| paper | d4-journals-mss-2019-0014 |
| authors | — |
| venue | ? |
| filed under | unclassified |
| judged by | gpt-6-luna / xhigh (triple__luna__xhigh__c6r1) |
| judge confidence | high |
| authors would recognise it | yes |
The proposed high-multiplicity SEEC computation is a coherent follow-on, but it adds an algorithmic problem around the paper’s characterization. Theorem 6 characterizes the allocation form and Theorem 7 establishes uniqueness; the explicit proportional-transfer formula is also presented as a solution characterization, without an algorithmic guarantee, hardness result, or computational formulation. The required computational-result bit therefore fails.
fails bit a — no named computational result to mirror
The proposed computation would cover the SEEC transfer vector and its induced schedule from Theorems 6–7; it leaves the paper’s other solution characterizations, fairness generalization, and transfer comparisons aside.
The strongest positive case is a new high-multiplicity computation problem built around the paper’s stochastic egalitarian constrained equal costs rule (SEEC). The paper does not, however, contain a named computational result of the kind your anchor requirement calls for.
The closest results are Theorem 6 and Theorem 7, both proved in the paper. Theorem 6 says that a solution has Local Symmetry—every unit increase in cost is shared equally among agents whose liabilities are not yet reached—if and only if it has the constrained equal costs form: an ex ante transfer plus the constrained equal awards allocation of realized cost against the remaining liabilities. Theorem 7 says there is a unique egalitarian solution with Local Symmetry, namely SEEC, whose ex ante transfers are chosen to make all agents’ disutilities equal. These are characterization and uniqueness results, not results that a problem is in P, hard, or encoded and computed. They therefore are not qualifying computational anchors.
A plausible follow-on is High-Multiplicity SEEC Computation. Consider a large firm with many divisions sharing a central service, where the total cost scales with the number of divisions. Each division has the same cost beliefs and preferences, and its type is its per-division liability \(\ell_j\). The society is a finite-support distribution \(\mu_j\) over liability types, with rational masses summing to \(1\), support size \(\tau\), and many divisions relative to \(\tau\). For a concrete version, let per-division cost \(C\) be uniform on \([0,\bar c]\), and let disutility be expected cost. Require \(\sum_j\mu_j\ell_j\ge\bar c\) and \(\min_j\ell_j\ge\bar c/2\), the per-capita versions of the paper’s admissibility and \(V\)-sufficiency conditions.
The question would be: compute transfers \(t_j\) and the resulting allocation schedule, where \(\sum_j\mu_jt_j=0\), \(t_j\le\ell_j\), and
\[
x_j(c)=t_j+\min\{\lambda(c),\ell_j-t_j\}.
\]
For each \(c\), \(\lambda(c)\) is chosen so that
\[
\sum_j\mu_j\min\{\lambda(c),\ell_j-t_j\}=c,
\]
and the transfers must make every type face the same expected cost:
\[
\frac{1}{\bar c}\int_0^{\bar c}x_j(c)\,dc=\frac{\bar c}{2}
\quad\text{for every }j.
\]
A solution is an exact transfer vector and its induced schedule, represented by its breakpoints and pieces. The paper’s Theorem 7 makes this a plausible computationalization of SEEC: the firm’s divisions can be grouped by complete type, and rational masses compactly describe many repeated divisions. I would tentatively expect tractability from the ordered, piecewise structure of the allocation, but the paper proves no running-time bound for this task.
The regime is recognizable in the authors’ own multi-divisional-firm setting: a large organization with many divisions sharing the service, and a small number of liability classes repeated across divisions. This is a population mirror, not a claim that the paper’s random cost is already a continuous society. The paper keeps a fixed finite set of agents; its continuity is in cost uncertainty.
So the positive case is for computationalizing a central result of the paper, not for claiming that it already reports a computational result. That is also the weakest point: under the requested anchor rule there are zero eligible anchors, and the proposed problem adds the algorithmic question the authors did not study.
The strongest negative point is categorical: the paper contains no named computational result to mirror. Theorem 6 characterizes Local Symmetry as the constrained equal costs form; Theorem 7 identifies the unique egalitarian solution satisfying that property. The other numbered results likewise concern existence, characterization, or transfer comparisons. None states a complexity bound or studies an algorithmic problem. By the proposed anchor standard, neither theorem is a qualifying anchor.
The proponent’s stronger version is coherent: take many divisions from finitely many liability classes, represent the classes by rational masses, and compute a mass-weighted SEEC schedule and its transfers. This does make the population continuous. The type can be just the liability when preferences are common, as in the paper; weighted transfers and type-level allocations preserve the relevant constraints. The paper’s stochastic cost is not itself a continuous population, but that does not invalidate this proposed population model.
What it does not supply is a computational result from this paper to mirror. It creates a new task—computing a rule that the paper characterizes axiomatically. That task might be worth studying, but its complexity, whether easy or hard, cannot turn Theorem 7 into a complexity theorem. So it is the best attempted anchor, but it falls outside the stated requirement.
I would not claim a stronger objection based on identity, scaling, or a degenerate limit: those do not defeat this model. The negative case is strong under the anchor requirement and weak as a universal claim that no worthwhile continuous follow-on exists. If “mirror” includes adding computation around an axiomatically characterized rule, the weighted SEEC proposal remains plausible, and I cannot honestly rule it out as worthwhile.
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.