Re: Airbus safety

From:         rdd@cactus.org (Robert Dorsett)
Organization: Capital Area Central Texas UNIX Society, Austin, Tx
Date:         02 Dec 92 03:49:48 PST
References:   1 2 3
Followups:    1 2 3
Next article
View raw article
  or MIME structure

In article <ByL8Hp.LM8@apollo.hp.com> nelson_p@apollo.hp.com (Peter Nelson) writes:
>>Remember that the standard definition of an airline pilot's job is 99.999%
>>pute boredom, and 0.001% pure terror (I forget where this quote came from,
>>and the ratios may be incorrect) - if this is anything like true, maybe
>>human pilots really are on the edge of extinction ?
>
>   _New Scientist_ had an article devoted to this about 3 issues ago.
>
>   Basically they said that as the % of "pilot error" crashes increases
>   we may already be at the point where more lives would be saved by 
>   pilotless airplanes.   

This is certainly a debatable contention.  Airbus certainly seems to believe
it: but it's also in the business of selling products "differentiated" by 
their style of protection.

The reality of the situation is that the safety record has remained pretty
much constant since the late 1970's--note: not the early 1980's, when the 
first automated aircraft were introduced.  It has stabilized at about 1500 
lives per year.  What IS true is that as mechanical failures are isolated
and fixed, the proportion of pilot-induced failures must, necessarily, in-
crease.  The problem facing the industry is how to get the death rate to 
zero: we don't seem to wish to recognize that, with current technology, it
may not be possible, and that we may see steadily diminishing returns in 
our efforts to do so.

The proportion of pilot-error incidents range from 60% to 95% of the total
number of crashes, with Airbus generally supporting the latter figure.  The 
problem, of course, is how one defines *pilot error*.  Is "pilot error" 
pushing the wrong switch?  Suddenly pushing, instead of pulling, on the 
yoke?  A psychotic break?  Naturally, none of these things: in all instances, 
"pilot error" has been a case of a broader *system* failure, the system being 
a combination of the pilot, his peers, the airplane, its interface, the 
airline, and the regulatory backdrop.  In precious few cases were the pilots
"asleep at the wheel," or criminally incapacitated.

What is debatable is how many of these factors can be eliminated, simply
by increasing automation, reducing oversight authority, or transferring 
responsibility for operations to ground controllers.  

It is EXTREMELY important to realize that we're struggling against an *ideal*:
no crashes.  It is also important to note that if, indeed, pilot error is
*increasing*, then it's probably a result of over-automation in the cockpit, 
since virtually no other part of the infrastructure has changed since the 
late 70's.  The simple, short-term solution is to reduce the degree of auto-
mation, or at least bring the pilot back into the loop (not necessarily 
exclusive concepts!).

You would have a hard time convincing me that the number of fundamental errors 
would not increase GREATLY with ground-based oversight, that the safety 
margins would not go DOWN, as people fundamentally distanced from the reality 
of a flight have a go/no-go say.  


>And moreover, the technology to do this either
>   already exists or is close at hand.

The technology isn't close to create safe, fully autonomous aircraft.  
And, in lack of that, we'd need ground-based control, with a high degree of 
automation in-flight.  The infrastructure needed to support this would be 
exhorbitantly expensive (and who would run it: the dispatch controller, who 
just sees a number on a status board, and wants to make his schedule?  A
government specialist?).  In addition, we'd almost certainly be replacing the 
existing social and interface problems that currently exist in the *air*, with 
a new, untried set of problems on the ground.

More than any other trend in aviation, this sort of talk, much of which seems
to originate with Airbus, and which deliberately, blithely underrates the
problems involved in reducing pilot authority, worries me that we've passed 
the point of negative returns.  The problem, again, is not automation: to 
paraphrase Don Norman, it's appropriate feedback.  Or, in mil-speak, the 
minimum capability needed to carry out the mission requirements.  There is 
abundant evidence that, in fact, this requirement can be met with *less* 
automation, *better* interfaces, and keeping the pilots in the loop.  
However, there is also evidence that flight deck design is engineering-and 
marketing-driven, and that "good" human factors does not play a primary role 
in flight deck design, except as a rubber-stamp on a pre-existing systemic 
intent.

There is also increasing evidence that hybrid designs: with high degree of
automation, and relegating the pilot to a passive, supervisory role, out of 
the loop, are *not* the way to go.  


>   They said, however, that it would be a public-relations nightmare and 
>   felt there was no hope of selling the idea to the public.  

Was Bernard Ziegler the author of this article, perchance? :-)  It's 
symptomatic of the technocratic solution: full-speed ahead with quantifiable 
solutions, damn the pilots.  Even if we don't fully understand the 
consequences of the resulting environment, when these solutions have to
ultimately interact with human beings, at least at some level.


>   People will
>   continue to cite those cases where coolness or quick thinking on
>   the part of the crew did save the airplane or at least many lives.  

I wouldn't.  Rather, I would ask how well we understand the *totality* of
in-flight incidents and actions, which are corrected by appropriate air-
manship.  An old, true saying, is that a good pilot is a pilot who doesn't
have to show he's a good pilot.  Is the capability of being able to maintain
control in a thunderstorm really that relevant, when 99% of all pilots would
simply have flown around the same thunderstorm?


We can automate easily quantifiable issues: simple tasks.  Judgement and
airmanship has thus far evaded us, on all levels.  Until we get a grip on
it, talk of fully autonomous aircraft or ground control is nothing more
than science fiction.




---
Robert Dorsett
rdd@cactus.org
...cs.utexas.edu!cactus.org!rdd