| paper | d4-journals-mss-2004-0026 |
| authors | — |
| venue | ? |
| filed under | unclassified |
| judged by | gpt-6-luna / xhigh (triple__luna__xhigh__c6r1) |
| judge confidence | high |
| authors would recognise it | yes |
The paper’s limit theorems concern a sensible continuous population regime, and the type-level limiting share can be stated precisely. But none of the named results asserts an algorithm, a computational formulation, or a complexity bound, so there is no qualifying computational anchor.
fails bit a — no named computational result to mirror
The proposed mirror concerns limiting Shapley–Shubik and Banzhaf power shares; the paper’s named results establish those limits under specified chain conditions, but make no computational claim.
There is a strong population-limit connection here, but no qualifying computational anchor. The closest results are Theorem 2.3, proved in the paper using Neyman (1982): in a replicative \(q\)-chain, if each voter’s relative weight vanishes while their replicas retain positive total relative weight, the Shapley–Shubik power ratio of any two voters converges to their weight ratio. Proposition 3.3, also proved here, establishes the corresponding Banzhaf result for half-quota chains satisfying the special local central limit condition. Theorem 3.6, proved here using a local central limit theorem, gives a sufficient condition for that Banzhaf result: finitely many positive integer weights, with the gcd of those occurring infinitely often equal to \(1\).
These are named mathematical results, but none asserts that a computational problem is in P, NP-hard, or otherwise classifiable by computational complexity. So I can’t honestly present one as the required anchor.
The most plausible mirror would use a large weighted-voting assembly whose members fall into a small number of standardized weight classes. A type is a weight \(w\); \(\mu_w\) is the fraction of members in that class. Members vote independently yes or no with probability \(1/2\), and a bill passes at the specified quota. The mirror question would be: given the finite-support distribution \(\mu\), compute the limiting share of total voting power held by each type as the assembly grows. Under the paper’s applicable conditions, the answer is
\[
\pi_w=\frac{w\mu_w}{\sum_v v\mu_v}.
\]
For Banzhaf power this uses the half-quota and local-limit conditions; for Shapley–Shubik power, Theorem 2.3 covers replicative chains at quota \(q\).
That is a credible continuous-population object: many members, few types, and type-level power remains meaningful as individual weight vanishes. A large cooperative or shareholder assembly with standardized voting classes is a plausible regime, and the authors would likely recognize the question as the limiting form of the weighted-voting problem they study. Evaluating the displayed formula is tractable. But the paper’s contribution is the analytic convergence result itself, not a complexity classification of this computation. The paper therefore supports the *sensible-regime* case for a continuous mirror, but does not supply a named computational result to anchor the requested case. The weakest point is exactly that distinction: its limit theorems are already the answer to the natural mirror, while computational questions such as convergence rates or finite-population approximation are follow-ons rather than results this paper establishes.
The strongest objection is that none of the proposed anchors is a computational result. Theorem 2.3, Proposition 3.3, and Theorem 3.6 establish convergence of voting-power indices along specified chains of finite games. They do not pose an algorithmic problem or classify its complexity. A large assembly with repeated weight classes is a sensible high-multiplicity regime, but that establishes the setting—not a computational question for this programme.
The best continuous version would give each weight class \(w\) a population fraction \(\mu_w\) and ask for its limiting share of total voting power. Where the paper’s conditions apply, the answer is
\[
\frac{w\mu_w}{\sum_v v\mu_v}.
\]
This is a faithful population-level restatement of the Shapley–Shubik result in Theorem 2.3 and, under the local-limit conditions, of the Banzhaf results in Proposition 3.3 and Theorem 3.6. The type-level quantity survives the limit even though an individual’s power vanishes. But computing it from the finite-support input requires only a weighted sum; the paper supplies an asymptotic identity, not a computational task with an algorithmic landscape.
The better variants do not give these anchors more computational substance. Theorem 2.3 concerns ratios for fixed individuals along a chain, and a limiting type distribution can lose exceptional individuals: the paper’s own example has limiting mass entirely on weight \(w\), while the power ratio of a single lighter voter fails to converge. Preserving such cases requires details of the growing sequence that a mass distribution omits. For Banzhaf power, the local-limit hypotheses likewise depend on sequence and lattice behavior. One could ask for a classification of such chains, but that would be a probability-theoretic limit question, not a computational mirror over societies given by finite-support masses.
Exact finite-size computation of voting power in a high-multiplicity weighted game is the strongest nearby algorithmic direction. It needs the population size and integer coalition structure: identical proportions at different population sizes can have different finite-game power indices. Including that information makes it a finite-game computation, rather than a problem determined by the continuous society, and it is not a computational result this paper proves.
So the case against is narrow but direct: this paper offers an excellent population-limit setting, yet no named computational result to mirror. The opposing case is right that the type-level society is sensible; its weak point is that the natural continuous questions attached to the three anchors are already limit statements, not the complexity questions ChoCo is organized to chart.
The adversarial triple: the proponent anchors on up to three named results; the opponent sees that case and must defeat every anchor; the judge decides which case convinced it. These are the pipeline’s own outputs, generated by tools/triple_run.py — no human edited them. The paper’s own text is not reproduced here beyond the quoted statement above.