Originally Posted by
silverstrata
The problem is rarely the computer. The problem (and this applies to all industries) is most often the computer-human interface.
In trying to cope with so many variables the flight computer gets ever more complex. But where the computer has to interface with the human, things have to be simplified down to basic commands. This difference in capability means the human operator has no chance of understanding what the computer is trying to do, behind the scenes; and so if there are any i/o errors or hardware malfunctions, there is no chance a human can instantly trouble-shoot the problem.
The answer is not to go back to basic stick and rudder flying. The answer is to have more redundancy in the inputs, and not allow the computers to be disconnected. (Just, as you would not dream of disconnecting the flight control hydraulics.)
I think that using the hydraulics as an analogy is not quite correct. Hydraulics have no power of decision they do what they are told and without a computer interface that is precisely what the pilot makes them do. The computers do precisely what the systems analyst and software designer wanted them to do when the pilot makes inputs and the environmental conditions are in particular states as sensed through various replicated sensor systems.
The crew will quite often be in the position of asking "what is it doing now? why did it do that? or what is it going to do next?" usually with the occasional
thrown in for emphasis.
The systems analyst, software designer and acceptance engineers often ask: "What did they do that for? Who would have thought a pilot would do that? or Surely the winds/temperatures can't do that?" and other comments as they have to consider every possible combination - not every likely combination but all of the unlikely ones too. To avoid getting too complex, they have an 'otherwise case'. So if the software finds things are not working the 'otherwise case' is to fail and hand the bag of bolts to the pilot. The pilot doesn't have to be a magenta line follower to be surprised by this and however good cannot know everything the computer was doing and why it was doing them. If possible the pilots are expected to keep, or get, the aircraft straight and level then try to solve the problems. This may not be easy if they have totally lost situational awareness and the computer system just dropped them the bag of bolts.
All this was known and discussed many years ago - see
http://csel.eng.ohio-state.edu/produ..._surprises.pdf Automation Surprises. (I made a similar comment on the Tech Log).
It is a pity that the pilot is not only not 'IN the loop', or 'ON the loop' but is often well 'OUT of the loop' when it comes to the automation and what and why it is doing things. It takes a level of concentration from the pilot to retain situational awareness, but if the computer system doesn't communicate - like an introverted PF with a steep authority gradient in CRM - it can lead to problems when things go awry.