| paper | d4-conf-aaai-2015-0446 |
| authors | — |
| venue | ? |
| filed under | unclassified |
| judged by | gpt-6-luna / xhigh (triple__luna__xhigh__c6r1) |
| judge confidence | high |
| authors would recognise it | yes |
Theorems 1 and 2 are representation results, while the propositions establish properties of the proposed utilities; none states an algorithm, computational formulation, or complexity result. The population version of the egalitarian score is a plausible extension, but the proposed menu-selection computation is not a result of this paper. The proponent concedes the missing computational anchor, so bit (a) fails.
fails bit a — no named computational result to mirror
The proposed lift covers evaluation and selection by the pessimistic egalitarian score; it does not continue the axiomatic claims in Theorems 1 and 2 or the paper's other propositions as computational results.
The strongest positive lead is the paper’s egalitarian representation, but it does not meet your anchor requirement: this paper contains no named computational result. Theorem 2, proved here, says that if the collective and individual preferences satisfy Axioms 1–6, Pareto unanimity, continuity on consequences, E, and PW, then there are a totally ordered scale \(U^{\mathrm{ext}}\), utility functions \(u_i:X\to U^{\mathrm{ext}}\), and an order-reversing function \(n\) such that lotteries are ranked by
\[
\min_{x\in X}\max\!\left(n(L[x]),\min_{i\in A}u_i(x)\right).
\]
That is an axiomatic representation theorem, not a result about complexity or an algorithm the paper computes with. Theorem 1 is also axiomatic; the propositions establish properties of the proposed utilities. None qualifies as a computational anchor under your instructions.
There is nevertheless a plausible, unanchored population mirror. Imagine a large public consultation with millions of residents but only \(\tau\) distinct utility types. A type records the resident’s complete utility function over a finite set of consequences; \(\mu_t\) is the fraction of residents of type \(t\). Keep the paper’s possibilistic lotteries and its finite consequence set. For a finite menu of lotteries, define
\[
U^-_\mu(L)=\min_{x\in X}\max\!\left(n(L[x]),\min_{t:\mu_t>0}u_t(x)\right),
\]
and choose a menu lottery maximizing \(U^-_\mu(L)\). This is a precise continuous-population version of the paper’s pessimistic egalitarian evaluation. For a finite menu and explicitly listed types, it is directly computable in time polynomial in the menu size, \(\tau\), and \(|X|\).
The population regime is credible: people can share the same decision-relevant utility type, and millions of residents can be represented by a short list of types and their population shares. A reader could plausibly recognize the score as the paper’s egalitarian collective utility applied to a continuum of residents.
The weakness is decisive for this assignment: the score depends on \(\mu\) only through which types have positive mass. The amount of mass assigned to each type does not affect a worst-off-agent criterion. And selecting from an explicit finite menu is an easy evaluation problem, not one of the paper’s computational results. So this candidate may motivate a future extension, but it cannot honestly be presented as a mirror anchored in a named result of this paper. The paper’s stated gap is axiomatic characterization under possibilistic uncertainty; it supplies no discrete computational hardness whose fate in a continuous population can be studied.
The decisive gap is that this paper gives no computational result to mirror. Theorem 1 and Theorem 2 characterize collective preferences from axioms; the propositions establish properties of the proposed utilities. None defines a computational problem or establishes its complexity. A continuous population could inspire a separate computational study, but that would not continue a complexity landscape this paper began.
The public-consultation scenario is plausible, and grouping residents by complete utility type is faithful to high multiplicity. But the paper’s pure egalitarian rule takes the least-satisfied agent, so a type’s population share has no effect. Its natural continuum form is
\[
U^-_\mu(L)=\min_{x\in X}\max\!\left(n(L[x]),\operatorname{ess\,inf}_{t\sim\mu}u_t(x)\right).
\]
With finitely many types, this depends only on which types have positive mass. A tiny positive subgroup counts as much as a large one; a type disappears from the calculation when its mass reaches zero. That captures worst-off welfare, but it leaves population proportions computationally idle. The paper’s weights encode agents’ ethical importance, so interpreting them as population shares would change the welfare criterion.
The best computational version remains simple. For an explicit menu of lotteries, evaluate each by scanning its outcomes and the \(\tau\) types. If the task is to maximize the pessimistic egalitarian utility over all normalized possibilistic lotteries on \(X\), a deterministic lottery at an outcome maximizing \(\operatorname{ess\,inf}_{t\sim\mu}u_t(x)\) is optimal: every normalized lottery assigns possibility \(1\) to some outcome and therefore cannot exceed that outcome’s worst-type utility. Hardness could enter through a succinctly represented, constrained outcome space, but then the computational burden comes from that external outcome-generation problem. The same objective and burden remain with one representative of each type.
The other named results do not supply a stronger anchor. Proposition 1’s ex-ante/ex-post equality and Theorem 2’s representation remain statements about lottery rankings and axioms; replacing agents by types simply replaces a finite minimum with an essential infimum. A richer continuum of consequences or utility functions could support a new axiomatic theorem, but it would still be an axiomatic project, outside ChoCo’s computational scope.
I cannot rule out every future application. A mass-sensitive equity criterion or a combinatorial choice problem over a huge outcome space might be worthwhile alongside this paper. But each would add the computational or population-sensitive structure that the paper itself lacks. The proponent’s candidate is therefore a credible modeling exercise, not a worthwhile continuous mirror of a result in this paper.
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.