| paper | d4-conf-aaai-2015-0773 |
| 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 uses its (MP4) formulation to fit weights and reports experiments, but neither is a numbered theorem, lemma, corollary, or proposition asserting a computational result. The opponent’s population-weighted reviewer model is a plausible extension, but it cannot supply the required paper result to anchor a mirror.
fails bit a — no named computational result to mirror
No qualifying named computational result is available to anchor; the paper’s weight-fitting formulation and experiments are not numbered theorem-like results.
This paper has no qualifying named computational result to anchor a continuous mirror. It presents LP and MIP formulations, labeled (MP1)–(MP4), and reports experimental results for its proposed weight-estimation method. But it contains no numbered Theorem, Lemma, Corollary, or Proposition asserting a complexity or algorithmic result. In particular, (MP4) is a formulation label, not such a result.
A possible follow-up would replace the finite expert panel with a distribution over evaluator types and fit objective weights against their mass-weighted ordinal comparisons. That could be a sensible population model, but it would be an analyst-added extension, not a mirror anchored in a named result of this paper. Under your anchor rule, there is no continuous problem here to classify.
The proponent is right about the narrow fact: the paper has no numbered complexity theorem. That fact does not establish that it has no computational result to mirror. Its proposed method, culminating in formulation (MP4), is a concrete algorithmic contribution: it fits a shared weight vector while allowing evaluator-specific thresholds and selecting basis functions. The paper also reports an empirical result for that method.
A plausible continuous question would take a large pool of reviewers assessing outputs from the same optimizer or simulator. Each evaluator type would encode a shared rubric and response pattern; a rational mass \(\mu_t\) would represent its share of reviewers. One could fit common weights and type-specific thresholds by minimizing mass-weighted slack, retaining (MP4)’s basis-function choices. The question could be parameterized by the number of features, evaluator types, comparisons, candidate basis functions, and input bit length. That is a population-continuous version of the paper’s estimation task, not of its outcome space.
The strongest case against it is about fit. In the paper, each expert supplies an observed table of judgments, and the method estimates one criterion from that finite evidence. To preserve those judgments exactly, an evaluator type may need to include their responses across all queried pairs; typically there may be few repeated types. To group evaluators by rubric instead requires a shared model of how rubrics generate responses, which the paper does not provide. The continuous version also estimates a population consensus criterion, rather than fitting the particular panel the paper studies. And its output is a scoring objective for another system to use, not a collective choice among alternatives. That makes it a plausible inverse-preference-learning extension, but a weak anchor for a continuous computational social choice programme.
I cannot honestly make the universal negative case airtight. The weighted version is mathematically natural, and a large pool of evaluators sharing rubrics is a plausible regime; the paper’s authors could recognize it as an extension of their method. The strongest negative conclusion is that the paper supplies no formal complexity result to mirror and that the promising continuous question needs extra modeling assumptions and a social-choice connection. That falls short of showing that no worthwhile 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.