I'm in the final stretch of my ATP rating, I’ve been thinking a lot how we can maximize survival and minimize the likelihood of becoming another aviation news headline.
I'm now convinced that we have the wrong mental model. Pilots like the simplicity of categorizing safety by powerplant: “Piston engines fail every 1k–10k hours.” “Jet engines are good for 100,000 hours, and with two of them you will never have to worry about it.”
Unfortunately, the real world is very messy, very gray, and very conditional.
A much better mental model that will keep you alive is to completely throw away those statistics and build a model for every flight you take. That’s much closer to reality.
The conditional probability of any given flight varies greatly. Some flights have a 1-in-10-million survival rate—you would have to repeat that flight 10 million times to have an incident. Some flights, if we’re intellectually honest, are probably terribly low odds, like 1 in 100. Repeat that flight 100 times and you will have a serious accident once.
Imagine flying a plane you’re not intimately familiar with yet, night time, IMC, cold front coming in so you need to fly now or not at all (any further delay on the ground does in fact increase the risk). You know the autopilot is not working - known squawk, one of the fuel gauges is sticky and one of the MP gauges is sticky. But it’s only 80 nm, you know both airports, you’ve been averaging 50 hours a month, and you’re super current and confident.
We’ve probably all been there and either made that flight or not made that flight. If you’re a serious pilot you come across these situations.
If you were to do an honest calculation of the risk of that flight, it’s insanely high. Stack two or three failures into that flight and it ends in a news headline. The likelihood of two or three failures on a flight like that? Probably high. Probably 1 in 100.
You take off normally, and 20 minutes into the flight your panel goes black. Shit. The alternator must be out again, and the battery was old and only held power for 20 minutes. Now you’ve got no comms, no PFD, you’re in the soup, hand flying, fumbling for your iPhone to try to get some light on the panel (it was only a 25-minute flight after all - you didnt think about losing power on this of all flights). You don’t know the EPs in this airplane well—doing a manual gear deploy (electric motor won’t work) while hand flying in IMC is going to be damn hard. All of your past experience and recent currency is no good. This might be your last flight.
The risk asymmetry in aviation is much, much greater than our human meat computers appreciate. Our cognitive capabilities are outstanding. Listen to Beethoven's 9th and final symphony for a concrete example. On the other hand, our ability to multiply multiple low probability events is terrible.
Train yourself to become computer-like at risk analysis, like an insurance model. Don’t come up with a binary “am I good to make this flight?” Come up with a numerical risk score like the one above: how many times can I repeat this flight with no incident?
Once you have a number , you can do a lot with it—and cancelling the flight is just one outcome. The risk mitigation part is way more interesting. You can start to think about “how do I move this from 1 in 100 to 1 in 10,000?” In the scenario above, if you fly it day VFR you can now easily handle two or three stacked failures incident-free, just by tweaking the timeline of the flight by 10 hours.
If we do the correct risk accounting—and use rigid numerical scores instead of emotional PPL-style “does this pass my personal minimums (Y/N)”—not only do we achieve much lower accident likelihood, we reduce or eliminate those hours sitting in the airplane wishing we were on the ground instead.
The problem with personal minimums is twofold:
1. The more you fly, the more you reduce the minimums until they’re basically gone, and you’re flying approaches to 200ft ceiling. This is great for skill increase and utility, but it's terrible risk mitigation—you keep cranking up the risk vector. Precisely what good risk mitigation should prevent.
2. Personal minimums are binary. Either you do or you do not make the flight. You don’t do the mental accounting for risk mitigation.
The objective of a well built risk-mitigation model isn't to "make or not make" the flight. It's to make the flight, but dramatically improve safety through deliberate risk mitigation before the flight starts.
I think the FAA is somewhat naïve here. They could do a much better job teaching this style of risk analysis. Their approach is to make you memorize acronyms like PAVE and IMSAFE and remind you about hazardous attitudes, but none of those lessons are highly actionable. They're shallow. They don’t get into the very interesting, unique nuances we encounter in real weather flying in these amazing machines.
The airlines solved this in a very different way. They don’t allow there to be any risk. Part 121 is amazingly safe not just because the pilots are exceptionally well trained, but because the *system* itself is designed to eliminate risk. The airplanes can’t fail, and the airspace/runway system is designed not to fail. Of course I’m exaggerating a bit—there are still failures and risk, —but the spirit of what I’m saying is the truth.
In Part 91 and Part 135, pilots have a lot more freedom. The system and aircraft are a lot more flexible to allow the pilot to make many more decisions. It offers amazing flexibility, but also ends up in unfortunate headlines and loss of precious life.
Comments
Post a Comment