| paper | d4-journals-mss-2012-0039 |
| authors | — |
| venue | ? |
| filed under | unclassified |
| judged by | gpt-6-luna / xhigh (triple__luna__xhigh__c6r1) |
| judge confidence | high |
| authors would recognise it | no |
Corollary 1 is a representation characterization, not a computational result with an input encoding, algorithmic bound, or hardness claim. The evaluator panel adds a new population model, while its masses do no work in the theorem's unanimous comparison.
fails bit a — no named computational result to mirror
No computational result is covered; Corollary 1, the Main Representation Theorem, Corollary 2, and the uniqueness theorem are representation results.
I cannot supply a qualifying computational anchor: this paper contains no named result asserting that a computational problem is in P, NP-hard, FPT, or similar. Its main result is a representation theorem, and the proof’s “constructive” approach is not an algorithm with a complexity claim.
The closest result is Corollary 1 (Expected multi-utility representation), proved here from the Main Representation Theorem. It says that, when a strict preference on lotteries over a finite prize set is bounded and satisfies the paper’s strict partial order, Archimedean, and independence axioms, it has a representation by a nonempty closed convex set \(U\) of utility functions: the derived weak comparison holds exactly when every \(u\in U\) gives the first lottery at least as high an expected utility, and the strict comparison holds exactly when every \(u\in U\) gives it strictly higher expected utility. The paper also gives a uniqueness result for the cone generated by the representing utilities.
The strongest possible population interpretation would be a panel of many evaluators choosing among lotteries, with each complete type a von Neumann–Morgenstern utility vector \(u\), and mass recording the share with that vector. A corresponding question could take a finite prize set \(X\), a compact rational polytope \(K\) of utility vectors supporting a full-support distribution, and lotteries \(p,q\); it asks whether \(u\cdot p\ge u\cdot q\) for every \(u\in K\), returning a violating \(u\) if not. With an explicit polytope, that comparison is an LP, so the imagined mirror would be tractable.
That is a weak bridge, not a sound ChoCo case. In the paper, \(U\) represents one decision maker’s incomplete preference; it is not a population. The comparison depends on every utility in \(U\), so the masses assigned to utility types do no work. The proposed question therefore lacks a meaningful population-mass role and is not a computational problem established by any named result in the paper. Its scope would cover only the risk representation in Corollary 1; extending it to acts would repeat the same problem. The proposal is the best intuitive analogy I see, but I would not present it as an anchored continuous mirror.
The strongest objection is that the paper supplies no computational result to mirror. The Main Representation Theorem and Corollary 1 characterize when an incomplete preference relation has a multi-utility representation; the uniqueness theorem characterizes the representing cone. None states the complexity of a decision, search, or optimization problem. Calling the proof “constructive” does not give it an input encoding or a running-time claim. So Corollary 1 is the closest anchor, but it is not a computational anchor of the kind this programme is meant to continuize.
The proposed panel of evaluators does not repair that gap. In the paper, lotteries are alternatives and their probabilities are over prizes; acts map states to lotteries. Neither is a distribution over people. If we make utility vectors into evaluator types and preserve the paper’s unanimous comparison, then \(p\) is weakly preferred to \(q\) exactly when \(u\cdot p\ge u\cdot q\) for every \(u\) in the population’s support. The weights assigned to those types do not matter: any two distributions with the same support induce the same comparison. Reweighting the evaluators changes nothing, while introducing a new type into the support can change the comparison regardless of how little mass it receives. An LP over an explicitly represented set \(K\) can check this condition, but that is a newly specified support-separation task, not a complexity result in the paper.
The more population-sensitive alternative changes the object being represented. If evaluators’ utilities are averaged by mass, the population has one aggregate utility, \(\bar u=\int u\,d\mu(u)\), and its lottery comparisons are complete; that removes the paper’s central subject, incompleteness. A threshold or quantile rule could retain some incomparability, but it would add a social aggregation rule the authors neither characterize nor analyze. The negative case is strongest on the absence of an anchored computational question: it cannot prove that no useful algorithmic project could ever be inspired by the theorem. Such a project would need to supply a new population model and computational task, rather than mirror a result this paper establishes.
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.