In the Airbus accident, the level of automation was a contributing factor to the accident. The fly-by-wire system prevents the pilot from placing the aircraft in an unsafe condition. However, this system was automatically deactivated (which is an extremely rare occurence) when the pitot/airspeed tubes were covered with ice and the computer hence couldn't make correctly informed decisions about the aircraft.

In that case, however, you could point to a big UX problem: It was not announced in big, red letters that the fly-by-wire system had in fact been deactivated. If I remember correctly, this was hidden under some menu in the flight computer and simply abstracted away to a "master caution" light or something in that vein.

> It was not announced in big, red letters that the fly-by-wire system had in fact been deactivated

It was announced in medium-sized orange letters reading “F/CTL ALTN LAW (PROT LOST)”. At least one pilot did notice this; the co-pilot who wasn’t pulling up the whole time said “alternate law, protections lost” out loud, as recorded on the CVR, when the message appeared.

I think he might have noticed, but not fully realized the implications of it. His behaviour following at least did not indicate that he was concerned about a stall, and there were no further mentions of this warning.

Would that not be a UX issue? Perhaps the warning needs to include the phrase "Failsafes deactivated" announced repeatedly.

Yep, I would say that, as far as I remember there were also several other warnings at the same time. There were some recommendations in the accident report regarding UX improvements.

I think there was a similar problems at Qantas Flight 32, where the crew had to spend a lot of time going through 54 error messages after an uncontained engine failure.

right, but there was also a pilot who was pulling back on the stick the entire time.

And the airplane averaged the inputs from the two sticks, which strikes me as a positively BRAINDEAD design choice. Why would you ever want to average the two inputs when flying a plane? Are you hoping that one pilot pushing and one pulling will somehow, on average, fly the plane correctly?

The braindead decision (in my uninformed, armchair pilot opinion) was made when the controls were no longer physically linked. Once you have controls that can move independently of each other, none of the options for handling conflicting inputs makes sense to me...

Sum them?

Only take the most extreme of the two inputs (in any given direction)?

Only take the most recently seen input?

Averaging makes more sense than any of those, given the premise of "Control sticks that can move in different directions"

You're right, but I'd go with "the secondary control is ignored and clearly marked as such" ("ignored" light turns on or something). But yeah, if they aren't linked, there's not much you can do that will make sense.

Nah, the right way to compute the final inputs in a case like this is, "This design is stupid and is going to get somebody killed. There is nothing I can do at this level to fix it. You must physically link the two controls."

When assuming that in the normal case the pilots are never doing the opposite of each other ... the difference between each stick is probably less than 5% or something. Agreed that in this case it seems the worst thing to do.

It is rarely discussed, but a major factor contributing to the Airbus crash was that the software did not derive airspeed from engine parameters and use it to verify the pitot tubes. The common mode failure of the pitot tubes could have been anticipated and protected against.

The problem there being when something unexpected happens that the automation can't cope with, your skills have atrophied.

I'm not a pilot but I am a technical diver, and we eschew many of the conveniences that recreational divers use, such as relying on dive computers. We work out dive plans on paper, and backup plans, and carry them with us on a slate. The only instruments we rely on are watch, depth gauge and compass, and these are of the simplest possible design, and we have backups for them. If something happens on a deep dive or a cave dive, you know exactly what the state of your system is and can reason about it. Not so if you were used to something else tracking it for you, and you had to start discovering it from scratch.

While mistakes in aggregate are inevitable, individual mistakes are not, and neither are certain categories of mistakes.

Not paying attention to your airspeed is the sort of mistake you should be able to eradicate.

I want to know why a pilot who says that he is "very concerned" with doing a simple visual approach is allowed to have a couple hundred lives in his hands in an airliner. This is Piloting 101.

Mistakes are inevitable, but we shouldn't just throw up our hands either, especially when this particular mistake comes down to straight-out incompetence and a pilot who's afraid of a basic and straightforward flight operation.

You keep reading an awful lot into a single line, taken from an interview after the accident, possibly with translation and/or language barrier, without even knowing how the question was phrased.

Would you feel better if he said he was "not concerned at all" with landing the airplane?

I'm using that single line as a summary, not the sole evidence.

> [The pilot] "stated it was very difficult to perform a visual approach with a heavy airplane."

From http://www.cbsnews.com/news/ntsb-asiana-crash-pilot-very-con...

This is, of course, complete bullshit.

From the same article:

> Lee said he told his instructors about his concerns in the flight's planning stages. He told investigators that as he realized his approach was off, he was worried he might "fail his flight and would be embarrassed."

More:

> Since “everyone else had been doing the visual approach … he could not say he could not do the visual approach,” according to NTSB

From http://atwonline.com/safety/asiana-777-pilot-sfo-visual-appr...

Would I feel better if he said he was "not concerned at all"? Of course I would. He was performing an easy landing in completely benign weather at an airport with nothing at all tricky or unusual about its approach. There was no reason whatsoever to be concerned.

Yep, and this is backed up by all the other airplanes taking off and landing visually that same day without any issues.

I think the main problem is the more automation still has to be accompanied with more training of the pilots, up until we have a situationally aware computer that can consistently make better decisions than their human counterparts. For safety critical aviation software, I think we are a few decades away from that.