| paper | d4-conf-ijcai-2021-0300 |
| authors | — |
| venue | ? |
| filed under | unclassified |
| judged by | gpt-6-luna / xhigh (triple__luna__xhigh__c6r1) |
| judge confidence | high |
| authors would recognise it | yes |
A recurring crowd with report types and exact masses is a plausible high-multiplicity setting, and the paper’s aggregation could be applied to it. But the only numbered result gives an asymptotic statistical guarantee, not a computational result under the stated standard; the paper’s own aggregation comparisons are empirical. Without a qualifying named result, there is nothing here to anchor a continuous computational mirror.
fails bit a — no named computational result to mirror
A report-mass version could compute the pairwise tournament and its resulting top alternative or ranking; it would not cover the paper’s experimental comparisons of accuracy, elicitation burden, or expressiveness.
The strongest positive case is for the *setting*, but this paper supplies no qualifying computational anchor. Its only numbered result, Theorem 1, is cited from Prelec et al. [2017]. It says that, under a condition on the prior and signal distribution, the true ranking is selected with probability tending to \(1\) as \(n\to\infty\). That is an asymptotic statistical guarantee, not a result about computational complexity. The paper’s own contributions are aggregation methods and empirical comparisons; it states no named theorem, lemma, corollary, or proposition classifying a problem as tractable or hard.
A plausible population regime would be a recurring, high-volume crowd asked to rank the same small set of alternatives. A type could be a participant’s complete vote-and-prediction report for a fixed elicitation format; \(\mu_t\) would be the fraction of the crowd giving report \(t\). Tens of thousands of participants could produce a much smaller set of distinct reports. Since the paper’s aggregation uses reported answers and their frequencies, the authors could reasonably recognize this as an idealized population version of their setting.
But that plausibility does not create a named computational result to mirror. The natural follow-up would be to compute the paper’s partial-information aggregation from exact report masses. I cannot present that as an anchored, prover-ready ChoCo question on the basis of this paper: there is no qualifying anchor, and the formal ordinal-to-cardinal aggregation details are deferred. The weakest point of the positive case is therefore decisive under the requested standard: the paper motivates a continuous crowd regime, but does not formulate or classify a computational problem for it.
The only named result the proponent offers is Theorem 1, and it is a statistical consistency guarantee attributed to Prelec et al.: as the number of respondents grows, the probability of selecting the truth tends to \(1\), under assumptions on the prior and signal model. It classifies no computational problem. The rest of this paper’s results are empirical comparisons of elicitation formats and aggregation methods. Under ChoCo’s stated standard, there is no computational result here to anchor a mirror.
The proposed recurring crowd is a sensible high-multiplicity setting: types could be complete vote-and-prediction reports, with their masses as the input. That makes a continuous version easy to state, and the paper’s pairwise aggregation can be applied to it. But its method already aggregates reported frequencies pair by pair. Replacing finite counts by exact masses changes the representation, not the computational question the paper studies. The paper asks whether elicitation formats help people recover a hidden ranking, and measures accuracy and burden experimentally; it does not ask for a complexity classification of computing the aggregate.
One could try to make accuracy the computational objective, but report masses alone do not specify the hidden ranking or the signal process that generated the reports. Supplying those would turn the task into a statistical analysis of the prior and signal model—the territory of Theorem 1—or add information unavailable to the paper’s principal. Neither supplies the missing computational anchor.
So the negative case is strongest on the programme’s own gate: the crowd setting is plausible, but this paper offers no named computational result whose continuous mirror would chart a complexity landscape. The case is not airtight if ChoCo is willing to treat evaluating the paper’s aggregation rule on exact report masses as sufficient; that is the best surviving candidate, though it is not a result the paper presents as a computational question.
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.