| paper | d4-journals-mss-2018-0066 |
| authors | — |
| venue | ? |
| filed under | unclassified |
| judged by | gpt-6-luna / xhigh (triple__luna__xhigh__c6r1) |
| judge confidence | high |
| authors would recognise it | yes |
Both anchor groups state computational results, and their population interpretations are explicit and plausible. But the paper itself defines the mixture weights as population fractions of ranking types and proves the proposed rationalizability and axis-recognition results in that model, so the mirror is already present in the paper. The proposed near-single-peaked extension is not one of the named anchors and does not change that collision.
fails bit b-already — the paper is already native to the continuum
The paper already formulates the mixture over rankings as population fractions and proves these same tests; the proponent acknowledges this collision but offers no reason it is not a collision under the stated gate.
fatal: True
The proposed mirror duplicates Theorems 1–4 on mixture feasibility and axis recognition; it leaves the model-power analysis and the proposed near-single-peaked extensions outside its scope.
The strongest positive case is unusually direct: interpret the mixture weights as shares of a large population, not as changes in one person’s preferences over time. The alternatives might be a fixed set of budget or policy options on a common left-to-right axis. A type is a complete ranking of those alternatives; \(\mu_m\) is the fraction of residents with ranking \(m\). In a large poll, the observed rate \(p_{ij}\) at which \(i\) is chosen over \(j\) is then \(\sum_{m:i\succ_m j}\mu_m\). With many residents and a fixed, modest number \(n\) of alternatives, the number of people can be enormous while the number of distinct types is bounded by \(2^{n-1}\), and may be much smaller. For this problem, rankings are the whole type: people with the same ranking are indistinguishable in every respect the model uses.
Lead anchor — Theorem 1 and Theorem 2, proved here. Theorem 1 gives a necessary and sufficient test for whether rational pairwise choice rates \(P\) can come from a mixture of rankings single-peaked with respect to a *given* axis \(\triangleright\): for every \(i\triangleright j\triangleright k\), \(p_{ij}\le p_{ik}\) and \(p_{kj}\le p_{ki}\). The paper notes that Dridi [12] first derived the condition, and supplies its reformulation and proof. Theorem 2 proves that this condition can be checked in \(O(n^2)\) time. The continuous problem is: given \(P\) and \(\triangleright\), find nonnegative masses \((\mu_m)\) summing to \(1\), supported on rankings single-peaked with respect to \(\triangleright\), that reproduce every \(p_{ij}\); decide infeasible if none exist. The objective is feasibility, and a solution is the mass distribution itself. The proof of Theorem 1 constructs one, using at most one ranking per distinct pairwise rate, plus a ranking to fill any remaining mass.
This is a good fit for ChoCo’s model: the unknowns are type masses, and the pairwise rates are linear aggregate measurements of those masses. It also has a clean discrete bridge. If the rates come from \(N\) people with complete rankings, they have denominator \(N\); the construction uses rational masses with that denominator, so feasibility corresponds to an actual finite electorate with integer type counts. This anchor is Class A, tractable. There is no discrete hardness result here whose difficulty dissolves in the continuum.
Second anchor — Theorem 3 and Theorem 4, proved here. Theorem 3 says that when the axis is unknown and at least one valid axis exists, Algorithm 1 constructs a PQ-tree representing exactly the set \(L_P\) of axes that work. Theorem 4 proves that Algorithm 1 runs in \(O(n^2)\) time. The corresponding continuous problem takes rational pairwise rates \(P\), asks whether there is any axis and a distribution over rankings single-peaked on it that produces \(P\), and, if so, returns a compact representation of all valid axes together with a supporting mixture. This is also Class A, tractable; the mixture witness follows by applying Theorem 1 to a valid axis. This extends the lead question by making recovery of the common ordering part of the computation.
The regime is plausible for repeated population polling: a large cohort chooses pairwise among a small set of options along an accepted policy dimension. The paper’s own population interpretation makes this an especially easy mirror for its authors to recognize. The scope should stay narrow: exact rationalizability of pairwise choice rates by single-peaked ranking types, with a known or unknown common axis. This does not cover manipulation, robustness, or near-single-peaked populations; the paper’s discussion of allowing some non-single-peaked mass suggests further questions, but does not give a named computational result for them.
The weakest point is also decisive for novelty. The manuscript explicitly says its mixture model can be used at the population level, with the mixture probabilities representing the fractions of the population holding each preference. Thus the continuous-population question above is not merely a plausible mirror waiting to be formulated: it is already part of the paper’s stated model and results. This makes the mirror exceptionally strong on author recognition and mathematical fit, but a strict ChoCo novelty test may call it a collision—the paper has already done the continuization. And while the paper contains no named hardness anchor, it does provide substantive algorithmic anchors; there is no basis here for claiming that continuization removes a hardness barrier.
The strongest case against the two anchors is a direct collision. The paper already defines mixture weights \(x_m\) over single-peaked rankings and requires \(p_{ij}=\sum_{m:i\succ_m j}x_m\). It then explicitly gives the population interpretation: \(x_m\) is the fraction of people with ranking \(m\). Identifying each ranking with a ChoCo type and setting \(\mu_m=x_m\) gives the same continuous society, constraints, and feasibility question. The input here is pairwise rates, with the type masses as a witness rather than as supplied data, but that is the paper’s own rationalizability problem—not a new population version of it.
That defeats the first anchor in its strongest form: given an axis, ask whether the rates have a supporting mass distribution and return one if they do. Theorem 1 characterizes feasibility, Theorem 2 tests it in \(O(n^2)\), and the proof constructs the mixture. Treating rates from \(N\) people as a finite electorate does not create a new mirror either: their rational mixture weights can be scaled to integer type counts.
The unknown-axis anchor has the same problem. The strongest version asks for all compatible axes and a supporting mixture. Theorem 3’s PQ-tree represents the compatible axes, Theorem 4 gives the \(O(n^2)\) bound, and Theorem 1 supplies the mixture for a returned axis. The paper already solves the joint population-rationalizability question. This objection is not that the population regime is implausible; the paper explicitly says it applies, and the regime is sensible. It is that these results already address it.
I can defend “no worthwhile continuous mirror of these named results” strongly. I cannot honestly defend the universal claim that no related continuous computational question is worthwhile: the paper’s §6.1 proposes bounding the mass of non-single-peaked types, which could motivate a distinct complexity question. That would be an extension of the paper’s model, not a surviving mirror of either anchor.
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.