| paper | d4-conf-ijcai-2017-0776 |
| authors | — |
| venue | ? |
| filed under | unclassified |
| judged by | gpt-6-luna / xhigh (triple__luna__xhigh__c6r1) |
| judge confidence | high |
| authors would recognise it | unclear |
The proposed shared-model consistency question is concrete and plausibly recognizable to the authors. But the only named result offered as its anchor establishes structural properties, while the computational guarantee appears in prose rather than a numbered result; the anchor therefore fails the required computational-result test.
fails bit a — no named computational result to mirror
The proposed mirror concerns consistency and inference only; it leaves the paper’s optimal-set results aside.
The strongest plausible mirror is Population Lex-Consistency. Let \(V\) be decision variables with finite domains. Each type \(t\) is a bundle \(\Gamma_t\) of preference statements in \(L'_{pqT}\), and \(\mu_t\) is the fraction of a large population with that bundle. The question is whether there is a lexicographic model \(\pi\) satisfying every statement held by a positive-mass type:
\[
\pi \models \bigcup_{t:\mu_t>0}\Gamma_t.
\]
A solution is such a model: an ordered sequence of variables and a value order for each. This fits a setting where many stakeholders constrain one shared policy or configuration preference model, and many stakeholders repeat a relatively small number of statement bundles. It mirrors the paper’s consistency problem directly.
The relevant named result is Theorem 2, proved in this paper. It says that every statement \(\varphi\in L_{pqT}\) is strongly compositional and that \(\pi\models^*\varphi\) iff \(\pi\models\varphi^{(\ge)}\). For non-strict \(\varphi\), its negation is compositional; when \(R_\varphi=S_\varphi\), that negation is strongly compositional, with \(\pi\models^*\neg\varphi\) iff either \(\pi\models\neg\varphi\) or \(V_\pi\cap S_\varphi=\varnothing\). The paper’s consistency algorithm uses these properties. Its stated \(O(|V|^2|\Gamma|)\) bound, however, appears in prose after Proposition 4, rather than in a numbered theorem asserting that consistency is in P. So Theorem 2 is a useful structural anchor, but it does not meet your strict requirement for a named complexity-class anchor.
For the proposed population problem, I would expect tractability: collect the statements from positive-mass types and run the paper’s consistency algorithm on their union. The masses matter only in identifying the support; their values do not otherwise affect the answer. The mirror covers consistency and inference, not the paper’s optimal-set results. Further questions include whether one can efficiently find a model satisfying the greatest total mass of statements, but that changes the problem and is not established by this paper.
The soft spot is that pooling heterogeneous stakeholders’ statements under one shared lexicographic model imposes a consensus assumption. The authors might recognize it as their inference problem in a repeated-user regime, but only when the aim really is to infer one common model. Also, since the result depends on support rather than mass values, this is a high-multiplicity version of their problem more than a new continuous optimization problem.
The anchor does not meet the requested standard for a named computational result. Theorem 2 establishes strong compositionality and an identity about \(\models^*\); it does not assert a complexity bound. The \(O(|V|^2|\Gamma|)\) consistency algorithm is described in the prose after Proposition 4. That is a computational result in the paper, but the proposed anchor is a structural theorem rather than a named complexity result.
The proposed population question also drops almost all of the distribution. It asks whether one model satisfies the union of statements held by every positive-mass type. A type with mass \(0.001\) vetoes just as much as one with mass \(0.9\). Replicating a type adds nothing, and in a continuum the question reduces to consistency over the essential support. The distribution’s mass values have no role in the computation. This is a support-compressed version of the paper’s consistency input, rather than a problem whose computational structure depends on a continuous population.
There is a further modeling step: the paper infers one latent lexicographic model from statements about a decision maker’s preferences. Pooling statements from many people presumes they constrain the same shared model and that every constraint is a hard requirement. That could describe a group checking whether shared policy requirements are jointly consistent. It is a plausible scenario, but it adds a consensus interpretation that the paper does not supply.
A mass-sensitive repair would ask for a model satisfying the largest fraction of the population’s statements, or for one satisfying at least a chosen fraction. That is a coherent collective inference problem, but it requires deciding whether a type’s bundle is all-or-nothing, how to count different numbers of statements, and what fraction is acceptable. The paper’s consistency result gives no direct answer to that optimization problem: choosing which constraints to discard is an added aggregation objective.
So the best case against is that the exact mirror is unusually thin: it has a plausible shared-specification use, yet its continuous masses collapse to support, and its named anchor is not a complexity theorem. I would call the negative case weak at the universal level. A group that genuinely needs all repeated requirements to fit one common model is a defensible scenario, and the resulting consistency question is well-posed. The case against rests on whether that support-only question is worthwhile for ChoCo, not on showing that no such scenario exists.
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.