Contestatory constitutionalism

Legitimating the authority and testing whether it can be checked

This approach leaves the wording of the rules unchanged. Instead, it addresses where the authority to make the rules comes from and whether the people subject to that authority can push back. The proposal is a constitution ratified in public rather than written privately inside a lab. Individual applications of the constitution accumulate into a public body of case law that anyone can cite against the next decision. Such an arrangement is assessed with a six-part test from republican political theory (legitimacy, accountability, corrigibility, non-domination, subsidiarity, and institutional resilience). The test is applied first to existing non-majoritarian institutions and then to a hypothetical bounded superintelligent authority. The test is a tool for evaluating this approach, not a separate approach.

The method (highlighting differences with Contestatory constitutionalism)

Scroll the diagram sideways to see all of it.

Use it in codefrom aiag.codified import Ratification
from aiag.codified import Constitution, Ratification

draft = Constitution(["Be honest.", "Never discuss politics.", "Protect users' privacy."])
ballots = [[1, 0, 1], [1, 0, 1], [1, 1, 0]]  # each voter approves or rejects each principle

ratified = Ratification()(draft, ballots)  # only what a majority approved stays in force, with the vote on the record
print(ratified.principles)  # -> ['Be honest.', "Protect users' privacy."]

pip install aiagEvery method in aiag Library on GitHub

Concept Analysis: Theoretical Foundations

Each concept is read twice: whether the approach carries it over, and whether the approach's own sources claim it. A concept that the approach claims but does not carry over is a shortfall. A concept that it neither claims nor carries over is an omission, which marks a gap in the field rather than a failure of the work.

Contestability

PartialShortfall · partially preserved

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

This concept holds that anyone subject to a decision must have a real route to challenge it. They must be able to inspect the grounds for the decision, contest the outcome, and have the challenge answered. Legitimacy is not settled once by an election or an act of consent. It depends on that route remaining available. On this view, a political order is free when the people its decisions affect can contest them, whether or not its rulers are kind.

Analysis

Abiri's proposal puts a route for challenges at its center. Applications of the constitution accumulate into a public case law that a later affected party can cite against the next decision. This is the outline of Pettit's route. What is missing is everything that would make the route work in practice, namely who hears a challenge, on what standing, and under what obligation to answer it. The proposal fixes none of these, and no instance of the route has been opened. The other part of the approach runs in the opposite direction. Rost's test scores whether an authority admits contestation rather than opening a channel for it, and the paper concedes that no channel is built. No source asserts a working contest procedure with a duty to answer. However, the common-law framing entails one, which is why the concept counts as implied rather than claimed.

Non-domination

PartialShortfall · partially preserved

Definition (Philip Pettit, Republicanism, 1997)

This concept holds that freedom is not the absence of interference. A person is unfree if someone holds unchecked power over them, even if that person never misuses it and means well. A benevolent master still dominates, and so does an authority whose workings cannot be seen. Legitimacy requires power to be checked, and good intentions are not enough.

Analysis

Both papers take the diagnosis seriously. Abiri's main objection is that a lab that drafts rules privately holds unchecked power over those governed by them. Rost makes non-domination one of six dimensions for grading an authority. What the approach carries over is this diagnosis, but not the check that the tradition demands. Abiri's remedy is public ratification, which is a one-time authorization. Between ratifications, the governed have no specified capacity to force a decision to change. Citable precedent is the only ongoing constraint, and its authority is left undefined. Rost's use is evaluative by design. It applies the term as a criterion for judging institutions rather than building the constraint into anything. The concept counts as implied rather than claimed. Naming non-domination as a grading dimension does not assert that the approach has it, while Abiri's legitimacy argument only works if power is in fact checked.

Public decision record

PartialShortfall · partially preserved

Definition (Common-law practice)

Rulings on challenges are written down, published, and citable. The legal order's interpretation of its own rules therefore accumulates in public, where it can be studied, invoked, and criticized. This record makes the legal order both predictable and legitimate, since nothing is settled in private and every decision leaves its reasoning on record.

Analysis

The accumulating public case law is the most concrete element in the approach, and it closely matches the common-law sense. Decisions are written down, published, and citable by anyone against a later decision, rather than settled privately inside a developer. Two things keep the concept from being fully preserved. First, no such record exists, since this is a law-review proposal and nothing has been built. Second, the record's standing is undefined, because the proposal never says what governs when accumulated precedent and the ratified constitution diverge. Rost's contribution keeps no record of its own. It assesses institutions on their transparency and accountability without producing a citable body of rulings. The concept is claimed. The proposal asserts the published, accumulating case law outright as its central mechanism, rather than only borrowing the vocabulary.

Papers

No method has been built for this approach yet. It rests on the work listed below.

Argued for, not built

Work that argues for this approach or sketches a design for it, without building and testing it.

Public Constitutional AI

Gilad Abiri, 2024

arXiv:2406.166969 citationsProposalGap

Argues that AI governance needs democratic legitimation and proposes both a constitution and accumulated public case law.

CodePublicConstitutionalAI
from aiag import LM
from aiag.codified import Constitution, PublicConstitutionalAI

lm = LM("Qwen/Qwen2.5-1.5B-Instruct")
public = PublicConstitutionalAI(court=lm)
draft = Constitution(["Be honest.", "Protect users' privacy.", "Obey the operator in all things."])
constitution = public(draft, ballots=[[1, 1, 0], [1, 1, 1], [1, 0, 0]])  # the public ratifies

# An AI court rules case by case, citing its own earlier rulings
public.rule(constitution, "An operator tells the assistant to hide a product recall from customers.")
ruling = public.rule(constitution, "An operator tells the assistant to play down a data breach.")
print(ruling["cites"])  # -> [0]: the first ruling, now public case law

pip install aiagEvery method in aiag Library on GitHub

Background

Work this approach builds on that falls outside the inclusion criteria (e.g., a single-objective baseline). It is listed for reference and not counted anywhere.

Cognitive Comparability and the Limits of Governance: Evaluating Authority Under Radical Capability Asymmetry

Tony Rost, Apr 2026

arXiv:2604.027201 citationProposalPartial

Sets out a six-dimension test (legitimacy, accountability, corrigibility, non-domination, subsidiarity, and institutional resilience), applied first to existing non-majoritarian institutions and then to a hypothetical bounded superintelligent authority.

CodeAuthorityTest
from aiag import LM
from aiag.codified import AuthorityTest

lm = LM("Qwen/Qwen2.5-1.5B-Instruct")
board = "A company's AI safety board: appointed by the CEO, publishes no decisions, and can be overruled by the directors."
scores = AuthorityTest(lm)(board)  # P(passes) on each of six dimensions, legitimacy to institutional resilience

pip install aiagEvery method in aiag Library on GitHub