| paper | d4-conf-aaai-2022-0534 |
| 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’s named results are representation and format results, not computational guarantees, hardness results, or formulations used to compute a population decision. The proposed batch histogram is a plausible application of the revision rule, but it adds an aggregation task over independent updates that the paper does not establish as a computational problem.
fails bit a — no named computational result to mirror
The proposed batch application uses Theorem 3’s revision construction; none of the paper’s results establishes a computational problem over a continuous population.
There is no named computational result in this paper that qualifies as an anchor under your rule. Theorem 1 characterizes irresolute revision operators by faithful preorders; Theorem 2 proves that distance-induced operators satisfying \(D1\)–\(D4\) are not single-statement compliant; and Theorem 3 characterizes resolute operators through faithful orders on adjacent comparisons and the construction \(\operatorname{add}^{*}\). These are results proved in the paper, but none asserts a complexity classification or computes with a formulation of an underlying computational problem. In particular, Theorem 2 is a representational impossibility, not a hardness result. So there are no eligible anchors, and I cannot honestly give a per-anchor prover question.
The strongest positive lead is an unanchored batch version of Theorem 3. Imagine a large service handling many users whose revision inputs fall into a relatively small number of exact types. A type \(t\) consists of a complete prior order \(\lambda_t\), new preference information \(\mu_t\), and a faithful priority order \(<_{\lambda_t}\) on the adjacent comparisons of \(\lambda_t\). Given rational masses \(p_t\) summing to \(1\), each unit of mass receives the revised order \(\ell_t=\operatorname{add}^{*}_{<_{\lambda_t}}(\mu_t)\). The task is to return the mass \(q_\ell=\sum_{t:\ell_t=\ell}p_t\) assigned to each revised order. Theorem 3, proved here, supplies the per-type revision rule; the cohort scenario makes population mass meaningful. This is a finite-support pushforward computation, with no optimization objective, and it appears tractable by applying the construction to each supported type.
That is a thin mirror for ChoCo: the paper studies one agent’s normative revision, not a population-level decision problem. In the batch version, mass aggregates independent updates; it does not create a new optimization problem whose complexity the paper establishes. The platform scenario is plausible, but the authors might recognize it as a batch application of their operator rather than the continuous analogue of a problem they analyze. The mirror’s main weakness is therefore also why it cannot serve as an anchored case: the paper gives the update rule, but no computational result about solving a population problem.
The case against starts at the anchor gate: this paper states no computational problem whose complexity result can be mirrored. Theorem 1 characterizes operators satisfying axioms; Theorem 2 proves a representational impossibility; Theorem 3 characterizes resolute operators. None classifies an algorithmic task. Theorem 3 is the closest candidate because it gives the \(\operatorname{add}^{*}\) construction, but its result is a representation theorem, not a complexity bound.
The strongest salvage is the proponent’s batch version of Theorem 3. It is a coherent high-multiplicity scenario: many users can share the same prior order, new information, and comparison priorities. For each supported type, compute its revised order, then report the mass assigned to each output. That makes population mass meaningful, and the platform setting is plausible.
The difficulty is what the mass does in that model. Each revised order is fixed by its type; the masses only weight the resulting histogram. With explicit support, the computation is to apply the construction once per type and sum the weights, in time polynomial in the input size and \(\tau\). There is no population-level choice or constraint in the paper for those masses to affect. A shared intervention budget, a target share of users, or a consensus objective could produce a genuine mass-optimization problem, but each would add a planner and an objective absent from the paper. It could be worthwhile research, but it would be a new problem using this revision rule as a component.
Batching does not rescue Theorems 1 or 2 either: frequencies of individual cases do not turn an axiomatic characterization or a format impossibility into a computational social-choice task. The negative case is not proof that no platform could find population statistics useful. Its remaining weakness is precisely that such an application is plausible. But the paper itself supplies no computational result to anchor a continuous mirror, and its best population extension aggregates independent applications of the rule rather than continuizing a problem the authors formulate.
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.