Everyone Claims to Support Psychological Safety. Almost No One Creates It.
I've sat in the retrospective. I've filled out the anonymous survey. I've heard a VP say, in a room where everyone was absolutely certain what the VP wanted to hear, "I really want your honest feedback."
Nobody said anything real.
Psychological safety surveys beautifully. It lives badly.
Amy Edmondson at Harvard Business School defined it in 1999 as "the belief that one won't be punished or humiliated for speaking up with ideas, questions, concerns, or mistakes." Google ran Project Aristotle in 2016 — one of the most intensive team effectiveness studies ever conducted — and landed on the same finding: psychological safety was the number one predictor of high-performing teams. Not skill, not composition, not clarity of goals. The shared belief that you could say something risky without getting punished for it.
That belief is earned through experience. Not through policy. Not through a survey. Not through a quarterly retrospective where the manager is in the room, visibly hoping nobody says anything that requires them to actually change something.
The conditions that make psychological safety real require leaders to absorb costs. Most don't.
Why Psychological Safety Initiatives Fail Even When Leaders Mean Them
The problem isn't bad faith — or at least, not usually. Most leaders genuinely believe they want challenge. They've read the research. They've attended the training. They understand, conceptually, that a team that can tell them bad news early is more valuable than one that hides it.
But psychological safety isn't about what leaders believe about themselves. It's about what their behavior teaches their team to believe.
Watch what happens in practice. A leader says "I want you to push back on my ideas." Then someone pushes back. And the leader — not angrily, not dramatically — just doesn't love it. A slight tightening. A fast pivot. The challenge gets acknowledged and then gently rolled over. The person who raised it notices. Everyone in the room notices. The lesson lands: the invitation is real; the welcome is not.
Edmondson's research frames this around interpersonal risk — the calculation each person makes before speaking. Will this cost me? Every day, in every meeting, teams are running that calculation based not on HR policy but on what they've actually seen happen to people who spoke up. You don't convince a team they're safe by telling them they're safe. You convince them by being consistently, visibly worth the risk.
That consistency is expensive. Most leaders aren't willing to pay.
Why Anonymous Surveys Don't Create Psychological Safety
The most common psychological safety intervention is some version of creating a structured opportunity to speak: the anonymous survey, the team retrospective, the "speak-up session," the skip-level meeting. These have value. But they have a predictable ceiling.
The ceiling is this: structured opportunities to speak don't change the underlying belief. They paper over it.
If a team has learned — through experience — that raising problems gets minimized, that challenging a decision earns quiet irritation, that bad news travels up slowly because good news travels up fast, then giving them a designated channel to share feedback changes the channel, not the belief. They'll use the channel carefully. They'll soften the edges. They'll pick which truths to send.
Anonymous surveys are the most honest version of this problem. When a survey has to be anonymous for people to tell the truth, you've measured the absence of psychological safety, not the presence of it. That's useful data. But it's not the intervention.
Running a follow-up survey to check if last quarter's survey improved things is not a feedback loop. It's a ritual. The ritual produces a score that can go in a deck. The deck goes in a QBR. And at the next retrospective, the VP says they really want honest feedback.
What Leaders Who Actually Build Psychological Safety Do Differently
The research on what actually works is less flattering than most leaders would like.
Edmondson's own follow-up work points to one consistent pattern among leaders whose teams demonstrate genuine psychological safety: they model vulnerability first, and they absorb the cost of bad news without transferring that cost to the messenger.
Modeling vulnerability first means this: before asking your team to share uncertainty, you share yours. You say out loud when you don't know something. You announce when you changed your mind, and why. You tell the story of a time you were wrong in a way that's specific enough to be credible and public enough to be witnessed.
This is uncomfortable. It feels like weakness. It isn't — but it feels that way, which is why most leaders skip it and go straight to asking others to be vulnerable first. That sequencing never works. You're asking for a thing you haven't demonstrated you can do safely yourself.
The second piece — absorbing the cost of bad news — is harder still. If someone brings you a problem early, when it's still fixable, the correct response is visible gratitude. Not tolerance. Not a polished "thanks for raising that" before moving on. Actual, active reward: this is exactly what I need, this is why I want you in the room, this helps us. The messenger should walk away feeling like they did something valuable, not like they got away with something.
That's the turn-toward response instead of the turn-away. It's a daily micro-decision, not a program. You can't schedule it. You can't delegate it. You do it or you don't, in small moments your team is watching more carefully than you think.
The Real Cost of Building Psychological Safety on a Team
Here's the part the organizational behavior literature is polite about: building psychological safety requires leaders to absorb costs that are genuinely uncomfortable, with payoffs that are genuinely slow.
Modeling vulnerability means accepting a reduction in perceived authority in the short run. Rewarding bad news means managing your own emotional reaction in real time, which takes energy — especially when the bad news is about something you care about. Changing your mind publicly means accepting that people noticed you were wrong. These aren't trivial.
And unlike most leadership behaviors, the payoff is invisible for a long time. A team that gradually starts trusting that it's worth the risk to say something hard — you don't see that in next quarter's numbers. You see it eventually, in the quality of the problems you know about before they become crises. In the ideas that actually get surfaced instead of staying in someone's head on the drive home. In the meeting where someone finally says the obvious thing nobody wanted to say.
Most leaders will nod at all of this and then run a retrospective anyway. The retrospective is easier. It looks like action. It produces a score.
The real thing is available. That's what I keep coming back to. Psychological safety isn't a myth, and the research isn't wrong. The teams that have it are different — measurably, functionally different — from the ones that don't. Google's Project Aristotle found that, Harvard confirmed it, and anyone who's been on both kinds of team has felt the difference without needing a study to name it.
But the path to it runs through the leader, and specifically through what that leader is willing to do daily that costs them something. The survey can't shortcut it. The program can't shortcut it. The initiative with the name and the deck and the quarterly cadence cannot shortcut it.
There's a parallel in how emotional availability gets misunderstood more broadly — the language sounds like care, but the structural conditions that would make care real are exactly what nobody wants to fund. Psychological safety runs on the same logic. The language of openness is cheap. The daily practice of absorbing the cost of it is not. Most organizations pay for the language.
The team that can tell you the truth is the most valuable team you'll ever lead. Most leaders spend their careers explaining why they can't quite build one.
Photo by Tima Miroshnichenko on Pexels