The Wrong Question
For decades, the dominant question in organizational psychology was: what makes some people more effective than others? Hire better people, train better people, manage better people — the unit of analysis was always the individual.
Google's People Operations team started from the same assumption in 2012. They called the project Aristotle, after the philosopher's claim that the whole is greater than the sum of its parts. Their hypothesis was that the best teams would be made up of the best individuals — that excellence was additive.
After three years studying 180 teams across Google, they concluded the opposite. The composition of the team — who was on it — was far less predictive of performance than the norms the team had developed. And the single most important norm, the one that explained more variance in team performance than any other factor, was psychological safety.
The discovery was not new. Amy Edmondson at Harvard Business School had identified the concept and its effects in 1999. What Google did was replicate it at scale, with their own data, in a highly competitive context where the incentives to find a different answer — one that pointed at individual talent rather than social dynamics — were substantial.
They found what Edmondson found.
What Psychological Safety Actually Is
Psychological safety is not comfort. It is not the absence of conflict. It is not a management style or a personality trait.
Edmondson's definition is precise: psychological safety is a shared belief that the team is safe for interpersonal risk-taking. The belief is collective, not individual — it is a property of the group, not of any one person. The risk is interpersonal — speaking up, disagreeing, asking what might look like a stupid question, admitting a mistake, proposing an unconventional idea.
Teams with high psychological safety argue more, not less. They surface more problems, give more critical feedback, and challenge each other's assumptions more directly. The safety is not that nothing is challenged. The safety is that the act of challenging does not carry a social penalty.
The opposite — what happens in psychologically unsafe teams — looks like smoothness from outside and functions like suppression from inside. People don't challenge ideas that seem wrong. They don't report problems that might reflect poorly on them. They agree in meetings and disagree in hallways. The official narrative of the team diverges from its actual state. Information that leaders need to make good decisions doesn't reach them because the people who have it are afraid to share it.
This is not a personal failing. It is a rational response to real incentives. When speaking up leads to punishment — being dismissed, being labeled a troublemaker, being passed over for promotion — silence becomes the rational choice. The suppression is structural before it is psychological.
The Counterintuitive Finding
Edmondson's 1999 paper originated in a study of medical teams. She was measuring medication error rates across teams and expected to find that better teams made fewer errors. What she found was the opposite: the teams she independently rated as higher-performing reported more errors.
Her initial interpretation was that she had the causation backwards — maybe teams that made more mistakes performed worse, which is why they were rated lower. But the performance ratings were independent of the error data. The better teams weren't making more mistakes. They were reporting more.
Lower-performing teams were suppressing error reports. Higher-performing teams had norms that made reporting safe — which meant problems were caught and corrected rather than hidden and compounded. The measurement revealed the suppression: what looked like better performance in lower-performing teams (fewer reported errors) was actually worse information flow leading to worse outcomes.
This is the core mechanism that makes psychological safety matter at scale. The most dangerous problems in complex systems — medical, financial, technological, organizational — are the ones that don't get reported. Errors that are reported are errors that can be fixed. Errors that are hidden compound.
Every major organizational failure in recent decades follows this pattern. The Boeing 737 MAX disasters: engineers who had concerns about MCAS suppressed them or were ignored. Theranos: employees who understood the technology didn't work couldn't safely surface the information. The 2008 financial crisis: risk analysts at multiple institutions identified problems that were not passed up the chain. Challenger and Columbia: engineers at NASA and Thiokol flagged safety concerns that were overridden by organizational pressure.
These were not primarily technical failures. They were psychological safety failures.
Project Aristotle: What Google Found
Google's study was not a controlled experiment. It was an observational study of existing teams — which means causation is harder to establish than in a laboratory setting. But the pattern across 180 teams was clear enough that the Google People Operations team treated it as decision-relevant.
The research team tried to explain team performance with structural variables first: team size, tenure, performance level of individual members, mixing of introverts and extroverts, colocated vs. distributed teams, gender composition. None of these explained much.
They then turned to group norms — the unwritten rules that govern how a team operates. They found five factors that distinguished high-performing teams:
- Psychological safety — Can team members take risks without fear of punishment?
- Dependability — Can team members count on each other to do quality work on time?
- Structure and clarity — Are goals, roles, and execution plans clear?
- Meaning — Is the work personally meaningful to team members?
- Impact — Do team members believe their work matters?
Psychological safety was listed first and ranked highest. Not because the others don't matter — they do — but because safety is the prerequisite that makes the others accessible. Teams can have clarity, meaning, and impact and still perform poorly if members cannot speak up about problems. Safety is what allows the other four factors to function.
Replication: The Google finding has been replicated across contexts:
- A 2017 meta-analysis by Frazier et al., covering 136 studies and 23,000+ participants, confirmed that psychological safety is positively related to team learning behavior, team performance, and employee voice.
- Studies in healthcare contexts (Nembhard & Edmondson, 2006) found that psychological safety predicted whether nurses would speak up about patient safety concerns — with direct implications for patient outcomes.
- Research in software development contexts (Edmondson & Lei, 2014) found that psychological safety mediated the relationship between team structure and software delivery performance.
- The DevOps Research and Assessment (DORA) organization's annual State of DevOps Report has found psychological safety to be among the top predictors of elite software delivery performance across years of data from tens of thousands of practitioners.
This is not a Google-specific finding. It is a finding about how groups function.
The Manager Lottery
Here is the problem with the current dominant approach to psychological safety: it treats the manager as the variable.
Most organizational interventions around psychological safety are aimed at managers. Train managers to model vulnerability. Train managers to invite dissent. Train managers to avoid punishing bad news. Create "leader standard work" that includes regular check-ins and feedback sessions.
These interventions can help. But they create what Edmondson herself has called the "manager lottery" — whether a team has psychological safety depends almost entirely on whether they happen to have a manager who creates it. The individual who wins the lottery with a skilled, psychologically aware manager gets safety. The individual who gets the wrong manager does not. The outcome depends on chance, not structure.
The data supports how variable manager quality is. Gallup's State of the Global Workplace report has consistently found that managers account for 70% of the variance in employee engagement. This is usually reported as evidence that managers matter. It is equally evidence of a structural problem: that the single most important variable in team dynamics is the individual judgment of one person, who was likely promoted for technical competence rather than social skill.
There is a better approach. Build the safety into the structure.
Cooperative Structures Create Safety Architecturally
The Supersociety thesis makes a specific claim: cooperative organizational structures — shared ownership, distributed authority, transparent governance, participatory decision-making — create psychological safety as a structural property rather than a personal one.
Here is the mechanism:
Shared ownership removes the primary source of interpersonal risk. In a conventional firm, the fundamental power dynamic is: management can fire employees. This creates a structural asymmetry in which speaking up against management's decisions carries real career risk. The individual calculus is rational: what do I gain from speaking up versus what do I risk? In many cases, staying quiet is the individually rational choice even when it is collectively harmful.
In a cooperative where workers are owners and governance is democratic, the power to dismiss does not rest in one person's hands. The risk calculation changes. Speaking up about a problem is not speaking up against someone who can end your livelihood — it is speaking up within a governance system where your voice has formal standing. The structural change modifies the risk calculus, which changes the psychological response, which creates safety not as a cultural aspiration but as a structural consequence.
Transparent governance creates information symmetry. When decisions are made in closed rooms by people whose reasoning is not visible, speaking up is risky in two ways: you don't know what you don't know (what factors went into the decision you're questioning?), and your challenge may be overridden for reasons that are never disclosed. Transparent governance — where proposals are public, reasoning is documented, and decisions are recorded — removes both of these risks. You can engage with the actual reasoning. You can understand why your challenge was accepted or rejected.
Participatory decision-making legitimizes disagreement. When you have formal standing to participate in decisions, disagreement is not insubordination — it is participation. The structural difference is significant. In a conventional firm, an employee who repeatedly challenges leadership decisions is often labeled as difficult or not a culture fit. In a cooperative or participatory structure, that same behavior is called doing governance. The label changes because the structure assigns legitimacy to dissent.
Distributed authority eliminates single points of failure. The manager lottery exists because psychological safety is entirely dependent on one person. Distributed authority — where different decisions are made by different people or groups with appropriate expertise and stake — means that a bad actor in one part of the governance structure cannot make the entire organization psychologically unsafe. Safety is decentralized.
The Evidence from Cooperative Workplaces
The theoretical argument for structural safety is supported by data from cooperative workplaces.
Mondragon Corporation, the Basque worker cooperative with 80,000+ members, conducts regular workplace surveys that consistently show higher reported psychological safety and lower fear of reprisal than Spanish private sector benchmarks. The mechanism is structural: workers elect their managers, not the other way around. The power dynamic is reversed.
John Lewis Partnership — a UK employee-owned retailer with 80,000+ partners — reports in its annual partner survey that partners feel substantially more able to raise concerns than employees at comparable conventional retailers. The 2024 Partner Survey found 78% of partners agreed they could speak up without fear of negative consequences, compared to a 58% average for UK retail workers in Gallup's engagement data.
Buurtzorg, a Dutch home care organization that operates as self-managing teams of nurses with no middle management, has demonstrated both higher patient satisfaction and higher nurse satisfaction than conventional home care organizations. A 2015 KPMG analysis found Buurtzorg's administrative costs were 40% lower than comparable organizations — not despite the flat structure but because of it. Nurses who are trusted with full decision-making authority do not spend energy managing up.
These are not isolated anecdotes. They are consistent with the Edmondson/Aristotle finding, approached from the other direction: if psychological safety predicts performance, and cooperative structures create psychological safety, then cooperative structures should produce better performance outcomes, especially in knowledge-intensive work where the quality of information flow matters most.
What High Psychological Safety Actually Looks Like
It is worth being concrete about the indicators, because "psychological safety" has become a corporate buzzword that frequently obscures what it actually describes.
In a team with high psychological safety:
- Team members finish each other's sentences and argue openly in meetings, then make a collective decision and commit to it
- People say "I was wrong about that" without visible distress
- Questions that might seem naive are asked without apology and answered without condescension
- Problems are surfaced early, when they are small
- Bad news travels up without distortion; leaders learn about problems before they escalate
- Experiments that fail are treated as data, not as career-damaging events
- New members are explicitly told what questions are encouraged and what the team's norms around disagreement are
In a team with low psychological safety:
- Meetings are consensus-producing machinery; real disagreements happen in hallways and Slack DMs
- Problems are held privately until they are unavoidable
- Questions imply competence, so they are not asked
- Failures are attributed to external causes or to individuals who can be blamed
- New members learn quickly which topics are safe and which are not by watching what happens to people who raise them
- The gap between what is said in meetings and what people actually think is wide and growing
The latter description fits most large conventional organizations more accurately than leaders in those organizations want to believe.
Where This Points
Google's Project Aristotle found what Ostrom's commons research found: the structure matters more than the individuals. The design of the institution shapes the behavior of its members more reliably than individual traits, training, or motivation.
This is not a pessimistic finding. It is an empowering one.
If psychological safety depends on having the right manager, you are at the mercy of the manager lottery. If it depends on the right structural conditions, you can build those conditions. If cooperative ownership, distributed authority, transparent governance, and participatory decision-making create those conditions architecturally, then the path from "team that relies on a skilled manager to feel safe" to "organization where safety is structural" is a governance design problem, not a cultural one.
Culture is downstream of structure. The culture of open communication and honest disagreement in high-performing cooperatives is not an accident of personalities. It is the predictable result of building organizations where speaking up is structurally encouraged rather than individually risky.
The Supersociety is built on this premise. Not as aspiration — as architecture.
The question for any cooperative institution is not "do we want psychological safety?" The answer to that is obvious. The question is: what decisions do we make in the open? Whose voice has formal standing in governance? How is authority distributed so that no single person can make the whole system unsafe by punishing the people who disagree with them?
Get the structure right. The safety follows.