Contestatory constitutionalism

Decisions laid out as argument networks that can be challenged

The decision is represented formally as a network of claims that support or undercut one another, so a challenger can see which grounds actually carried the outcome. Because the links carry weights, one can compute how sensitive the conclusion is to each link and work out exactly what would have to change for the outcome to differ. Contestation here is built into the system's machinery rather than its interface. It is an exchange the system itself can carry out step by step, revising the decision when a challenge succeeds.

The method (highlighting differences with Contestatory constitutionalism)

Scroll the diagram sideways to see all of it.

Use it in codefrom aiag.codified import Argumentation
from aiag.codified import Argumentation

graph = Argumentation()
graph.add("Deny the loan", 0.5)
graph.add("Missed two payments last year", 0.7, supports=["Deny the loan"])
graph.add("Income has doubled since", 0.6, attacks=["Deny the loan"])
print(graph()["Deny the loan"])  # -> 0.55

# The applicant contests a supporting argument, and the decision weakens
graph.contest("Missed two payments last year", "The bank confirmed the missed payments were its own error", base=0.9)
print(graph()["Deny the loan"])  # -> 0.235

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

The approach builds working mechanisms for two of Pettit's three requirements. The grounds can be inspected, because the decision is laid out as named arguments and the links between them. A challenge can be acted on, because the gradient method computes which link weight must change, and by how much, for a chosen conclusion to reach a chosen strength. The third requirement is missing entirely. There is no forum. The deciding system assesses the grounds offered against itself, nobody has standing to file anything, and whoever runs the algorithm sets the target strength.

Non-domination

PartialNot claimed

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

Pettit's paradigm case of domination is an authority whose workings you cannot see. An argumentation framework addresses that part of the objection by laying out the grounds in a form a person can examine. However, the check stops at visibility. The operator builds the framework, sets the weights, chooses the target strength, and decides whether the revision is applied. The power that has been made visible is therefore not made answerable to anyone subject to it.

Public decision record

AbsentOmission

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

Nothing is written down, published, or made citable. Worse for the tradition, the mechanism erases its own history as it works. A successful contestation changes the edge weights until the framework yields the desired strength, and nothing preserves what was contested, on what grounds, or what changed. Common law keeps the overruled case on record so that the line of interpretation stays visible. Here, the earlier state of the framework is simply lost.

Papers

Contestability in Quantitative Argumentation

Xiang Yin et al., Jul 2025

arXiv:2507.11323MethodBuiltAuthority: Affected public (open)

A method that takes an argument network with weighted supporting and attacking links, computes how sensitive a chosen conclusion's strength is to each individual link weight, and then adjusts the weights step by step until that conclusion reaches a desired strength.

CodeEdgeWeightedQBAF
from aiag.codified import EdgeWeightedQBAF

graph = EdgeWeightedQBAF()
graph.add("Deny the loan", 0.5)
graph.add("Missed two payments last year", 0.7, supports=["Deny the loan"])
graph.add("Income has doubled since", 0.6, attacks=["Deny the loan"])

print(graph.attributions("Deny the loan"))  # G-RAEs: how sensitive the decision is to each link's weight
trace = graph.revise("Deny the loan", desired=0.45)  # step the weights along them until the decision drops to 0.45
print(trace)  # -> [0.55, 0.515, 0.48, 0.445]
  • EdgeWeightedQBAFProcedureaiag/codified/contestatory_constitutionalism.py:579 Contestability in quantitative argumentation: links carry weights, gradient-based relation attribution explanations give how sensitive a topic argument's strength is to each link, and the weights are moved along them step by step until the topic reaches a desired strength (Yin et al., 2025).

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.

Contestable AI needs Computational Argumentation

Francesco Leofante et al., May 2024

arXiv:2405.10729ProposalGap

A position paper arguing that a contestable AI system must be able to hold an exchange with humans or other machines. Such a system must explain its output and reasoning step by step, weigh the grounds offered against it, and revise how it decides when a challenge succeeds. The paper argues that computational argumentation is the technique suited to supporting this.