| paper | d4-conf-atal-2010-0264 |
| authors | — |
| venue | ? |
| filed under | unclassified |
| judged by | gpt-6-luna / xhigh (triple__luna__xhigh__c6r1) |
| judge confidence | medium |
| authors would recognise it | yes |
other→P discrete result → its continuous mirror
Answer: Polynomial-time solvable: weighted isotonic regression computes an exact optimal rational design with quota \(q=\tfrac12\). · discrete side: other · adjudication: proved
full proof & adversarial review →
Corollary 12
statement extracted from the paper’s text layer
Given rational type masses \mu_t>0 summing to \(1\), rational target densities \pi_t\ge 0 with \sum_t\mu_t\pi_t=1 , and rational \varepsilon>0 , find rational non-increasing type weights w_1\ge\cdots\ge w_\tau\ge0 normalized by \sum_t\mu_tw_t=1 and a quota q\in(0,1] whose objective is within \varepsilon of its infimum. For each admissible replication size N , form a finite game with N\mu_t members of type t , each of weight w_t/N , and quota q ; if \beta_{t,N}(w,q) is one such member’s normalized Banzhaf index, minimize \limsup_{N\to\infty,\;N\mu_t\in\mathbb Z\;\forall t}\sum_t\mu_t(N\beta_{t,N}(w,q)-\pi_t)^2 .
A society is a rational distribution \mu over interchangeable governance classes. Choose one normalized weight per class and a quota; evaluate the squared error between target influence densities and the type-level normalized Banzhaf densities induced by finite replications of \mu .
The limsup objective may capture replication-specific oscillations rather than a well-defined continuum power density, and the proponent gives no convergence result.
fatal: False
The mirror covers Corollary 12’s exact inverse power-index design for canonical weighted voting games in a class-based high-multiplicity regime, using squared error for Banzhaf power densities. It does not directly mirror the enumeration structure in Theorems 4 and 8–11 or design for other power indices.
The best positive case is for a high-multiplicity version of the paper’s inverse power-index design problem. Take a very large cooperative federation with members in a fixed number \(\tau\) of statutory classes. Members in a class have the same governance role and desired influence. The federation chooses one voting weight per class and a quota. Here \(N\) may be enormous while \(\tau\) stays small; \(\mu_t\) is the fraction of members in class \(t\). A class is a genuine type because its members are interchangeable for this design problem.
The paper has no numbered hardness result for inverse design. Its strongest eligible anchor is Corollary 12, proved here: for any power-index function \(f\) computable within exponential time, ExactEnum is an exact anytime algorithm running in exponential time for \((f,g,\mathcal G_{\mathrm{cwvg}},L_{\mathrm{weights}})\)-PVGD, where \(g\) is the paper’s sum-of-squared-errors objective and \(\mathcal G_{\mathrm{cwvg}}\) is the class of canonical weighted voting games. It enumerates candidate games and keeps the one whose power-index vector is closest to the target. The paper’s cited polynomial-time result for recognizing a weighted game from its minimal winning coalitions supports the construction, but Corollary 12 is the anchor.
My lead mirror is Continuous High-Multiplicity Banzhaf Design. Its input is rational type masses \(\mu_t>0\), summing to \(1\), and rational target influence densities \(\pi_t\geq 0\), ordered with the types and satisfying \(\sum_t\mu_t\pi_t=1\). A design chooses rational per-member weights \(w_1\geq\cdots\geq w_\tau\geq0\), normalized by \(\sum_t\mu_tw_t=1\), and a quota \(q\in(0,1]\). In the continuum game, a coalition may contain any mass \(x_t\in[0,\mu_t]\) of each type, has weight \(\sum_t w_tx_t\), and wins exactly when that weight is at least \(q\).
To define power without assigning positive power to an infinitesimal voter, use finite replications of that society. For each admissible \(N\), make \(N\mu_t\) voters of type \(t\), each with weight \(w_t/N\), and let \(\beta_{t,N}(w,q)\) be one such voter’s normalized Banzhaf index. Ask for weights and a quota whose objective
\[
F(w,q)=\limsup_{\substack{N\to\infty\\N\mu_t\in\mathbb Z\ \forall t}}
\sum_t\mu_t\bigl(N\beta_{t,N}(w,q)-\pi_t\bigr)^2
\]
is within a given rational \(\varepsilon>0\) of its infimum. This is a precise continuous limit problem: \(\pi_t\) is the type’s target influence density, and the decision variables are its per-member weight and the quota. The objective comes directly from the paper’s squared-error objective: setting each type-\(t\) voter’s target to \(\pi_t/N\) gives finite error \(\sum_tN\mu_t(\beta_{t,N}-\pi_t/N)^2\); multiplying by \(N\), which does not change the finite instance’s minimizer, yields the expression above.
This is a restricted, canonical high-multiplicity version of their design problem, not a claim that Corollary 12 already solves it. The choice of normalized Banzhaf power is covered by the corollary’s \(f\)-generality. The restriction to common weights within a type is the point of the regime: governance is assigning a rule to classes of otherwise interchangeable members. I would expect Class A in regular regimes, if the limiting influence map can be computed and optimized efficiently in the \(\tau\) type weights and quota. The natural questions are whether the limit exists beyond the \(\limsup\) formulation, how fast finite replications approach it, and whether the optimization is tractable as \(\tau\) grows.
The weakest point is that the paper studies small weighted games and its main algorithm depends on enumerating their discrete structures. In the continuum limit, normalized individual power vanishes; rescaling it into a type-level density may wash out much of the quota and game-enumeration structure, perhaps leaving an almost weight-only optimization. That would make this a sensible continuous limit of the authors’ inverse-design question, but a much easier one. The positive case rests on the preserved design objective and high-multiplicity regime; it does not assume the discrete algorithmic difficulty survives.
Corollary 12 is a genuine computational anchor, so I cannot reject this paper on the ground that it has no result to mirror. The strongest objection is about what its design problem measures. In a finite weighted voting game, Banzhaf power is defined through a player’s ability to change the result when added to a coalition. The paper’s algorithm enumerates games on a finite set of players and evaluates a vector of individual powers. A distribution over types does not, by itself, specify those one-player changes or a probability measure over coalitions.
The proposed mirror handles that by defining power through exact finite replications, then rescaling each player’s normalized power by \(N\). That gives a coherent asymptotic problem, but the scaling and the replication rule supply essential structure: the mass threshold game alone has no individual whose Banzhaf power can be measured. A more complete version could make the mirror algorithmic by asking, for rational input \((\mu,\pi,\varepsilon)\), whether the infimum of the specified limiting objective is at most \(\varepsilon\). It would still be a question about an asymptotic power index induced by finite coalition sampling. Changing the index or the sampling convention can define a different limit problem, so the paper does not single out one continuous extension.
That is the best case against the anchor: its continuous version needs a particular microscopic convention to make power well-defined, and its target is a rescaled limit of finite-player influence. But the proposed federation of interchangeable members is a credible high-multiplicity regime, and exact replication makes the convention explicit. Those choices are compatible with a continuous mirror of inverse power-index design; they are not a fundamental objection. I therefore cannot honestly sustain the requested universal claim that no worthwhile mirror exists. The negative case is weak here.
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.