What is institutional debris?
Institutional debris is the record of what people did to keep a formal system functioning: workarounds, overrides, tickets, grievances, and exception codes.
The ideas the standards rest on, one page each: what the term means, a made-up case where it matters, what to look for in a real system, and which standards use it.
In the order the argument runs: where the evidence of an error collects, who can challenge it, what authority has to be renewed, and how the repair is recorded.
Institutional debris is the record of what people did to keep a formal system functioning: workarounds, overrides, tickets, grievances, and exception codes.
The people who absorb a system's errors are the first to see failures and the last able to act, unless they have standing to challenge the system.
AI contestability is the ability to challenge an AI or other automated decision and get a review by someone with the authority to change it.
A grant of authority is held for a term against a renewable justification, not owned outright.
A delegation lets a system decide for an organization. It is justified only while its permission, its evidence, and the capacity to correct its errors all still hold.
A repair log records each repair made after a system harms someone, what it cost the people repaired, and the receipt and owner for each.
Ethotechnics takes no side on open or closed weights. It classifies a deployment by nine properties and governs the assembled system, not the model.
Technical, operational, and institutional reversibility are separate claims, and only the third decides whether a correction will ever be used.
A typed decision model returns a choice and a confidence, not text. It is governed like any system allowed to decide, and the standards name six places its authority sits.
Ethical interrupts are deliberate checkpoints that pause automation when risk thresholds are crossed.
A burden index measures the minutes, retries, and documents a system demands from people who comply, challenge, or recover.
A consent journey maps how consent is requested, renewed, revoked, and recorded over time.
The permission surface is the set of authority grants in force, each with an evidence basis, a scope, a state, and an expiry.
A safety valve routes cases to a logged fallback path, often human review, when automation is overloaded, uncertain, or error-prone.
An intervention specification replaces “human in the loop” with six answerable questions about who can change the system.
Design authority names the accountable owner who can approve, change, or halt a system’s design.
Stoppability is the ability to halt an automated process: for the person inside it, without penalty; for the institution, by a named person, in a set time.
Time to halt (TTH) is the time between a request to halt an automated system and the moment it stops acting.
Arguments that need more than a definition, a checklist, and phrasing to reuse.
Governance measured by what a running system can be made to do: answer a challenge, change the rule, halt, and reverse.
A dated log of how governance terms drifted into labels, with definitions to cite instead.
The eight mechanisms a remedy process needs, grouped by when each one acts.
Seven pass/fail questions to run over a decision flow before it ships or when it is audited.
Questions a person can send when an automated decision is made about them, one ask each.