| paper | d4-journals-geb-2019-0142 |
| authors | — |
| venue | ? |
| filed under | unclassified |
| judged by | gpt-6-luna / xhigh (triple__luna__xhigh__c6r1) |
| judge confidence | medium |
| authors would recognise it | unclear |
Theorem 3.2 gives a constructive formula, so the paper does contain a named computational result despite the proponent’s claim that none qualifies. But the suggested continuous version changes the arbitrary coalition-value input into a function over type masses and asks about a limiting allocation, without specifying a computational task that mirrors the anchored result. No continuous computational question survives in the cases as stated.
fails bit b-changes — the continuum limit changes the question
The suggested type-mass limit would address the Shapley allocation result in Theorem 3.1 and the formula in Theorem 3.2; it leaves the full component-game decomposition, weighted variants, and restricted-cooperation results without a continuous computational formulation.
I don’t see a qualifying anchor in this paper. It discusses computation, but none of its numbered Theorems, Lemmas, Corollaries, or Propositions asserts a complexity classification.
The closest results are Theorem 3.1 (proved here), which characterizes the component games and shows their grand-coalition values equal the Shapley values; Proposition 2 (proved here), which characterizes each component game as the unique normalized solution of \(L v_i=L_i v\); and Theorem 3.2 (proved here), which gives an explicit formula for those components. These are mathematical characterizations and formulas, not claims that a computational problem is in P or hard.
The paper does state in §2.1 that computing the Shapley value is #P-complete, but that statement is unnumbered and attributed to Deng and Papadimitriou (1994). A continuous, type-based Shapley question might be worth studying, but presenting it as a mirror of a qualifying named result would violate the anchor requirement. That is the limit of the positive case here.
The strongest case against a mirror is that the paper proves no computational result of its own. Theorem 3.1 characterizes the component games and recovers the Shapley value from axioms; Proposition 2 characterizes those components through a Laplacian system; Theorem 3.2 gives an explicit formula. None classifies the complexity of a problem. The paper’s statement that Shapley computation is \(\#\mathrm{P}\)-complete is attributed to Deng and Papadimitriou, not established here. That makes this a poor target for the programme’s paper-by-paper search for computational mirrors.
The best population version of Theorem 3.1 would group interchangeable players into types, describe coalition values by a function \(V\) of type masses, and ask whether per-player Shapley values converge to a continuum allocation. That is a coherent question, but it is a limit or characterization question, not a computational mirror of Theorem 3.1. The normalization from individual payoff to payoff density is manageable; it does not change that mismatch.
Proposition 2 and Theorem 3.2 offer a more computational-looking route: study a limit of the finite coalition-count Laplacian, or of the Green’s-function formula, and ask how to compute the resulting allocation. But the paper’s game is an arbitrary table of \(2^n\) coalition values. A continuous version needs a different input, such as a function \(V\) over type-mass vectors. Whether that function is given by a table, formula, circuit, or oracle changes the complexity question. Choosing one could yield worthwhile new work, but the paper supplies no canonical continuous input model or complexity claim for it to mirror.
The cited \(\#\mathrm{P}\)-completeness result is the strongest possible starting point for a new project on high-multiplicity Shapley computation. Yet a type-count representation and its complexity would have to be specified independently; it does not follow from the paper’s Hodge decomposition results. I would therefore vote against prioritizing this paper as a continuous-mirror target. The negative case is strongest as a judgement about this paper’s contribution, not as a proof that no worthwhile continuous-population computational project on cooperative games could be designed.
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.