| paper | d4-journals-jair-2014-0011 |
| authors | — |
| venue | ? |
| filed under | unclassified |
| judged by | gpt-6-luna / xhigh (triple__luna__xhigh__c6r1) |
| judge confidence | medium |
| authors would recognise it | unclear |
Theorem 4 is computational, but the proposed bloc model leaves the population outside the game: masses only rescale weights in a finite τ-player instance. Independent members restore multiplicity, but an atomless individual’s Banzhaf pivotality vanishes; a type-aggregate limit could provide a mirror, though neither case establishes that limit or that the authors would recognize it as their design problem.
fails bit none — no continuous question survives
In the stated model, the masses only rescale weights and the Banzhaf coalitions are subsets of the τ blocs, so the population itself is not continuous.
fatal: True
The stated bloc model is finite; what remains undecided is whether a canonical type-aggregate Banzhaf limit under independent replication and proportional quotas is a well-posed continuous version the authors would recognize. A sequence-independent limit result, or a proof of path dependence or degeneracy, together with that recognition judgment would settle it.
The proposed bloc version covers normalized Banzhaf power-index design through enumeration on τ players; it does not cover member-level aggregate power without a limiting index, or the paper’s generic design framework and other results.
The strongest mirror I can defend is a type-bloc version of the paper’s inverse weighted voting game problem. It fits a large federation or membership organization where millions of members fall into a small number \(\tau\) of constituency classes, and each class casts one instructed bloc vote. Members of a type share the same mandate and desired per-member power \(r_t\); the type has population mass \(\mu_t\), with \(\sum_t\mu_t r_t=1\). The target power of the bloc is therefore \(p_t=\mu_t r_t\). Here the population is large relative to the number of types, while the blocs are the game’s decision-making units.
The mirror problem is: given \(T\), rational masses \(\mu\), and target \(r\), choose a quota \(q\) and nonnegative per-member weights \(a_t\). Bloc \(t\) receives weight \(\mu_t a_t\), and a coalition \(S\subseteq T\) wins exactly when \(\sum_{t\in S}\mu_t a_t\ge q\). Compute each bloc’s normalized Banzhaf index \(\beta_t\), counting subsets of the *other blocs* that it can turn from losing to winning. Return a weighted game minimizing \(\|\beta-p\|_2\), with an exact optimality guarantee. This continues the paper’s power-index design question: choose a weighted voting game whose Banzhaf distribution is closest to a specified target.
My anchor is Theorem 4, proved in this paper. It states that Algorithm 1, which enumerates all canonical weighted voting games on \(n\) players, runs in \(O^*(2^{n^2+2n})\) time. The algorithm generates games through their minimal winning coalitions, tests candidate extensions for weightedness, and removes duplicates. In the type-bloc problem, enumerate games on \(\tau\) blocs, compute each game’s Banzhaf vector and distance from \(p\), and return the best one. Relabeling the canonical games to match the target types adds at most a factor of \(\tau!\), absorbed by the stated bound; converting bloc weights to per-member weights adds only polynomial overhead in the input bit length \(L\). This gives an exact algorithm with running time \(O^*(2^{\tau^2+2\tau})\operatorname{poly}(L)\).
I expect the difficulty to track the finite \(\tau\)-bloc problem: the large membership count \(N\) drops out once the type masses and type targets are given. My tentative prediction is therefore toward hardness transferring if hardness for discrete PIWVGD is established; the paper itself reports no such hardness result, so that is a prediction rather than a classification. Questions this mirror raises include whether the design problem is FPT in \(\tau\), whether its exact search can avoid full game enumeration, and whether there is a principled bridge from individual-member power to aggregate type power.
The scope is narrow: it covers the paper’s weighted-game design problem with normalized Banzhaf targets. The paper’s generic enumeration framework could support other design objectives, but I am not claiming mirrors for those here.
The weakest point is the bloc assumption. Banzhaf power is defined through discrete coalitions of players; in this mirror, a coalition selects whole blocs, so \(\beta_t\) is bloc power rather than the aggregate limit of independent members’ Banzhaf power. That co-modification is plausible where constituencies vote collectively, and the paper itself treats parties, companies, and states as voting players. If its authors require every member to be an independent swing voter, this mirror would not preserve their power measure.
Theorem 4 is a real computational anchor: it gives a runtime for enumerating canonical weighted voting games, and that enumeration supports exact game design. The problem is that its \(n\) counts independent players, each of whom can join or leave a coalition. It does not count members represented by those players.
In the proposed bloc mirror, type \(t\) remains one player. Its total weight is \(w_t=\mu_t a_t\), and its Banzhaf index is calculated over subsets of the \(\tau\) blocs. Since any nonnegative \(w_t\) can be represented by choosing \(a_t=w_t/\mu_t\), the masses add no new structure to the game. The model is the original finite-player design problem with relabeled players and aggregate targets. That may be a sensible institutional problem, but its population is not continuous: the blocs, rather than their members, are the voting agents.
The stronger alternative is to let members vote independently, group identical members into types, and ask for each type’s aggregate power. That gives a genuine high-multiplicity question, but it exposes a problem with carrying over Banzhaf power. Its definition counts coalitions of individual players. In the atomless limit, a random coalition has exactly half of each type’s mass almost surely. Adding one individual changes no mass, so it cannot change whether the coalition wins. Every individual’s pivotality vanishes; the normalized index becomes \(0/0\).
One could define a useful finite-population asymptotic instead, summing each type’s individual Banzhaf power and taking a limit. But the result would be a new influence functional, whose definition depends on how finite populations and quota thresholds approach the limit. That may be worth studying, but Theorem 4’s enumeration of finite games does not itself provide a computational result for it.
So the negative case is strongest against the anchor as stated: its block version is finite, while its genuinely continuous version loses the paper’s exact power measure unless that measure is replaced by an asymptotic one. I would not claim the universal “no scenario” conclusion is airtight. A canonical limit of aggregate Banzhaf power in replicated, independently voting populations is a plausible mirror; whether its asymptotic definition is sufficiently natural to count as the authors’ problem remains the weak point in this case.
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.