| paper | d4-journals-mss-2022-0050 |
| authors | — |
| venue | ? |
| filed under | unclassified |
| judged by | gpt-6-luna / xhigh (triple__luna__xhigh__c6r1) |
| judge confidence | high |
| authors would recognise it | no |
Neither anchored theorem asserts that a problem can be computed or that computing it is hard. The proposed sign check for a limiting mass-weighted margin is coherent, but it replaces the paper's probability claims over ordered competence sequences with a new problem over type masses. The paper's practical discussion of weights motivates that follow-up, but does not make it a computational result of this paper.
fails bit a — no named computational result to mirror
The proposed model touches Theorem 3.3 and Theorem 4.3, but neither supplies a computational result to mirror; the remaining probability and weighted-rule results are also analytic.
The strongest honest positive case is a plausible large-jury population model. Let a voter type be \(z=(p,\eta)\), where \(p\) is the probability of voting for the correct option and \(\eta\) describes weight-assessment error. A society is a distribution \(\rho\) over these types; \(\rho_z\) is the fraction of voters of type \(z\). Under the paper’s independence assumption and bounded weights \(w_z\), the limiting weighted-vote margin is
\[ \sum_z \rho_z w_z(2p_z-1). \]
A natural continuous-population question is whether this margin is positive, so that the weighted majority selects the correct option in the large-population limit. A setting with many jurors or voters sharing a small set of competence and assessment profiles is plausible, and the authors would likely recognize the connection to their weighted Condorcet Jury Theorem. For rational, finite-support input, though, checking the sign is just a sum. It is a possible follow-up, not a computational result of this paper.
The paper’s closest named results are Theorem 3.3, proved here in Appendix A, and Theorem 4.3, proved here in Appendix C. Theorem 3.3 says that for a centered product measure \(\mu\) on competence sequences—whose reference distribution has mean \(1/2\), with coordinate measures satisfying the stated absolute-continuity and summable-divergence conditions—the probability that the independent Condorcet Jury Property holds is zero: \(\mu(C_I)=0\). Theorem 4.3 says that with the paper’s truncated-Gaussian weight errors, if the baseline measure assigns positive probability to competence \(p>1/2\), then there is an exponent \(k\) such that sufficiently large \(W\) and \((W-1)/\sigma_W\) make the weighted Condorcet Jury Property hold \(\mu\)-almost surely. These are asymptotic probability claims, not complexity classifications or algorithms for computational problems.
So there are no qualifying computational anchors here, and therefore no paper-anchored prover questions to state. The main limitation is also the key scope issue: the paper’s \(\mu\) is a meta-probability over infinite sequences of voter competences, not a society represented by masses of voter types. Turning it into the population model above is intelligible, but it changes the object being studied; it cannot supply the missing computational result.
The strongest objection is that the paper has no computational result to mirror. Theorems 3.3 and 4.3, Theorem 3.12, and Proposition 4.1 are statements about almost-sure convergence under probability measures. The paper defines no encoded computational problem and proves no complexity bound, algorithm, or hardness result. That matters here because a computational question one can invent from the model is not, by itself, a mirror of a result in the paper.
Theorem 3.3 is the clearest example. Its object is a prior \(\mu\) on infinite, ordered sequences of voter competences, and its conclusion is \(\mu(C_I)=0\). Replacing the sequence with a society distribution \(\rho\) over competence types changes what the theorem concerns: \(\rho\) records type masses, while the theorem concerns the behavior of every finite prefix under a second-order prior. The paper even gives an example where CJP depends on the initial voters. An unordered mass distribution forgets that prefix information. Preserving it by retaining each voter’s position restores the sequence model, rather than a high-multiplicity society. Putting a prior over \(\rho\) instead restores uncertainty at the meta-probability level; it does not turn the theorem into a computational problem about a continuous society.
Theorem 4.3 has the same gap. A society over joint competence and weight-error types gives a natural limiting weighted margin, \(\int w(2p-1)\,d\rho\). Asking whether that margin is positive is coherent, but it is not the theorem’s almost-sure claim over random voter sequences. Nor does the paper optimize a weight assignment: it establishes conditions under which a specified family of weights works. Choosing weights subject to implementation constraints could be a sensible follow-up, especially given Appendix D, but its objective and constraints would have to be supplied as a new problem. It would not mirror a computational result in this paper.
The positive case is weakest where it suggests that continuous populations themselves are implausible. A large jury with recurring competence and assessment profiles is quite plausible, and high multiplicity could be sensible for that regime. The problem is that the paper’s results use voter-specific competence sequences and stochastic errors, often with continuously distributed values; finite type masses do not preserve those sequence-level claims. One could also study finite-population reliability for rational masses of types, but that would be a different target from the paper’s asymptotic CJP results.
So I would reject the proposed margin and weight questions as mirrors, not because their answers are known or uninteresting, but because they introduce computational tasks the paper never states or analyzes. The negative case is weaker against such independently motivated follow-ups: the paper’s practical discussion gives a reader a real reason to formulate a weight-design problem. That possibility makes “no worthwhile mirror in any scenario” hard to prove, but it does not supply a computational anchor in 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.