A Manifesto for Institutions You Can Stop

A debt letter arrived naming a sum the recipient had never spent: a year’s income, averaged across the weeks when there was no income, and the difference called a debt. To escape it, the recipient had to disprove the government’s arithmetic, document by document. The machine that sent the letter was never asked to prove its own sum.

That letter is the definition of domination in miniature: a decision the institution made, a correction it refused to carry, and a person left to make up the difference. Dependence on powerful institutions is permanent, and a politics that promises escape isn’t serious. The aim is interdependence in which neither side subordinates the other.1

The argument rests on five propositions.

The first is that human capacity is finite. Time, attention, energy, health, patience and the ability to adapt are scarce, and none of them is free infrastructure.

The second is that any collective arrangement needs adaptation. Institutions act through categories, records, models and rules, all partial, and someone has to carry the remainder. Human adaptability may be the subsidy that lets rigidity survive.2

The third is that power shapes who supplies that adaptation. Part of power is the ability to make others adjust while you stay the same. Responsibility should track authority, and by default the institution that benefits from a simplification bears its cost. Until you know who made up for what an institution left undone, you can’t credit it with competence or trust the numbers it publishes about itself.3

The fourth is that domination is the power to leave your errors on other people and refuse their corrections. It becomes durable when authority over an arrangement, adaptation pushed onto others and insulation from the consequences come together. Then the people carrying the costs can’t force a change.

The fifth is that freedom needs enough capacity left over, after necessary coordination, to have a life substantially your own. Part of freedom is what it costs to say no, and part is a working route to make an institution that misreads you change course. On this account, an institution has lost its claim to legitimacy once none of the people it governs can make it undo a mistake.4

The political question underneath all five is who gets to spend whose life.

What is new here

None of the pieces is. Organizational sociology, fiduciary law, standpoint epistemology and aviation safety each supplied one. The combination is new, and so is the definition. Locating domination in the refusal of correction makes exception handling, usually written off as administrative friction, the main way modern power extracts unpaid adaptation. Robodebt looked efficient because the people it misread were correcting it for free.5

Where the framework sits

The established theories each locate power somewhere. This framework keeps those locations and adds a question they leave open. When the abstraction collides with reality, which side has to move?

Two institutions with identical databases and categories can differ completely in power, because the direction of correction differs when model and reality come apart. When people manage to conform, that shows how much they can absorb and nothing about whether the norm was justified.

Where difficulty lives

Every institution puts the difficulty of its work somewhere, either in its own coordination, memory and safeguards or in the lives of the people it serves. People who piece together fragmented records and remember what the institution forgets are doing its processing in the wrong place. The line where its failure becomes their problem can be moved.6 Redesign moves the function to the institution and checks that the compensation hasn’t reappeared elsewhere. Avoidable friction also decides which people get through a process and, where exported costs stay invisible, which organizations win.7

An institution that can’t be corrected still fails. Its mistakes put wear, repair work and lost time on the people they land on, and that metabolic debt never shows up on its books.

Standing

Every institution makes mistakes, so the design question is where the error goes next. In a corrigible arrangement it becomes a challenge that can climb from the record and the decision to the rule, the model, the incentives and finally the institution’s claim to that power.8 Standing is the hinge. It turns what someone lived through at the edge of the rules into evidence the institution has to admit, and lets the person the error landed on be the one who detects it. Rights, appeals, unions, judicial review and independent expertise all let someone contradict the institution with consequences, and each is judged by whether it keeps the institution open to correction.9 Corrigibility is the capacity of the people a system gets wrong to make it change course. It still lets a matter be settled enough to act on, without making the institution the last word on what’s true.

Diagram

The closed circuit and the open circuit

View source

%% title: The closed circuit and the open circuit
%% caption: When the error stays with the people it lands on, their adaptation reads as stability and the model confirms itself. When they can contest it, the error comes back as evidence and an independent review can force a fix.
graph TD
    subgraph Closed["When the error stays with the people it lands on"]
        C1[Formal Model] --> C2[Unresolved Residual]
        C2 --> C3[Crumple Zone]
        C3 --> C4[Compensatory Adaptation]
        C4 --> C5[Apparent Stability]
        C5 -->|the model confirms itself| C1
    end

    subgraph Open["When the people it lands on can force a fix"]
        O1[Formal Model] --> O2[Unresolved Residual]
        O2 --> O3[Situated Signal]
        O3 --> O4[Enforceable Standing]
        O4 --> O5[Independent Diagnostic Review]
        O5 --> O6[Upstream Recalibration]
        O6 -->|Closed Loop| O1
    end

What knowledge does not license

Governing at scale needs abstraction and expertise, but knowledge does not confer jurisdiction. Capacity, knowledge, usefulness, dependence and existing authority create no further jurisdiction on their own, and every extension of authority needs its own justification.10

New capability tests that rule by dissolving protections that came from inability, as privacy held while surveillance was expensive. Dependence tests it too. The usual conversion runs from “you need us” to “we require more discretion over you,” and it should run the other way. The more a society depends on an infrastructure, the more that infrastructure owes in reliability, due process and repair, and the less discretion it should hold.11

The same limit protects a person. A model may know things about someone that she doesn’t know herself, and its accuracy gives it no authority over her.12

The remedies all take one form: a legally binding stop on an institution’s power to push its costs onto others, pulled by the people who would otherwise absorb them.

Follow the burden

The method runs the theory in reverse, rebuilding an institution from its failure boundaries. Follow the appeal, the workaround, the shadow spreadsheet, the unpaid administrative labor and the person who knew the official category was wrong. The official record shows what the institution thought it was, and the residual shows what it took to make it work.13

At each boundary, two questions find domination. The first locates it where a contradiction gets resolved. The second catches the loop in which people’s adaptations become training data for governing them further.

When the institution and reality stop fitting, which one has to move?

Does the weaker party’s successful accommodation become permission for the stronger party to demand still more?