| paper | d4-conf-aaai-2014-0356 |
| authors | — |
| venue | ? |
| filed under | unclassified |
| judged by | gpt-6-luna / xhigh (triple__luna__xhigh__c6r1) |
| judge confidence | high |
| authors would recognise it | unclear |
The paper has no numbered theorem, lemma, corollary, or proposition asserting a computational result, so bit (a) fails. Its equations define methods and an optimization objective, while the reported accuracy is empirical; the proposed type-based formulation would create a new question rather than mirror a named result.
fails bit a — no named computational result to mirror
No named computational result is available to anchor a mirror; the paper's voting methods and empirical accuracy results remain unmirrored.
This paper has no numbered Theorem, Lemma, Corollary, or Proposition asserting a computational complexity or algorithmic result. Its numbered equations define aggregation methods and an optimization objective; its table reports empirical accuracy. Neither supplies a qualifying named anchor.
So there is no anchor on which to base the required per-result continuous problem. A continuous-population version of crowdsourced answer aggregation may be imaginable, but presenting it here as a mirror of a named result would violate the anchor requirement.
The proponent raised no named result to mirror, and that is the strongest point in the negative case. Equations (1)–(5) define voting procedures and an optimization objective; Table 1 reports accuracy on one dataset. The paper states no theorem about computational complexity or algorithmic tractability. So there is no qualifying result to anchor the requested mirror.
The best candidate for a new mirror is the participant-weighted method. For a fixed batch of questions, define each type by its complete answer-and-confidence history and let \(\mu_t\) be the mass of that type. One could ask how to infer an answer \(x_q\) for each question while estimating type weights \(w_t\), with \(\sum_t \mu_t w_t=1\). That is a coherent high-multiplicity formulation: grouping participants with identical histories preserves the information the method uses.
But it would be a new computational question, not a continuous version of a result the paper establishes. For one question, majority and confidence-weighted voting reduce to summing mass over a fixed set of answer-confidence types. For the multi-question method, the paper does not establish a complexity result, and a substantive formulation would need to specify the latent-truth and reliability assumptions that make jointly inferred answers meaningful. Different choices would yield different problems.
That leaves a defensible case against spending ChoCo’s effort on this paper: there is no named computational result to continue, and its strongest algorithmic candidate needs a new formal problem before there is anything to classify. This is not a universal objection to continuizing crowds. Repeated crowds with recurring answer histories could support a sensible high-multiplicity model; the negative case is weaker if ChoCo is willing to treat new aggregation problems as mirrors.
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.