| paper | d4-conf-atal-2011-0186 |
| authors | — |
| venue | ? |
| filed under | unclassified |
| judged by | gpt-6-luna / xhigh (triple__luna__xhigh__c6r1) |
| judge confidence | high |
| authors would recognise it | unclear |
The named results characterize incentive properties and conditions for scoring-rule existence; they do not state a computational result that can anchor this mirror. A population of experts would require specifying how reports are aggregated and how payments work, while the paper's central result concerns one expert. No concrete computational question over a continuous society survives from these anchors.
fails bit a — no named computational result to mirror
No result is covered by a continuous-population mirror. Theorem 4.1 and Corollaries 5.1–5.2 concern scoring-rule incentives; Section 6 constructs quasi-strictly proper rules, and Section 7 discusses multi-expert markets, but neither supplies a computational result over a population.
I don’t see a defensible positive case under the required anchor rule. The paper has named mathematical results, but none asserts that a computational problem is in P, NP-hard, FPT, or otherwise classifies its complexity.
The closest candidate is Theorem 4.1, proved in this paper. It characterizes when a regular scoring rule is proper or strictly proper for a decision rule \(D\): such a rule exists exactly when it has the stated form derived from a convex or strictly convex function \(G\), with subgradient coordinates zero for actions that \(D\) does not select. That is a characterization of incentives, not a computational result. Corollaries 5.1 and 5.2 give sufficient conditions for the existence or nonexistence of strictly proper scoring rules; they do not classify the complexity of deciding either property. The paper’s reference to a “simple test” does not supply a computational complexity claim.
A large forecasting platform with many experts could suggest a population model: types might be experts’ complete conditional belief matrices, mass their share of the forecasting population, and the decision-maker could choose an action from aggregate reports. Section 7 makes that setting recognizable to the authors. But the paper discusses incentive problems in decision markets rather than giving a computational result about them. And its probability simplices are spaces of reports, not a continuous society. Treating those simplices as the mirror would continuize the report space, not the population.
So there is no honest anchor here from which to state the required continuous problem. The paper may be useful analytic background for such a future problem, but it offers no named computational result to mirror; trying to provide one would mean inventing an anchor or changing the paper’s question.
The proponent names no eligible anchor, and the paper supplies none. Theorem 4.1 characterizes proper scoring rules through convexity; Corollaries 5.1 and 5.2 give sufficient conditions for their existence or nonexistence. Neither classifies the complexity of a computational problem. Open Problem 2 asks about incentive properties, not computational complexity. The paper’s reference to a “simple test” does not specify an algorithmic problem or its running time.
A population of experts is imaginable, and Section 7 gives a reason to study one. But making experts into types and aggregating their reports requires choosing a new aggregation and payment model. Keep the single-expert setup, and adding identical experts gives the theorem no population-level role. Move to a multi-expert setup, and the question becomes a new mechanism-design problem, not a continuous mirror of a result in this paper. That might be worthwhile research, but this paper offers no named computational result to anchor it.
This is a strong case against a *mirror of this paper*, not proof that no related continuous-population problem could ever be worthwhile. The negative case is weakest on that broader universal claim: a new computational question about a population of experts remains conceivable.
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.