In the early 1990s, a second-year PhD student set out to answer a question that seemed almost too simple to be interesting… Do better teams genuinely make fewer mistakes?
What ensued was six months of collecting data across a group of hospitals, the measures being team quality on one side and medication errors on the other. The hypothesis seemed a foregone conclusion, better teams would have fewer errors.
And yet the reality was to the contrary. The highest-performing teams had the highest reported error rates.
"Maybe the better teams don't make more mistakes," Amy Edmondson realised. "Maybe they report more mistakes."
From that sentence came the concept of psychological safety, now one of the most cited ideas in modern management research. It is also one of the most misunderstood ideas in safety.
The underlying insight was easy to miss, despite being what mattered most. It wasn’t that better teams were safer due to making fewer errors. It was something else. Better teams were in fact safer because the truth about what was going wrong was reaching someone who could act on it. The difference was visibility, not the error rate.
And that brings us to the crux of this article; is the truth about what's happening on your site actually reaching the people who can do something about it before it’s too late? It’s a question most safety programmes quietly fail, not because their people don't care, but because the way organisations are run tends to keep hard truths from travelling.
Pixaera is built on this very idea, the notion that safety training should build the judgement to act. not just prove a briefing happened.
Amy Edmondson is the Novartis Professor of Leadership and Management at Harvard Business School, the author of The Fearless Organization and Right Kind of Wrong, and was ranked the world's number one management thinker by Thinkers50. She joined Mousa Yassin on the Safety Heroes Podcast to trace those discoveries back to their origins, and to be blunt about what the safety profession still gets wrong about it.
It’s not about being nice
Many organisations start in the wrong place, and this is a good starting point.
"Psychological safety is a belief that speaking up is safe," Amy explains. "That you can speak up with ideas, questions, concerns, mistakes. Not that it's easy or fun to do that, but that you believe it's expected and welcome."
Expected and welcome. Not comfortable. Not easy.
The most common failure mode she sees is organisations hear "psychological safety" and immediately think softness, lowering the bar, avoiding the hard conversation, and protecting people from difficult feedback.
"Psychological safety is not about being nice or soft," she says. "That's actually disrespectful, the tyranny of low expectations. Excellence comes when you have both candor AND high standards."
This distinction matters more in heavy industries. These are often sites where nobody speaks to you about what's actually happening. These sites can be dangerous. And this is the first condition for truth to travel: people have to believe surfacing it is expected of them. If that belief is missing, the flow of information stops long before it reaches anyone who could have used it.
What vulnerability actually means
A leadership buzzword, Amy has little patience for the version of “vulnerability” that means oversharing.
"Vulnerability is honesty," she says. "And therefore it's strength, because you're courageous and confident enough to tell the truth about reality so that we can work on it together."
I've never run a project like this before. I need your input. I'm not sure how to do this, can you help? The realities of it in practice are unglamorous and specific.
If psychological safety is the condition for truth to travel, this is the act. Someone has to actually say the true thing, and the thing standing in the way is often us.
"We are our own worst enemy," Amy says. "We puff ourselves up, we try to look good like I never made a mistake in my life. And that makes us actually quite unappealing."
Three kinds of failure, and only one you should want more of
While the majority of organisations run just a single failure policy, Amy argues there are three distinct kinds. Treating them the same is why so many learning programmes stall.
First are basic failures; single cause, familiar territory, usually human error. Examples could be texting while driving or the wrong ingredient in a known recipe. These are preventable, and we should be intolerant of them.
Then there are complex failures; system breakdowns. Several small shortcuts and deviations aligning at the wrong moment. These can be collapses in the supply chain, pandemic responses. These must call for vigilance, not blame.
Finally there are intelligent failures. "The undesired result of a thoughtful foray or experiment in new territory." There are four conditions to this: there's no existing solution, you're pursuing a real goal, you've done your homework, and you've kept it appropriately small. Scientists, she notes, fail for a living.
Confusing these three is costly in both directions. Punish an intelligent failure and innovation stops. Tolerate a basic failure and someone gets hurt.
The red/green ratio
This was one of the most immediately valuable ideas in the conversation, and it doesn’t take long to apply.
In your next meeting, listen to the balance. How much of what you're hearing is green; good news, on track, agreement? And how much is red; concerns, dissent, bad news, I'm not sure?
"If you're hearing an awful lot of green rather than red, something's wrong," Amy says. "We're living in a complex, ambiguous world. How likely is it that everything's green? Not very."
Contrary to conventional wisdom, an all-green meeting is not evidence of a healthy operation. It's evidence that the red isn't reaching you and the comfort of an all-green report should worry you. It's the same comfort a completion certificate gives you in that it’s proof of a process running, not proof that anyone is actually ready.
Why blame kills learning
Amy picks up a point made by Todd Conklin, who appeared on Safety Heroes earlier this season, on what blame actually does to an investigation.
"As soon as there's blame, it shuts down the learning," she says, "because everybody else is suddenly off the hook and they no longer have to reflect on what they did or might have done."
This is the real cost. Blame doesn't just harm the person carrying it, it excuses everyone else from thinking.
This is the heart of Human and Organisational Performance, the principle a growing number of the best safety teams now build around: you cannot learn from an event and blame someone for it at the same time. The two are mutually exclusive. The moment an investigation lands on who, it stops being able to answer why, and why is the only thing that transfers to the next site, the next crew, the next shift. Blame is not just unkind. It is the most effective way there is to stop the truth from travelling.
Pause, breathe, choose curiosity
When asked for the one thing she'd want a safety leader to take away, Amy didn't reach for a framework.
"Knowing is the automatic state. I look around and I think I see reality. I don't, I see reality filtered through my own biases. When I override that sense that I know and instead choose curiosity, it's more joyful, I'm more interested in what's happening around me, and it promotes quality, safety and all sorts of other good things. Pause, breathe and choose curiosity. I think that's the most important choice we make."
While this might sound small, in a high-hazard environment, it is often the difference between walking onto a site to confirm what you already believe and walking on to find out what's true.
And this is the actual work of a safety culture.
Not producing reports, recording attendance, or proving a briefing was delivered. These things measure whether a process happened rather than if the truth is moving. The organisations that get this right aren't the ones with the fewest failures on paper, they're the ones where the truth arrives before the incident does. That was the finding hiding inside a PhD student's messy data thirty years ago, and it's still the one most of the industry hasn't fully acted on.
So pause, breathe, and ask the only question that matters: is what's really happening on your site reaching the people who can do something about it? And if you're honest, how would you know?
Three questions for your team this week
- In our last leadership meeting, what was the red/green ratio? If it was all green, who wasn't speaking?
- Of the last five failures we responded to, which were basic, which were complex, and which were intelligent? Did we respond differently to each?
- When did a leader here last say "I don't know how to do this" out loud, and what happened next?
If that last question is a hard one to answer, it's the reason we built Pixaera; decision-led training that shows you who's actually ready, not just who showed up.