| paper | d4-journals-mss-2005-0015 |
| authors | — |
| venue | ? |
| filed under | unclassified |
| judged by | gpt-6-luna / xhigh (triple__luna__xhigh__c6r1) |
| judge confidence | high |
| authors would recognise it | no |
The paper's theorems characterize which preference axioms yield Choquet integral representations; none states an algorithm, complexity bound, or computational hardness result. A continuum of recipients could motivate an axiomatic extension, but turning it into optimization or learning would add a computational problem the paper does not study.
fails bit a — no named computational result to mirror
The paper's named results characterize preference relations and capacities; none supplies a computational result for a continuous mirror to cover.
I can’t make a positive case that meets your anchor rule: the paper has no named computational result asserting tractability, hardness, fixed-parameter tractability, or a similar complexity claim. Its observation that a general capacity needs \(2^n\) values motivates restricting attention to \(k\)-additive capacities, but the paper does not analyze the complexity of a computational problem.
The closest result is Theorem 10, proved in the paper: a preference relation satisfies A1, A2, A4, A5, A6, and A9 if and only if it has a unique representation as the Choquet integral of a 2-additive capacity. That is an axiomatic characterization, not an algorithmic result, so it cannot serve as an anchor under your rules.
There is a plausible continuous-population analogy: treat \(X\) as a continuum of welfare recipients, let \(f\) be a measurable income assignment, and evaluate it with a capacity-based Choquet integral. In the symmetric case, the value could depend on the distribution of incomes rather than recipients’ identities. But that would continue the paper’s axiomatic question, not formulate a computational problem it studied. Indeed, Theorem 7 cites Schmeidler’s representation result for finite or infinite \(X\). A task such as optimizing redistribution or learning a capacity would add a decision problem absent from this paper.
So there is no honest named-anchor continuous computational question to hand to a prover, and no basis here to predict whether such a question would be tractable or hard.
The strongest negative is that this paper contains no named computational result to mirror. Theorem 10 is the closest candidate, but it characterizes which axioms on preferences yield a Choquet integral with a 2-additive capacity. It gives no computational problem, algorithm, or complexity bound. The same is true of the other theorems. The introduction’s observation that a general capacity has \(2^n\) coefficients motivates restricting the model; it is not a complexity result.
The most favorable population reading is to treat \(X\) as welfare recipients and extend income profiles and capacities to a continuum. That could lead to a worthwhile axiomatic or analytic question. But the paper’s results ask which preference axioms characterize a representation; they do not ask how to compute, optimize, or manipulate one. The cited infinite-domain result of Schmeidler is relevant background for such an axiomatic extension, not an existing answer to a computational question. Turning the model into optimization or learning would require adding a new task and its input model. Even a carefully designed mass-level version of A9 would not create a computational result in this paper to mirror.
So the proponent raised no surviving computational anchor: Theorem 10 does not qualify. This is a strong case against a *continuous computational mirror of this paper*. The broader claim that no scenario could ever yield worthwhile research is weaker: one could build a new computational programme around Choquet welfare models, but it would be a new programme rather than a mirror of the results here.
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.