| paper | d4-conf-wine-2015-0012 |
| authors | — |
| venue | ? |
| filed under | unclassified |
| judged by | gpt-6-luna / xhigh (triple__luna__xhigh__c6r1) |
| judge confidence | high |
| authors would recognise it | unclear |
Theorem 3 rules out mechanisms satisfying the paper’s stability criterion for a finite-agent prior, but it does not state an algorithmic task or a computational complexity result. The proposed mass-contract model is a plausible adjacent question, but it changes individual bilateral deviations into positive-mass contracts and introduces shared cohort preference shocks.
fails bit a — no named computational result to mirror
Theorem 3 is an impossibility result about mechanisms, not a result about computing or the hardness of computing; the mass-flow decision problem is newly formulated.
fatal: True
The proposed mass version tracks Theorem 3 and Example 3; it leaves Theorems 1 and 2 and computational complexity questions unaddressed.
The honest limit is that this paper has no named computational-complexity result: it does not classify a problem as polynomial-time solvable, NP-hard, or parameterized tractable. Its formal results concern whether stable matching mechanisms exist. The strongest basis for a continuous mirror is Theorem 3 (proved here): no mechanism is ex-ante pairwise stable. In full, no mechanism is immune, across all common priors over preference profiles, to a blocking contract between at most one man and one woman, where both strictly prefer the contract’s partner-rank lottery by first-order stochastic dominance. The proof gives a stronger, fixed-prior example: for the prior in Example 3, no mechanism is ex-ante pairwise stable at that prior.
A concrete continuous question is Ex-Ante Pairwise Stability for Matching Masses. Take three cohorts on each side, each of mass \(1/3\), with complete rankings over the opposite-side cohorts and the unmatched option. A finite rational prior \(\Psi\) specifies possible realized ranking profiles. In the mass-lift of Example 3, cohorts \(M_1\) and \(W_1\) have the paper’s three-ranking lotteries, independently, with probabilities \(1-2p,p,p\), for rational \(p\in(0,1/4)\); the other cohorts have the fixed rankings in that example. Members of a cohort share its realized ranking in each state. Thus the population can be arbitrarily large while each realized society has only three preference types per side.
A mechanism maps each realized mass profile to a feasible matching flow \(x^\omega_{ij}\): the mass matched from male type \(i\) to female type \(j\), with row and column totals bounded by the available masses. The flow gives each unit of mass its lottery over partner types. The instance asks whether there is a mechanism for which no positive mass \(\delta\) from a male cohort and an equal mass from a female cohort can contract in advance to match each other, with both sides’ ex-ante rank lotteries strictly improved in the first-order stochastic-dominance sense. The solution is either such a mechanism or a blocking cohort pair and contract. This is an existence decision problem, not an optimization problem.
The lift preserves the paper’s central ingredients: preferences are uncertain when contracts are considered; the clearinghouse acts after preferences are realized; and agents compare partner lotteries ordinally by stochastic dominance. It also has a credible high-multiplicity setting: large cohorts of near-identical graduates and positions, with common market or policy shocks affecting each cohort’s preferences. The finite prior and cohort masses are rational inputs, and the number of types stays small as the population grows.
The expected answer is no for the lifted Example 3 instance. In the paper, any ex-ante pairwise-stable mechanism would have to give \(M_1\) first, second, and third choices with probabilities \(1-3p+4p^2,p,2p-4p^2\). Contracting \(M_1\) and \(W_1\) gives each first, second, and third choices with probabilities \(1-2p,p,p\), which strictly first-order stochastically dominates the mechanism’s lottery for \(p\in(0,1/4)\). In the mass model these are per-capita probabilities, so the contract can be made between equal positive masses of the two cohorts. The expected outcome is that this obstruction survives the continuum—a B-shaped preservation of the finite result, not a hardness classification.
The main weakness is that the paper’s blocking pair is two named individuals, while an atomless population gives any individual zero mass. This mirror therefore makes the co-modification explicit: the block is a contract between positive masses of two homogeneous cohorts. The common preference shock within each cohort also differs from the independent individual draws in Example 3. That is plausible for some labor-market cohorts and is consistent with the paper’s general allowance for priors, but the paper does not prove the mass version. A proof would need to show that the finite argument survives when matchings are mass flows and contracts involve cohorts.
I would keep the scope to Theorem 3. Theorem 2’s interim-stability result involves agents learning from others’ willingness to contract and a larger coalition; translating that information and coordination structure into mass types would require more modeling choices. Further questions for the proposed mirror are whether the mass-flow version can be characterized for general priors, how its decision complexity depends on the number of types and prior support, and whether the result persists under independent individual preference draws within cohorts.
The strongest negative is that the paper has no named computational result to continue. Theorem 3 proves a structural nonexistence claim: for a particular prior, no mechanism is ex-ante pairwise stable. It does not classify the complexity of an input problem or give an algorithmic task over societies. The proposed “Ex-Ante Pairwise Stability for Matching Masses” defines a possible decision problem, but that problem is an extension of the paper, not a continuous mirror of a computational result in it. To make it a complexity project, one would need to specify and study a broader input family of priors and mass types—work the paper itself does not set up.
The strongest mass lift also changes substantive primitives. In Example 3, individual preferences are drawn independently, and a blocking contract is between two named people. Replicating that market to high multiplicity makes an individual pair negligible in the population. The proposed lift avoids that by giving whole cohorts shared preference shocks and allowing positive masses from two cohorts to contract. That preserves a meaningful aggregate lottery, but it replaces independent individual uncertainty with group-level correlation and replaces a bilateral contract with a mass deviation. Neither change follows just from normalizing population size; together they define a different matching model.
A more careful version could keep bilateral contracts and represent a positive mass of them as a continuum of pairings between cohorts. That is plausible, and may be worth studying. But it needs a new account of how pairs are selected before preferences are known, how the realized preferences of contract participants are distributed, and how identity-specific partner rankings are represented by type-level flows. The finite theorem does not settle those choices. Keeping only the paper’s three-by-three example gives a population-scaled illustration, not an algorithmic study; broadening it enough to pose a computational question creates a new model whose connection to Theorem 3 must be established.
So the negative case is strongest at the programme boundary: this paper offers no computational-complexity anchor, and its best proposed mirror depends on new assumptions about correlation and contracting. It is weaker as a universal claim. A well-designed continuous matching model with mass contracts could be a worthwhile adjacent problem, but the paper alone does not make that a compelling ChoCo mirror.
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.