4 ms·

If you think that's the entire premise, you didn't read much of the article. Keep scrolling, it's a well well reasoned discussion of why Malcolm Gladwell's comm

If you think that's the entire premise, you didn't read much of the article. Keep scrolling, it's a well well reasoned discussion of why Malcolm Gladwell's commentary on previous KA crashes was misinformed and possibly even distorted to prove his point. Gladwell's book was the source of the "Korean culturalism is incompatible with emergencies" myth you cite. > With Airplane pilots, Crashes are so exceedingly rare, and the backups, both mechanical, human, and training so well designed to prevent them Air travel is exceptionally safe, but it falls prey to the law of large numbers: 99.9% success still means some people die every year when there are thousands of planes in the air. This has everything to do with individual decision making on the part of the pilots, which will necessarily be faulty occasionally. It's basically just random chance that some planes every year will crash because of pilot error. > the issues associated with a crash are typically systemic, and little to do with an individual. What sounds better in an NTSB report: "the pilot fucked up and flew into the ocean, whoops" or "we recommend adjusting the height of the foo so the pilot can better see it while barring"? Obviously everyone wants actionable outcomes from post-crash investigations, even when they're just trying to explain away a random accident. Y

gizzlon 13y ago
Air travel is exceptionally safe, but it falls prey to the law of large numbers: 99.9% success still means some people die every year 99.9% success would mean ~475 plane crashes during take-off and landing on Heathrow alone each day =)
- gizzlon 13y ago
  Wait, that can't be correct, can it? :) Was going off this: http://en.wikipedia.org/wiki/World%27s_busiest_airports_by_aircraft_movements http://en.wikipedia.org/wiki/World%27s_busiest_airports_by_a... Edit: Silly me, the stats are by year not by day as I was expecting. Guess it means only ~1.3 crashes at Heathrow each day
- protothomas 13y ago
  I doubt there are 475000 take-off and landings per day from Heathrow despite how busy it is!
ghshephard 13y ago
re: "Entire Premise" - sorry, I didn't mean to say that was the "Entire premise" of the article, I was just trying to say that his "entire premise regarding the golf game analogy" was off. Landing Planes is systemically guaranteed to be safe, it's no longer a function of individual performance, whereas there is nothing systemic about golf games - it's all about individual performance. The OP was trying to compare and contrast a "System"'s behavior with an "Individual's" behavior, which was where the article went awry. Re: Gladwell - Gladwell was most definitely not the source of "Korean Culturalism" regarding Air Safety. He simply wrote a popular chapter on an issue that already well documented. The cause of any pilot error is not law-of-large numbers - pilot error has been, for the most part, systemically eliminated from flying. What's left now is system-flaws. One of the obvious flaws at SFO was it's GlideScope Indicator was out of service. I have to believe that this will turn out to be a major contributor to the accident, and, in the future, Airport's may be much more cautious about disabling such safety equipment unless absolutely unavoidable.
alphaBetaGamma 13y ago
> It's basically just random chance that some planes every year will crash because of pilot error. When you design a system where errors can cause a large number of deaths (airplanes, trains, nuclear reactors,...), your objective is to have very few accidents (nobody can guarantee zero). Having someone to blame when an accident happens -- the pilot did not follow procedure -- is not enough. Your system design must be human error tolerant, because you know these will happen. To take an extreme example, if a drunk pilot error causes an accident it's not only his fault, but there probably also is a systemic error which caused this criminal behavior (which is a possibility you expect) to cause an accident: maybe the cockpit culture was such that the co-pilot was frightened to stand up to the drunk pilot.
- alanctgardner2 13y ago
  I guess my point was that airline safety hits a sort of horizontal asymptote as you put in effort. You'll go years and years without accidents, have one, and then struggle to generalize a rule from this incredibly freak event. My biggest complaint is that the whole 'East Asian culture causes deference to authority, crashes planes' argument is a huge generalization with very little data to back it up. In any other context, the number of events we're talking about would be anecdotal.
mikeash 13y ago
Actually, air travel doesn't really fall prey to the law of large numbers. Air travel in the US is so safe that it's actually getting difficult to measure it and difficult to identify problem areas. The last fatal airline crash before Asiana 214 was years ago. Some people are most emphatically not dying every year anymore, despite thousands of planes in the air. We go years at a time without a single airline fatality in the US. And yet, those few accidents that do happen all have identifiable causes with clear remedies. This Asiana crash may have been pilot error, but it was definitely not just random chance. Pilots letting their airplane get too low and too slow on landing with no adverse factors is a clear failure of training and proficiency. One fairly obvious remedy for this one would be to require all airline pilots to hand-fly a purely visual approach at least X times per year to ensure that the skill stays fresh. I dare you to find a first-world airline crash from the past couple of decades that has no actionable outcomes from the investigation. Shouldn't take you too long to go through all the crashes, since there won't be many to look at in the first place. I guarantee you that every one had actual, useful recommendations come out of the accident investigation, not just bureaucratic CYA "recommendations".

gizzlon 13y ago

Air travel is exceptionally safe, but it falls prey to the law of large numbers: 99.9% success still means some people die every year 99.9% success would mean ~475 plane crashes during take-off and landing on Heathrow alone each day =)

gizzlon 13y ago
Wait, that can't be correct, can it? :) Was going off this: http://en.wikipedia.org/wiki/World%27s_busiest_airports_by_aircraft_movements http://en.wikipedia.org/wiki/World%27s_busiest_airports_by_a... Edit: Silly me, the stats are by year not by day as I was expecting. Guess it means only ~1.3 crashes at Heathrow each day
protothomas 13y ago
I doubt there are 475000 take-off and landings per day from Heathrow despite how busy it is!

ghshephard 13y ago

re: "Entire Premise" - sorry, I didn't mean to say that was the "Entire premise" of the article, I was just trying to say that his "entire premise regarding the golf game analogy" was off. Landing Planes is systemically guaranteed to be safe, it's no longer a function of individual performance, whereas there is nothing systemic about golf games - it's all about individual performance. The OP was trying to compare and contrast a "System"'s behavior with an "Individual's" behavior, which was where the article went awry. Re: Gladwell - Gladwell was most definitely not the source of "Korean Culturalism" regarding Air Safety. He simply wrote a popular chapter on an issue that already well documented. The cause of any pilot error is not law-of-large numbers - pilot error has been, for the most part, systemically eliminated from flying. What's left now is system-flaws. One of the obvious flaws at SFO was it's GlideScope Indicator was out of service. I have to believe that this will turn out to be a major contributor to the accident, and, in the future, Airport's may be much more cautious about disabling such safety equipment unless absolutely unavoidable.

alphaBetaGamma 13y ago

> It's basically just random chance that some planes every year will crash because of pilot error. When you design a system where errors can cause a large number of deaths (airplanes, trains, nuclear reactors,...), your objective is to have very few accidents (nobody can guarantee zero). Having someone to blame when an accident happens -- the pilot did not follow procedure -- is not enough. Your system design must be human error tolerant, because you know these will happen. To take an extreme example, if a drunk pilot error causes an accident it's not only his fault, but there probably also is a systemic error which caused this criminal behavior (which is a possibility you expect) to cause an accident: maybe the cockpit culture was such that the co-pilot was frightened to stand up to the drunk pilot.

alanctgardner2 13y ago
I guess my point was that airline safety hits a sort of horizontal asymptote as you put in effort. You'll go years and years without accidents, have one, and then struggle to generalize a rule from this incredibly freak event. My biggest complaint is that the whole 'East Asian culture causes deference to authority, crashes planes' argument is a huge generalization with very little data to back it up. In any other context, the number of events we're talking about would be anecdotal.

mikeash 13y ago

Actually, air travel doesn't really fall prey to the law of large numbers. Air travel in the US is so safe that it's actually getting difficult to measure it and difficult to identify problem areas. The last fatal airline crash before Asiana 214 was years ago. Some people are most emphatically not dying every year anymore, despite thousands of planes in the air. We go years at a time without a single airline fatality in the US. And yet, those few accidents that do happen all have identifiable causes with clear remedies. This Asiana crash may have been pilot error, but it was definitely not just random chance. Pilots letting their airplane get too low and too slow on landing with no adverse factors is a clear failure of training and proficiency. One fairly obvious remedy for this one would be to require all airline pilots to hand-fly a purely visual approach at least X times per year to ensure that the skill stays fresh. I dare you to find a first-world airline crash from the past couple of decades that has no actionable outcomes from the investigation. Shouldn't take you too long to go through all the crashes, since there won't be many to look at in the first place. I guarantee you that every one had actual, useful recommendations come out of the accident investigation, not just bureaucratic CYA "recommendations".