Category: Systems Alignment & Culture
Few concepts have travelled as far, or been as thoroughly softened in transit, as psychological safety. The term originates in the organizational research of Edmondson (1999), who defined it with some precision as a shared belief that a team is safe for interpersonal risk-taking, and it entered mainstream managerial consciousness when Google’s Project Aristotle identified it as the factor most strongly distinguishing the company’s high-performing teams (Duhigg, 2016). The concept has undergone a quiet transformation in the course of this journey. Psychological safety has come, in its popular usage, to mean something closer to emotional comfort — a warm climate, a feeling of being cared for, the absence of stress. Leaders speak of creating a safe space as though safety were an atmosphere they might generate through a softening of tone and an accumulation of good intentions. This is, I want to argue, a consequential misreading, and correcting it changes what a leader is actually obligated to do.
Return to Edmondson’s definition and attend to what it does not say. It does not say that people feel comfortable. It does not say that the environment is low in stress or free of conflict. It says that people believe they can take specific interpersonal risks — speaking up, asking a question, admitting an error, challenging a decision — without incurring specific negative consequences such as humiliation or punishment. Safety, on this definition, is not a feeling at all. It is a prediction, an assessment that people make about what will happen to them if they speak. Predictions, unlike feelings, are built from evidence. This is the crux of the matter: people do not come to feel safe because they have been told they are safe. They conclude that they are safe, or that they are not, by observing what actually happens to those who take risks. Psychological safety is therefore a system condition rather than a mood, because the system — its behaviors, its consequences, its processes — generates the evidence from which the prediction is formed.
The question that ought to occupy any leader, then, is what evidence the system is generating, and here three structural sources prove decisive. The first is leadership behavior under pressure, and in particular the behavior that follows bad news. Each time a person brings forward a problem, admits a mistake, or challenges a decision, the leader’s response constitutes data — not only for that person but for everyone observing the exchange. A leader who meets a surfaced error with curiosity and a turn toward problem-solving generates evidence of safety; a leader who meets it with blame, even occasionally, generates evidence of danger that no quantity of stated openness can overwrite. It is worth being honest about the asymmetry here: people weight the punishing response far more heavily than the reassuring word, because the cost of being wrong about danger is high. A single public humiliation teaches a lesson that a hundred invitations to speak freely cannot un-teach.
The second source is the accountability structure, and specifically whether consequences are predictable and proportionate. Psychological safety is frequently and mistakenly opposed to accountability, as though a safe environment were one in which standards had been relaxed. The opposite is the case. Edmondson (2018) situates psychological safety and accountability on two distinct axes, locating the high-performance zone where both are high. Arbitrariness is what corrodes safety, not accountability — consequences that are unpredictable, disproportionate, or unevenly applied. The rational response, when people cannot anticipate how the system will respond to a given action, is to minimize their exposure: to say less, to risk nothing, to withhold difficult information for as long as possible. The third source is policy and process — the formal architecture that determines whether speaking up has a channel and whether that channel connects to any outcome. A team may possess a genuinely warm interpersonal climate and remain unsafe in the consequential sense, if no actual mechanism exists through which a raised concern travels to a decision. A demanding and high-intensity environment may conversely be genuinely safe if it is equipped with robust and trusted processes for surfacing and addressing what is not working. The feeling follows the structure in each case, and not the reverse.
The reason this distinction is practical rather than merely conceptual is that it dictates what a leader does. A leader who believes psychological safety to be a feeling will attempt to produce it through affect — through encouraging language, team-building exercises, statements of values, an open-door policy announced but not structurally supported. Such efforts are not without worth, but they are insufficient, and when they fail to move the condition they are intended to move, the leader is apt to conclude that safety is soft and unmeasurable and to turn elsewhere. A leader who understands psychological safety as a system condition works instead on the system: examining how leaders actually conduct themselves when bad news arrives, and altering that conduct because it is the loudest signal the system emits; auditing the accountability structure for predictability and proportionality, because arbitrariness is the quiet destroyer of candor; and building and sustaining real channels through which concerns travel to decisions, while making the connection between voice and outcome visible, because people calibrate their willingness to speak against the evidence that speaking matters. This is the principle that runs through the whole of the work: the conditions an organization experiences as feelings — safety, trust, morale — are downstream of structure. One cannot exhort one’s way to psychological safety. One can only build the conditions under which safety becomes the rational conclusion for a person deciding, in the middle of a meeting, whether it is worth the risk to raise a hand.
