A hospital disciplines a nurse. She administered a drug the system says she should have caught. The allergy flag sat in a scanned note the pharmacy module never read. The unit was running two staff short, as it had run for a year. The committee that decides her case can discipline a nurse. It cannot change the staffing rule that put her there, the procurement rule that bought a pharmacy module that cannot read scans, or the design rule that let the allergy live in a document no system ingests. Every level of that stack could have prevented the error. Only one level can be blamed for it, because only one level is a person.
Nobody on the committee is cruel. The cruelty is in the arithmetic, and the arithmetic runs unspoken: responsibility pools where it is affordable. It settles on the employed, the insured, the credentialed, the one with a badge to take away. Authority pools somewhere else entirely — in the rules nobody on the committee can reach. The two quantities never meet. That is the pattern, and it is so ordinary that the committee's verdict will not even feel like a decision.
The principle
Responsibility is legitimate only where the authority lived: the power to have altered the outcome, at the time it was still alterable, with what the person actually knew.
State it that way and the nurse's case dissolves before it starts. She could not have altered the staffing. She could not have altered the ingestion pipeline. She knew what her shift actually contained, which is to say she knew it the way people know a flood is rising — as a condition, not a choice. The only decision in reach was administration, and administration of what the system itself supplied. Strip her of her license and the stack that produced the error is exactly as capable of producing it again, minus one competent person.
The principle is not a call for mercy. It is a constraint on allocation. Where authority is divided, responsibility must divide with it. Where authority was absent, responsibility may not appear at all — not softened, not shared, not assigned "for accountability's sake." Blame assigned where power never lived is not accountability. It is the disposal of accountability.1
Why the blame lands low
If blame were assigned by authority, it would travel upward, and institutions cannot survive that. So it is assigned by survivability instead. The person with the least exit takes it. The committee blames the nurse because the nurse has no committee.
You can watch the sorting rule run in every direction. The patient who missed appointments under a booking system that books nothing is noncompliant. The caseworker whose quota makes careful work impossible is underperforming. The clinician who cuts corners in a unit that scheduled the corners away is negligent. In each case the institution found the one participant who could absorb the verdict without changing the system, and gave it to them.2
This is why the crumple zone and the scapegoat are the same position. The moral crumple zone — Madeleine Clare Elish's term — names who absorbed the force of a failure. Responsibility without authority names why the force went there: blame follows the crumple zone because the crumple zone is who the failure could be given to, and authority never was. Both describe a single act of allocation — the institution handing its failure to the part of the stack that can hold it.
The forms it takes
The misallocation rarely looks like blame. It looks like responsibility generously conferred.
There is the committee that "provides oversight" — performing the rituals of responsibility without the power to force a sacrifice, so that when the dangerous thing proceeds anyway, someone was officially watching it. There is the testimony that is fully believed and binds nothing, which grants the speaker recognition while declining them any share in the decision their words are supposed to govern.3 There is the metric that professionals are scored on but cannot contest, so that they answer for outcomes a formula assembled out of their reach.4 And there is the quiet repairer: the caseworker whose competence keeps the numbers true, with nowhere to put what she catches. Her absorbed corrections are the audit trail of an authority gap. Every repair she makes in silence is a decision that had no owner.5
Each of these confers the costume of responsibility on someone whose authority over the outcome is zero. Each then uses the costume as proof that the outcome was governed.
The rule
An institution may not assign responsibility for an outcome until it can name the decision. Which decision could this person have made, with what information, against what alternative, at what cost — and was that cost one an institution had any right to require? If the answer to the last question is that saying no would have destroyed them, the institution has not found the responsible party. It has found the one that could not refuse.
The rule bites in both directions. A decision made with full authority is fully owned, even by people who turn out to be right about that — authority is not exculpation for being wrong, and "the system made me" is not an excuse for the person who had the pen. What the rule removes is the free transfer of ownership downward. What it protects is the ability of anyone to establish, with evidence, that the decision was never theirs — and institutions resist exactly that, because an audit that establishes where authority actually lived is an audit of everything they preferred not to count.6
The rule also ends an inquiry that ought to stay open. Blame is how an institution declares its model of a failure complete. Once the nurse is named, the staffing rule stops being examined, the procurement rule stops being examined, and the system that produced the error keeps its model intact by spending a person. An institution that assigns responsibility without authority has not corrected anything. It has purchased the appearance of correction, and the purchase price is whoever was cheapest.7
What survives
The alternative is not a blameless institution. The blameless institution is the fantasy of every system that prefers its errors unexamined. The alternative is an institution whose blame lands where decisions are made — which is also the only institution whose corrections can land there, because correction and blame obey the same rule. You cannot route the fix to someone who lacked the power, and you cannot honestly route the fault to them either.
A hospital can build that. It has to know which rule set the staffing, which rule bought the module, which rule let the allergy live unread, and each of those rules has to have an owner whose ownership is real — someone who can be answerable for it, and whose answers can change it. That is harder than disciplining a nurse. It is the difference between an institution that learns from its failures and one that processes them.
Notes
The ethics that performs responsibility without the power behind it — the oversight that watches so the institution can proceed watched — is the subject of Toothless Ethics: Why Principles Don’t Stop Machines.
How the sorting rule finds the participant who can absorb the verdict without changing the system: You Were Free to Choose.
Testimony that is believed and binds nothing: Institutions Decide Who May Speak First.
The metric the scored cannot contest is one of the instruments in Dependence Without Domination.
The quiet repairer, and the audit of who an institution actually depends on: The Metric She Repairs Says She Is Not Needed.
An audit that establishes where authority lived is resisted for the same reason a valid finding binds nothing unless something makes it bind — the response function of the challenged institution is the subject of Stop the Machine, Not the Person.
What a blame-that-ends-the-inquiry protects, and what a valid finding must be able to do to it: The Rule Is Never on Trial.