerza Aérea Uruguaya 571 is an example of such an accident. Feb 8 at 9:16
• @jcaron: AF447 was definitely pilot error. The reason the pilots are there is precisely to deal with the situations in which the computer is not sure what to do, including the cases where an instrument fails. They had a perfectly functioning plane with a single sensor mulfunction; the pilot flying never understood what was happening, and the two others never completely overrode him, all three ignoring the majority of their instruments. Feb 8 at 17:09
• Which is why, @Opifex, driverless cars are so scary. When (not if) he automation fails, the already poor (American, at least) driver will be so far out of practice, he won't have a clue what to do. Feb 11 at 18:24

Yes the flight crew is the key factor in the majority of crashes. It's for this reason that two big innovations are largely responsible for the massive drop in fatal accident rates since the 90s (there hasn't been a crash of a heavy, loaded with passengers at least, in the US in over 20 years - think about it... 20 years, and that was a pilot-induced rudder-fin failure).

The first and biggest is Crew Resource Management theory, which started to be implemented in the West in the 80s. CRM was so successful that its concepts have spread far and wide outside aviation, from policing to surgery. The other one is Flight Management Systems and the two dimensional map navigation display made possible by the "glass cockpit" (also in the 80s). It's hard to exaggerate how much 2D moving maps reduced crew workloads in terminal areas and on approaches, by reducing the dependence on a pilot's internal mental "map making" abilities to maintain situational awareness.

Fatal crashes caused by system failures are exceedingly rare, because at the certification level, system design requires a mathematical probability of better than one in a billion of a component failure leading to an airframe loss. MCAS can be discarded as a Black Swan event, a culmination of a set of very unique circumstances and decisions at the certification and development level (I mean, so unique that the project chief test pilot was charged criminally for hiding information from the FAA). MCAS can be thought of as "pilot error" induced by a bad configuration that slipped through the cracks you could say, but which some crews could have coped with.

Double and triple systems redundancy is required to maintain this less-than-one-in-a-billion probability of disaster. It's not perfect, but it's the main reason airplanes don't fall out of the sky regularly when components fail (and things are breaking down on complex aircraft all the time as you can see in an airline's dispatch delay rate - those regular delays leaving the gate are usually because the plane arrived with something broken that has to be fixed on the spot).

"Automation" in itself? Yes automation could probably be said to have prevented some disasters, but in the bigger scheme of things, not as much as something like CRM. FADEC makes engine operation a little easier and more convenient for the crew, but it mostly drives costs down. Fly-By-Wire and other system automation controls? They help reduce crew workload overall, and maybe cover for sloppy flying from time to time, but they are also mostly cost reduction drivers (FBW eliminates a massive amount of hardware). On the C-Series/A220, the systems automation level is just about at the pinnacle of current development, and the main benefit (and selling point) is a huge reduction in maintenance man-hours, more than safety.

• Wasn’t the point of FADEC to eliminate the Flight Engineer, which reduced both opex and human error? Feb 7 at 18:10
• there hasn't been a crash of a heavy in the US in over 20 years. Well, there's GTI 3591. That was a 767. But, once again, pilot error. Feb 7 at 18:14
• FADEC came about long after FE's were already being eliminated by the new gen of 2 crew heavies like the 767. But to your point, yes pre-FADEC you had electronic engine control, basically computer trimming of fuel schedules, and computer controlled monitoring and sensing with EICAS synoptics presented directly to the front seaters, and and those allowed FEs to be eliminated. FADEC is mostly about weight saving - eliminating cable runs as with FBW - and the ability to run engines much closer to their margins with much tighter control over temperatures, stall/surge, improving efficiency, etc. Feb 7 at 18:35
• @TomMcW hadn't heard of that one. Thanks. There was also a crash of a heavy in the Arctic, flown by First Air in the mid 2000s where it ran into a hill because the capt became obsessed with regaining LOC and the FO was too timid to stop what was happening, but that was in Canada. Are their any other US heavy crashes in that time frame? Feb 7 at 18:39
• Asiana 214 happened in the US, and that was less than 10 years ago. First loss of a 777, multiple fatalities. Feb 8 at 3:01

I'd like to point out a difficulty in providing you with an answer:

We know well the accidents caused by pilot error and automated features. However, we cannot be sure of how many cases would have gone downhill if the automated features weren't there to assist the pilots. Sometimes the pilots may have understood it but the cases weren't logged, and other cases the features might have provided support in the background and "saved the day" stealthily. (I'll try to get some examples in the future).

Now, I am hardly in a position to answer this, as my only experience is youtube watching and a couple hours flight simulator. But the maturity of the aviation industry as a whole amazes me (nothing is perfect, but I've not really seen any better). Each automated feature was born to assist, and many times the reason for it was to prevent conditions of some accident to happen again. So, in general, I'd say the features help.

To interject some personal ipinion: MCAS was one of the rare black sheep, with multiple entities working in a bad way. But it is rare. I guess it is a small reflection of the aviation paradox: it's the safest transport method, but if an accident happens in the other end of the world, you will learn about it and it will create fear, while the (thousands?) people dying via eg car is nothing interesting so it's not in the news.

Also, after stevec's observation in the comments, the opposite is also true: we also cannot be sure of how many cases the pilots saved the day by disabling the automated features. In order to keep supporting my pro-feature point, I will argue that features that are consistently creating trouble will be noted as such by pilots and reviewed....hopefully.

• Black sheep=black swan? Feb 9 at 17:39

In comparing the lethality of automation induced accidents with pilot induced ones, you are paying very much attention to a very small aspect of a very big story. As a result, the direct answer to your question (a simple 'Yes') is bound to not make you much wiser.

Moreover, automation induced accidents are not always due to automation malfunction. Quite often they happen as a result of misapplication of automation devices that themselves do not malfunction, like in the case of MCAS. In the same fashion, pilot error is often the cheap and easy way out of having to explain what caused an accident. Remember how even the famous 'Miracle on the Hudson' was initially claimed to be the result of pilot error?

In general, the big difference between automated decision making and piloting, is due to the fact that pilots are unique. It may sometimes be hard to categorically upgrade all automated devices of a certain type after a malfunction of just one of them, but it can still be done. Doing the same with pilots is virtually impossible. This speaks strongly in favor of automation, even in a world with generally highly skilled, well trained, current and proficient pilots.

The claim as stated is a generalization and as such it is true. There will be exceptions, but any argument against automation is doomed to die soon anyway , as aviation solutions more and more come to rely entirely on automation. An increasing number of in flight decisions is made automatically, simply because no human can ever be capable of making them. This makes room for ways of flying that up to now were strictly confined to science fiction. If anything will ever be able to make you fly like Tinkerbell, it's automation.

By the way, the single most deadly crash in aviation history, when a KLM Jumbo crashed into a PANAM one on take off from Tenerife , was caused by pilot error. You must have been following from quite some distance to miss that one.

• The "Miracle on the Hudson" was never once claimed to be pilot error; you'd only think that if your only knowledge of it was the film. Chesney Sullenberger forced the studio to invent alternative names for all the investigators in the film because he hated so much that they were being libelled that way. There certainly have been military crashes caused by faulty controls which were (falsely) put down to pilot error, most notably with RAF Chinooks in the 1990s, but it's far less likely in civilian life. Feb 9 at 11:39
• ... And MCAS certainly did malfunction, due to a single faulty sensor. More than that, the design was inherently faulty because nothing relating to flight controls should be reliant on a single sensor. As an engineer who's worked in safety-related systems for cars, we wouldn't tolerate the processes in MCAS on a car dashboard display, never mind on a car engine controller; so the idea that something so fundamentally incompetently designed could go into a plane is horrendous. There are no two ways around it - MCAS was the sole cause, and the "engineers" should be in jail for murder. Feb 9 at 11:45
• @Graham There are sources other than the film that refer to a challenge similar to what was presented in the film, but honestly, I don't know what to believe any more. I do believe the film gave a generally false impression of the NTSB as in what objective they pursue , but that's my personal opinion. Funny thing is, it still works as an illustration here.
– user55607
Feb 9 at 15:14
• @Graham The sensor that fed data to the MCAS malfunctioned. The training malfunctioned. The design malfunctioned. The process for certification malfunctioned. But the MCAS as in the device itself, never did anything it wasn't supposed to do. I have personally been heavily into the 737 MCAS story. Ask anyone here :). Over time I have come to understand that there is way more than meets the eye to it. Blaming it on any one person or group of persons is more dangerous than not blaming it on anyone. If I would have to point the finger in any one direction, I would point it at culture.
– user55607
Feb 9 at 15:25
• @Graham: If pilots were properly trained to react to an MCAS malfunction, then MCAS would not have been a safety-critical system. On the flip side, a failure to train pilots to recognize that the 737 Max has different natural flight characteristics from the 737 could easily have caused a crash under some circumstances where MCAS was working as well as it physically could, but flight conditions were outside the range where the MCAS could make the plane handle like a 737. Feb 9 at 21:17