Distributed adjudication refuses to let any one authority decide the whole person: institutions get jurisdiction over problems, not ownership over participants. That refusal invites the sharpest objection in all of governance. If every decision can be referred to another body, and every body must itself be authorized, funded, staffed, and reviewed, then who adjudicates the adjudicators? Distributed authority looks like it merely pushes the problem up a level forever — into infinite regress, or into deadlock when jurisdictions collide, or into the quiet capture of whoever was supposed to be independent. This page answers that objection directly: how the regress actually terminates, and how distributed decision-making avoids collapsing into either paralysis or a new sovereign wearing a kinder face.
Operational finality, not final truth
The move that dissolves most of the recursion is a distinction. NWG does not need a decision-maker who is epistemically correct. It needs a process that produces decisions operationally final enough to act on while remaining institutionally revisable. Decision finality is not truth finality. Every governance system eventually reaches someone who says for now, this is the decision — the mistake is letting that become this authority has therefore determined the truth about the participant.
A decision can be actable tonight and still be time-limited, documented, open to review, reversible, compensable if it was wrong, and barred from spreading into domains it was never meant to touch. The framework already reasons this way — it decides provisionally, and its adjudication sequence ends by asking whether a decision can be undone if later evidence proves it wrong. Stated as a principle: some decisions must become final enough for action without becoming final descriptions of reality. There is one carve-out the distinction cannot paper over. Where an action is irreversible in effect — an involuntary hold that costs someone their home, job, or custody; a removal; a shutdown — “reviewed and reversed afterward” is a fiction, because the review arrives too late. There the consequence ladder governs: the greater the irreversible consequence, the higher the bar before any decision may be treated as final at all.
The regress terminates in three levels
The recursion ends because authority is arranged in three levels, not an endless chain of equals.
First, a domain decision: the actor with ordinary, narrow jurisdiction decides — a clinician on a clinical question, a housing body on tenancy, a court on a legal dispute, an administrator on eligibility. Second, an independent challenge: a different institution can examine the evidence, the procedure, the jurisdiction, the proportionality, and any conflict of interest, which is where participant advocacy and appeal sit, and where appeals cross institutional boundaries so that no forum becomes the last word on itself. Third, a constitutional boundary: a final body does not relitigate every fact forever. It asks only the boundary questions — did the deciding institution actually hold jurisdiction, was minimum necessary jurisdiction exceeded, was standing violated, was due process available, has the authority outlived its legitimate duration?
That is what terminates the regress. The constitutional layer decides the boundary of authority, not the substantive truth of every underlying dispute — so it does not need to be an infinitely wise last instance, only a thin one that polices the limits of power. And it is not exempt from what it imposes: it is legitimate only while it, too, remains correctable. The recursion ends not in a perfect institution but in a contestable settlement the ecology keeps re-earning.
Design around error, not ideal judgment
Underneath the three levels is a design stance. Advocates, reviewers, and adjudicators will sometimes be wrong, biased, captured, overloaded, or ideological — so the framework assumes it, and stops asking the question it cannot answer. The question is not how do we find trustworthy adjudicators? but how do we design a system in which an untrustworthy one cannot easily acquire irreversible jurisdiction? The answers are structural: narrow jurisdiction, multiple routes of review, recorded reasons, reversibility, time limits, separation of functions, conflict disclosure, and real remedies for error — the same bounded, provisional, reviewable machinery the model already builds, and the same reason it prefers plurality to one perfect institution. It is the treatment the framework gives AI, applied to human institutions: depend on structure, not on benevolence.
What keeps a single error from becoming a verdict is non-cascading adjudication — the sibling of non-cascading standing. An adjudication binds only its own domain, duration, and parties, so a finding in one place never propagates into the rest of a life: an incapacity to make one particular decision does not settle finances, housing, parenting, or voting, and a housing violation does not become a portable risk score that follows a person everywhere. The system does not have to produce flawless adjudicators if it contains the blast radius of their mistakes.
Polycentric governance has its own failures
More decision centers are not automatically better, and the objection is right to name the pathologies. Each has a structural counter rather than a wish for good behavior.
Deadlock — two legitimate jurisdictions collide, each acts reasonably, and the person is left unhelped between them — is answered by conflict-resolution jurisdiction: a predetermined rule for which concern temporarily dominates when jurisdictions conflict, running from immediate irreversible harm, to basic standing, to existing rights, to ordinary administrative interest. It is not a permanent sovereign; it is a bounded forum for the collision itself. Forum-shopping — seeking whichever decider is likeliest to favor you — is answered by defining jurisdiction beforehand, by issue, with appeal routes fixed in advance, so no one chooses the tribunal after seeing whether they like its answer, while genuine alternative pathways remain open where they protect against capture. Capture of the independent advocate is answered by making independence structural: funding not controlled case-by-case by the reviewed institution, protected tenure, a separate reporting line, record access, conflict prohibitions, plural providers, and participant choice — the test being that the advocate’s survival can never depend on pleasing the institution whose decisions it challenges.
The fourth failure needs its own section, because it is the one that hurts the vulnerable most.
Someone must own the whole failure
Diffusion of responsibility is the polycentric failure that leaves a person on the sidewalk while every institution is technically correct: the police say it is a health matter, health says housing, housing says behavioral, behavioral health says the person refuses treatment, and everyone has partial jurisdiction so no one owns the outcome. The principle no one decides the whole person must never curdle into no one is responsible for whether the person falls through the whole system.
The counter is continuity responsibility: a function that owns the coordination failure — responsible for whether the interfaces between bounded jurisdictions actually connect. Its design is delicate, because it is the very place a coordinator could quietly become the whole-person authority the architecture exists to refuse. It cannot be given jurisdiction over the participant in order to do its job. Its jurisdiction runs over the institutions’ duty to coordinate, enforceable against them — a standing to say collectively, this system is failing this participant, and these bodies are required to resolve the interface, and nothing more. No one owns the person; someone owns the gap.
Emergency compresses; it does not rewrite
Emergencies punish distributed process, because someone is dangerous, or a system is failing, now, and there is no time to consult every forum. The framework already permits emergency authority that decays — a definable trigger, a scope limited to the threat, automatic expiration, documentation of why ordinary procedure was bypassed, mandatory review, and restoration unless the authority is independently renewed. The recursion view adds one line and one limit. The line: urgency may temporarily compress governance, but it may not rewrite the constitution — it concentrates decision-making for a moment without transferring jurisdiction, which is exactly the red line the register keeps against emergency authority becoming permanent. The limit is the finality carve-out again: an irreversible emergency action cannot be made operationally final on compressed process, because there is no later review that can undo it.
Between too much authority and too little
The whole architecture lives between two failures, not one. Jurisdictional concentration — one actor decides too much — is the danger the framework names most often. Its mirror is jurisdictional dissipation — responsibility so distributed that no institution retains enough authority to remedy a recognized standing failure. Both break the same thing: correctability. So “distributed” is not itself the objective; the objective is jurisdiction that is appropriately bounded and sufficiently capable — enough to act, distributed enough to stay correctable. Dissipation belongs on the irreversibility register beside the concentration red lines, as their matched pair.
That boundary also settles which inefficiency to accept. Distributed jurisdiction is sometimes slower, costlier, and duplicative — but appeal only looks wasteful until the first bad decision, independent review only looks redundant until the primary institution is captured, and plural jurisdiction only looks messy until one body tries to own a whole person. That is productive friction, and it is the price of correctability; the framework removes waste, not all resistance. Deadlock is different — friction that prevents action rather than protecting it — and it is a failure, not a safeguard. The open question the architecture can never fully close is how much protective friction an ecology can carry before it can no longer act at all: too much, and governance becomes non-traversable from the inside, paralyzed by its own safeguards rather than captured from without.
What the architecture actually promises
The recursion does not end in a wise final body, and it does not run forever. It ends in bounded domain authority, finite independent challenge, and a thin constitutional layer that governs jurisdiction rather than truth — held open by the fact that it, too, can be corrected. What NWG asks of a decision system is not that it be right, but that it stay bounded, answerable, and reversible enough that no single error and no capture can become permanent.
No one decides the whole person. Someone stays responsible for the whole failure. And no authority — not even the one that judges the others — gets to place itself beyond correction.