Taxonomy / Contestatory constitutionalism / Reporting deployed failures into shared catalogs

Reporting deployed failures into shared catalogs

Opens a route for someone who was not in the room when the system was built to report that it failed in the world, and builds the plumbing that carries the report to whoever can act on it: standard machine-readable formats with rules of engagement, disclosure programs borrowed from bug bounties with legal safe harbor for the reporter, and a conduit that fans a single submission out to the several developers, coordinators and registries it concerns. Design frameworks in this group also cover schemes where the submitter is the deployer rather than an outsider, specifying enforcement, anonymity and what happens after a report. At the far end sit open, searchable archives of real-world failures and the editorial problem of indexing incidents whose cause, severity or technical detail is unclear.

The method, against Contestatory constitutionalism

Scroll the diagram sideways to see all of it.

Concept Analysis: Theoretical Foundations

Each concept is read twice: whether the approach carries it, and whether the approach's own sources claim it. A concept that is absent and was never claimed is a gap in the field rather than a failure of the work, and is marked out of scope.

Contestability

PartialClaimed · partial

Definition · Philip Pettit, On the People's Terms (2012)

Anyone subject to a decision must have a real route to challenge it — to inspect the grounds it was made on, to contest the outcome, and to have that challenge answered. Legitimacy is not settled once by an election or an act of consent; it is held open by that route staying available. The test of a free system is not that its rulers are kind, but that their decisions can be contested by the people they affect.

Analysis

A route exists and runs, which is more than the rest of this cell manages: FLARE-AI takes one submission and pushes it to several developers and registries, and the AI Incident Database has archived more than 1,000 reports. What it is not is a route for the person a decision fell on. Reports come from researchers who found the flaw, the grounds on which a system acted appear nowhere in the file, and no recipient owes an answer, so remediation is something the papers hope to accelerate rather than something anyone can demand.

Non-domination

PartialClaimed · partial

Definition · Philip Pettit, Republicanism (1997)

Freedom here is not the absence of interference. You are unfree if someone holds unchecked power over you — even if they never misuse it and mean well. A benevolent master is still a master, and so is an authority whose workings you cannot see. Legitimacy requires that power be checked, not merely well-intentioned.

Analysis

The strongest republican element here is the legal safe harbour, a protection for the reporter that would not depend on the reported firm's forbearance, which is the structural rather than dispositional check Pettit demands. Wei and Heim go further and treat enforcement and mandatory channels as design choices a national regime must settle. Those are precisely the parts not in force, and the safe harbour is asked of legislators and providers rather than supplied by anyone here. What actually runs is voluntary on both sides, so a developer that ignores every report filed against it breaks nothing.

Public decision record

PartialClaimed · partial

Definition · Common-law practice

Rulings on challenges are written down, published and citable, so the system's interpretation of its own rules accumulates in the open, where it can be studied, invoked and criticized. The record is what makes the system predictable as well as legitimate: nothing is settled in private, and every decision leaves its reasoning behind it.

Analysis

The AI Incident Database is archived and searchable by outsiders, and the 750-plus incidents reviewed four years later show it accumulating as intended. What accumulates is harms rather than holdings. No one rules on a report, so nothing is settled by it, and no entry can be cited as authority for how a developer's own published rules should be read; the editors' own conclusion is that cause, extent of harm and severity frequently cannot be pinned down at all.

Concept Analysis: Newly Introduced

Near-miss reporting

Added

A common-law record contains only disputes that actually arose: someone had to be harmed and someone had to bring a case. Wei and Heim make the level of risk materialization a design dimension in its own right and build near-miss channels into their specification, so the record fills with failures that injured nobody and could never have reached a court. Nothing in republican or common-law practice provides for a docket of harms that did not happen.

Papers

No method has been built here yet — this approach rests entirely on what follows.