| paper | d4-journals-mss-2016-0052 |
| authors | — |
| venue | ? |
| filed under | unclassified |
| judged by | gpt-6-luna / xhigh (triple__luna__xhigh__c6r1) |
| judge confidence | high |
| authors would recognise it | yes |
The continuous Minimax reversal-bias question is direct and well-posed, but the paper’s named results classify when a social-choice property holds. They assert no algorithm, computational hardness, or formulation used to compute, so there is no computational result to anchor a ChoCo mirror.
fails bit a — no named computational result to mirror
The proposed continuous classification mirrors Theorem 2’s type-1, type-2, and type-3 immunity results over ranking masses, with Propositions 6–8 giving related witness characterizations. It leaves the Borda and Copeland results, graph lemmas, and correspondence comparisons aside.
There is no qualifying computational anchor in this paper. It gives no named result about a problem being in P, NP-hard, FPT, or another complexity class. The main result, Theorem 2—proved here—characterizes when Minimax is immune to reversal bias: for type 1, exactly when \(h\le3\), or \(n\le3\), or \((h,n)\in\{(4,4),(5,4),(7,4),(5,5)\}\); for type 2, exactly when \(h=2\), or \(n\le3\), or \((h,n)=(4,4)\); and for type 3, exactly when \(n=2\), or \((h,n)=(3,3)\). Propositions 6–8 give related witness-existence characterizations. These are substantive named results, but they concern a social-choice property, not computational complexity, so none meets your anchor requirement.
The population mirror itself is unusually direct. In a large election, let a type be a complete ranking of the \(m\) candidates, and let \(\mu_t\) be the fraction of voters with ranking \(t\). For candidates \(x,y\), define the mass defeating \(x\) by \(y\) as \(d_\mu(y,x)=\sum_{t:y\succ_t x}\mu_t\). The continuous Minimax winners are
\[
M_\infty(\mu)=\arg\min_x\max_{y\ne x} d_\mu(y,x).
\]
Reversing every ranking gives \(\mu^r\). The three reversal-bias definitions carry over directly by comparing \(M_\infty(\mu)\) and \(M_\infty(\mu^r)\). Here the many agents might be millions of voters in a recurring election, grouped into ranking types; the support \(\tau\) may be far smaller than the population. There is no intervention, action variable, or objective in the paper’s problem—the object is a property of the outcomes.
The closest honest continuous research question is: for each \(m\) and bias type \(j\), characterize whether every rational distribution over rankings is immune to type-\(j\) reversal bias under \(M_\infty\). A witness is a finite rational distribution exhibiting the bias; a proof of immunity must cover all such distributions. This is recognizably the paper’s Minimax reversal-bias question with the population made continuous. It is a plausible mathematical classification problem, but it is not an anchor-backed computational problem. I would not assign it to Class A, B, or C on the evidence here.
The strongest point for the mirror is that fractions of a large electorate and weighted pairwise defeats preserve the paper’s central objects exactly. The weakest point is decisive for ChoCo: the paper studies a property classification, not algorithms or computational hardness. Its finite-\(h\) exceptions may also disappear when arbitrary rational masses are allowed, so the continuous theorem could be simpler than the result being mirrored.
The strongest case against a ChoCo mirror is that this paper contains no computational result to continue. Theorem 2 classifies when Minimax is immune to three reversal-bias properties as a function of \(h\) and \(n\). Propositions 6–8 restate those classifications as witness-existence claims. They are substantive structural results, but none concerns an algorithm, running time, or computational hardness. The other named results likewise concern social-choice properties or correspondences.
The population model itself is natural. In a large election, let each ranking type \(t\) have mass \(\mu_t\); pairwise defeat mass is \(d_\mu(y,x)=\sum_{t:y\succ_t x}\mu_t\), and reversal maps each ranking to its reverse. This gives a coherent continuous Minimax correspondence. There is no sound objection based on voters being individuated or the limit becoming trivial.
But the proponent’s proposed mirror—classifying which \(m\) make every rational distribution immune—remains a property classification, not a computational question of the kind ChoCo studies. The same is true of the continuous versions of Propositions 6–8. For a supplied distribution, one can compute all pairwise defeat masses, evaluate both Minimax outcomes, and compare them in time polynomial in \(m\), \(\tau\), and the input bit length. To get a harder algorithmic problem, one could add a task such as finding a minimum-cost mass change that creates or removes reversal bias. That would introduce an intervention and cost model absent from the paper; searching across all distributions for a witness would likewise be a new synthesis problem, not a computational result the paper establishes.
So the negative case is firm on the programme’s scope, but not universal in the broader mathematical sense. The paper has a well-posed continuous-population analogue that may interest social-choice theorists. What it lacks is a computational anchor that makes that analogue a ChoCo mirror.
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.