[error_logic]

As your sibling comments mentioned, there's a difference between giving a chance for someone to learn from a single mistake without punishment, and allowing them to make the same mistake twice without taking matters out of their hands after.

If it's a really critical role, the training will have realistic enough simulation for them to make countless mistakes before they leave the training environment. Then you can assess their level of risk safely.

[hinkley]

This whole thread is missing the fact that the NTSB had a theory that transparency leads to safer airplanes, they tried it, and it works. People hesitate to self-report when it comes with punishment (fines, demotions, or just loss of face among peers). You need a formal “safe space” where early reporting is rewarded and late reporting is discouraged.

Safety is a lot about trust, and there is more than one kind of trust. At a minimum: are you capable of doing this thing I need you to do? Will you do this thing I need you to do?

[zettabomb]

It's not just the NTSB, it's part of things like the Toyota Production System. There's ample evidence to show both that punishment discourages safety and that lack of punishment encourages safety, across multiple industries.

[nyrikki]

Yes this is cross industry best practices.

Goodhart's law also applies, as in the case of the edoor bolts, Spirit intentionally bypassed safety controls to meet performance metrics.

The Mars Climate Orbiter is another example. While unit conversion was the scapegoat, the real cause of the crash is that when people noticed that there was a problem they were dismissed.

The Andon cord from the Toyota Production System wasn't present due to culture problems.

Same thing with impact scores in software reducing quality and customer value.

If you intentionally or through metrics incentivize cutting corners it will be the cost of quality and safety.

I am glad they called out the culture problem here. This is not something that is fixable under more controls, it requires cultural changes.

[StableAlkyne]

> The Mars Climate Orbiter is another example. While unit conversion was the scapegoat, the real cause of the crash is that when people noticed that there was a problem they were dismissed.

Challenger too. Multiple engineers warned them about the O-rings. They weren't just ignored, but were openly mocked by the NASA leadership. (https://allthatsinteresting.com/space-shuttle-challenger-dis...)

A decade later a senior engineer at NASA warned about a piece of foam striking Space Shuttle Columbia and requested they use existing military satellites to check for damage. She was ignored by NASA leadership, and following (coincidentally) a report by Boeing concluding nothing was wrong, another 7 people were killed by a piss-poor safety culture. (https://abcnews.go.com/Technology/story?id=97600&page=1)

[ethanbond]

But but but what about my intuition and gotcha questions about how this could never work in practice?

[hinkley]

The dirty secret of why traffic circles reduce accidents? Stoplights feel safer than they actually are, while circles feel more dangerous than they actually are. That nervousness becomes vigilance, which reduces accidents. It’s also why people intuitively hate them. They’re actually right, but also wrong.

Feeling safe is an illusion that governments try to maintain for their people. It’s one of their biggest jobs. But the illusion has dimensions and it’s hard to keep several going at once.

[s1artibartfast]

I think there is more nuance to it than that. Not everything is a mistake, not every mistake is recoverable, and not all skills are trainable.

The fundamental goal is to distinguish between recoverable errors and those that are indicative of poor employee-role fit.

[nyrikki]

Mistakes are the problem, as they will always happen.

The point is to build a culture where you value teamwork and adjust and learn from failures.

This isn't an individual team problem, this is an organization problem.

It is impossible to hire infallible, all knowing employees.

But it is quite possible to enable communication and to learn from pas mistakes.

When you silence employees due to a fear of retribution bad things happen.

People need to feel safe with calling out the systemic problems that led to a failure. If that ends up being the wrong mixture of skills on a team or bad communication within a team that is different.

Everything in this report was a mistake, and not due to gross incompetence from a single person.

The E door bolts as an example was directly attributed to metrics that punished people if they didn't bypass review. The delivery timelines and defect rates were what management placed value on over quality and safety.

Consider the prisoner delema, which is resolved by communication, not choosing a better partner.

[s1artibartfast]

I don't disagree with what you said about this instance, but I'm trying to push back on the knee jerk sentiment that there are no bad employees only bad systems- There are both. cultures that are too permissive of bad actors degrade the system.

Part of maintaining quality culture is maintaining red lines around integrity.

Like I said above, not all errors are recoverable or honest mistakes.

I work in medicine and a classic example would be falsifying data. That should always be a red line, not a learning opportunity. You can add QA and systemic controls, but without out integrity, they are meaningless. I have seen places with a culture of indifference, where QA is checked out and doesn't do their job either.

[filleduchaos]

> I work in medicine and a classic example would be falsifying data

Certainly nobody has ever thought about that before. In fact, there definitely isn't a second sentence in the definition of aviation's just culture that is being completely ignored in favour of weird devil's advocacy.

> 4) Just Culture- errors and unsafe acts will not be punished if the error was unintentional. However, those who act recklessly or take deliberate and unjustifiable risks will still be subject to disciplinary action.

Oh wait.

[s1artibartfast]

I have no problem with the stated safety culture.

I simply agree that "that everyone has the right to work every job" is not a reasonable interpretation of them.

as stated above, a reasonable reader should understand:

> Not everything is a mistake, not every mistake is recoverable, and not all skills are trainable. The fundamental goal is to distinguish between recoverable errors and those that are indicative of poor employee-role fit.

[filleduchaos]

Who is claiming that "everyone has the right to work every job", though? The only person to even bring up the sentence is someone who's handwringing about an interpretation that nobody was making to begin with.

This is why I called it weird devil's advocacy, because what exactly is the point of jumping to caution people about something they aren't doing?