| paper | d4-conf-ijcai-2007-0251 |
| authors | — |
| venue | ? |
| filed under | unclassified |
| judged by | gpt-6-luna / xhigh (triple__luna__xhigh__c6r1) |
| judge confidence | high |
| authors would recognise it | yes |
Proposition 1 is the only proposed anchor, and it concerns whether local rules can reproduce a global scoring rule, not whether a computational task can be solved or is hard. Its population version is statable, but that does not give this paper a named computational result to mirror.
fails bit a — no named computational result to mirror
Proposition 1 is an impossibility about the existence of local rule tuples, not a computation algorithm, computational hardness result, or computational formulation; the proponent already concedes this.
fatal: True
The proposed mirror restates Proposition 1's nondecomposability claim. It leaves the paper's computational motivation of computing outcomes from compact preference descriptions without a named computational result.
A continuous population is plausible for this paper’s setting. Think of a large community choosing a bundle with many attributes, where voters share a common dependency graph and fall into a modest number \(\tau\) of preference types. A type would be a complete ranking over bundles, and \(\mu_t\) the fraction of voters of type \(t\). Direct positional scoring would maximize the mass-weighted score of a bundle; sequential voting would choose each attribute’s value from the relevant conditional voter preferences. The authors would recognize this as the same voting problem with voter counts replaced by masses.
The closest numbered result is Proposition 1: *no positional scoring rule is decomposable*. In full, for a product domain and profiles whose rankings follow a common variable order, no positional scoring rule can be reproduced on every such profile by sequentially applying local voting rules to the individual variables. This has a natural continuous restatement: given a rational distribution \(\mu\) over rankings compatible with a common graph \(G\), ask whether there are local rules whose sequential outcome agrees with the direct positional-scoring winner for every such \(\mu\). A solution would be the tuple of local rules, or a finite-support counterexample. Since finite profiles are empirical distributions, Proposition 1’s nonexistence result carries over.
That is a structural impossibility result, though, not a computational-complexity result. The paper has no numbered theorem, proposition, or corollary showing that a problem is in P, NP-hard, or otherwise classifying its complexity. Its introduction mentions hardness of computing rules from succinct preferences, but cites Lang (2004); that claim is neither a numbered result of this paper nor an anchor under your criteria. The closest proposition therefore cannot support a Class A, B, or C prediction.
The weakest point is that the paper’s computational obstacle is the enormous bundle space and the structure of preferences, rather than the number of voters. A moderate-support population is a sensible regime for voter aggregation, but it does not resolve that obstacle. Also, Proposition 1’s direct-scoring comparison uses complete rankings, while the paper’s compact CP-net descriptions may leave rankings incomplete. So the honest positive case is that a population mirror is plausible, with Proposition 1 offering a useful structural near-miss; strictly, there is no eligible named computational anchor and thus no anchored continuous complexity question to submit.
The strongest case against is that this paper gives no named computational result to continue. Its numbered propositions concern rule decomposability, Condorcet winners, and aggregation axioms. The introduction’s statement that computing rules from succinct preferences can be hard is background attributed to Lang (2004), not a result established here. So Proposition 1 is the only anchor offered, but it is a structural impossibility claim, not a complexity classification.
The most faithful population version of Proposition 1 is easy to state: can local rules reproduce a direct positional-scoring winner for every rational type distribution compatible with the common variable order? But that is already the finite-profile question in different notation. For any such distribution \(\mu\), choose a common denominator \(q\) and make \(q\mu_t\) voters of each type. Weighted scores and weighted local majorities then reproduce the finite profile’s outcomes. Conversely, every finite profile gives a rational distribution. Thus a universal claim over these distributions is settled by the proposition’s finite counterexample; it asks no new algorithmic question. This is not a problem because the continuous answer would be the same. It is a problem because the proposed mirror has no computational task or complexity landscape to study.
The stronger formulation does not fix that. Using complete rankings as types avoids ambiguity about CP-net extensions, and repeated rankings are plausible in a large menu-setting community. But it leaves a property of voting rules, not a computational problem. Asking whether the rules agree for one input distribution instead changes the question: the paper’s result is about whether a fixed tuple of local rules works across all profiles. And a complete ranking over exponentially many bundles is itself an enormous representation, obscuring the compact-preference motivation of the paper.
One could pose a worthwhile new problem: compute a global scoring winner from a type distribution encoded by compact conditional preferences. That would engage the paper’s motivating difficulty, and I would not reject it because the bundle space may preserve hardness. But this paper neither classifies that problem nor supplies a named computational result for it. It would be a new research lead, not a continuous mirror of Proposition 1. On the paper-mirroring question, the negative case is strong; the broader claim that no continuous computational study of structured preferences could ever be worthwhile is not established by this paper.
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.