The Essence of Safety podcast

the word fin is written on a black background . ALT: the word fin is written on a black background .

That’s it – the end of my very first thread.
Did you like it? What are your thoughts?
I’d love to hear from you.

Asking those questions reveals the deeper story: about systems, conditions, decisions, and pressures.
And that’s the insight we need if we want to prevent future accidents.

Instead, finding human error is the starting point.
The real questions are:
Why did the error occur?
Why this time – and not the countless other times when everything went fine? Or did it occur previously, but didn't cause an accident? Why not? That's how you learn about the accident causes.

So where does that leave us? How do we deal with human error in an investigation?

One thing is crucial:
Human error must never be the *ending* point of an investigation.
If you are an investigator and your only conclusion is that human error occurred, you haven’t done your job. At all.

I’ve heard of surgeons saying a *good* surgeon can still operate flawlessly after a 24-hour shift (i.e you need to be 'good," then you can do it). Most pilots would never consider flying after being awake that long, because they know that fatigue will affect them, no matter how "good" they are.

Human performance is shaped by many factors: fatigue, design, commercial pressure – just to name a few.
This doesn’t mean humans have no agency, but it does mean they’re not as fully in control as we like to think. As an example:

So the interesting question is *not* whether there was human error – but:
Why did it lead to an accident THIS time, and NOT on previous times?
Why did things go right all the other times, until they didn’t?

And one more thing: you’ll always find human error.
Todd Conklin, a safety guru, once said: “Looking for human error is incredibly boring – because it’s always there.”

Every system has human involvement. And wherever humans are involved, some form of error is inevitable.

Nothing was technically “wrong,” but today, that design would probably never be certified.

Design matters. Context matters. Humans make errors, but those errors are shaped by the systems they're working in, and how tolerant the systems are to human error.

A famous example: During WWII, US B17 pilots often mixed up flap and gear levers, causing accidents. Were they poorly trained? Overwhelmed? Or is there another explanation?

As if it turned out, the flap and gear switches looked alike, felt alike, and were close together.

Engine failures shouldn't happen – and crews train for them.
So was it technical or human error? The answer: both. You can't reduce it to one aspect. Any accident is usually both.

You also can't understand human error without understanding the technological and sociological context.

A common misunderstanding: that you can clearly separate technical failure and human error. People think it's either one or the other. But reality is messier.
Imagine an airliner suffers an engine failure during takeoff. The crew mishandles it and the plane crashes.

Sometimes, the same crew can perform worse than expected and then better than expected – even on the same flight.
Example: Air Astana 1388. They failed to detect the flight control issue before departure – but later managed to regain control in flight, against all odds.

One key concept in understanding human performance is human performance variability. Human performance isn't steady – it can be worse than expected (many accidents show that) but also much better than expected (think Sioux City or the Hudson ditching).

Please note that this thread is not going to be about suicidal actions and mental health. These are important topics, but they deserve their own discussion.

A pictogram symbolising human error, using the head of a person and an aircraft

Recently, the topic of pilot error / human error / human factor has come up a lot again, and I thought it might be useful to start a 🧵 summarizing how modern safety science views the human role in safety-critical systems – and particularly in accidents. It's going long, but worth it, I promise!

Take an engine failure on take-off resulting in a crash. Is it the crew's fault because they should normally be able to handle it, or is it the fault of the engine?

Any issue that can't be covered reliably by crew actions should normally result in urgent modifications of the entire (ie worldwide) fleet.

Mind you, it is almost never as black and white as in "it is either the crew's fault or it is technical fault."

AVHerald update on AI171:

"It is becoming [...] clear [...] that there was probably 0 negligence in the cockpit [...]. The probability of a technical cause is high."

avherald.com/h?article=528f…

FEHLER EINS - Folge 47 - Unfall Vnukovo FEHLER EINS - Daily Error · Episode

Eckhard Jann and I sat down and talked about the 2014 Falcon 50 accident at Moscow-Vnukovo that killed Total CEO Christophe de Margerie [in German].

open.spotify.com/episode/5upw...

Nearing the end of the final report on Aeroflot 1492, I think I'm close to a full understanding of why this crash happened.
I have a record 30 pages of notes, and I still need to read the Aeroflot and Rosaviatsiya dissenting opinions and the MAK's responses before I assemble the final text. (1/2)

A must-read. Yet another instance in which the non-public military accident investigations uncover troubling technical problems and the public report just blames the pilot.

Medevac Learjet 55 crashes near Philadelphia | Flightradar24 Blog A Learjet 55 air ambulance crashed shortly after departure from Northeast Philadelphia Airport on 31 January.

The NTSB released its preliminary report on the crash of a Medevac Learjet 55 last month. Of note, the NTSB says, “The CVR did not record the accident flight and ... it was determined that the CVR had likely not been recording audio for several years.”
www.flightradar24.com/blog/flight-...

TONIGHT! Free #aviation lecture, RAeS Solent Branch, 5 March, 'Accident Investigation - The Human Perspective' #avgeek buff.ly/Wliyx8A

Today I learned that Ron Schleede died this month. He was a well-known air crash investigator that appeared on many Mayday episodes. My condolences to his family and all of his friends and colleagues.

*I'm planning a podcast episode, not an article 😄

Excellent article! I am planning an article on commercial pressure (and how hard it is to prove that it played a role in an accident), using the Polish Presidential flight and the accident of a Gulfstream at Aspen as case studies. I may include the Gulfstream accident you mentioned, too!

“The role of the operator in an accident is to add the final garnish to a lethal stew whose ingredients have been long in the cooking” No one has ever said it better.