Intel's Ronler Acres Plant

Silicon Forest

Tuesday, September 15, 2015

Airline safety

Airliner burns at Las Vegas airport last week.
I asked a question on the Wikipedia Reference Desk:
I have always heard that "flying is safer than traveling by car" on a per mile basis, and I believe that is true. But what about on a per trip basis? Say I make about two car trips a day (I drive to work and then I drive home). In 40 years that could add up to 25,000 trips. I've had two accidents that were serious enough to make my car undrivable. I don't know how many airline flights I've taken, not many, maybe 100. My point is all you can do about these kinds of things is you can either decide to make the trip or not, so the odds of an accident on a per trip basis are more meaningful then on a per mile basis. — Preceding unsigned comment added by (talk) 17:55, 12 September 2015 (UTC)
I got back a couple of replies. I thought this one was especially good.
Indeed; ultimately, what we choose as our denominator when we normalize statistics for comparison is purely a heuristic. Should we compare number of trips? Number of miles? Number of passengers multiplied by number of miles? Number of dollars spent? Whichever value we choose represents a heuristic model of our threat: if we count miles, that implies that (for some reason) we believe the risk is uniformly distributed over distance; if we choose number of trips, that suggests that we believe the risk is quantized - e.g., because of the adage that the riskiest part of a flight is the takeoff and landing. There is no universally correct answer.
I feel safer when I fly myself than when I fly on a commercial airline, even though accident statistics very clearly show that airline transport is safer than general aviation. This may be an illusion, but it is a real psychological effect. The process of Aeronautical Decision Making is the "systematic approach to the mental process of evaluating a given set of circumstances and determining the best course of action." It means becoming fully informed about all the pertinent objective facts of a situation. It also means to think clearly through the details, and to be aware of our own mental limitations. Finally, it means to take a reasonable course of action, based on all available data.
It is also worth emphasizing the formal distinction between probability and statistics, because this is relevant to making good decisions about outcomes.
In my short career as a (non-commercial) aviator, I've seen many mechanical failures - on the ground and in the air. Thus far, none of these interesting occurrences have resulted in a fatal accident. But from this perspective, I recognize the widespread fallacy of fixating on Gaussian distributions. The mean and median event rate for a large population has zero impact on when I will experiece an event. This is the causation-correlation fallacy. It is tragic that in the basic tiers of formal schooling, we spend so much time studying bell-curves for large populations. We ought to spend more time studying Poisson distributions and their effect on probability. Bell curves are fantastic ways for institutional regulators to study safety on the macro-scale, and do provide actionable information if your decisions can affect large numbers of events. However, I am only one individual. I do not represent 300 million air travelers; I do not actually feel effects of n-accidents-per-hundred-million. What I care about is likelihood of a single event - one single event, not n-events-per-mille - and all I care about is how that single event will affect me (and my aircraft and my passenger). Recognition, and realistic understanding, of these types of probability distributions, is more useful for me to inform my judgement than all the bell-curves in the world. My aircraft will not suffer an accident because the national average for aircraft predicts it. My aircraft will only suffer an accident if a mechanical or systems failure occurs, or if there is a fire, or if I command the aircraft to do something unsafe, or if some outside occurrence creates an unrecoverable situation. None of that is caused by nation-wide average. Quite the opposite: the nation-wide accident rate is caused by the aggregation of all of these individual events.
This perspective completely changes the way I evaluate, and reduce, my risk - whether the risk is related to automobile traffic, operating or riding in aircraft, or participating in any of the other uncertain activities of ordinary life.
Nimur (talk) 15:44, 14 September 2015 (UTC)

No comments: