| paper | d4-conf-ijcai-2017-0525 |
| authors | — |
| venue | ? |
| filed under | unclassified |
| judged by | gpt-6-luna / xhigh (triple__luna__xhigh__c6r1) |
| judge confidence | high |
| authors would recognise it | unclear |
Theorem 1 rules out mechanisms satisfying participation; it says nothing about computing or computational hardness. The proposed mass-flow policy-synthesis task is a plausible extension of online organ matching, but it adds a new computational problem rather than supplying a computational result from this paper.
fails bit a — no named computational result to mirror
The proposed question concerns only Theorem 1's participation property. It leaves the MIN-versus-BOX experiments and their welfare, waiting-time, swap, and delay-manipulation measurements as empirical findings; the real-valued score scales and stochastic arrivals do not themselves make the population continuous.
The strongest honest positive case is a narrow one about online mechanism design. The paper’s only numbered result is Theorem 1, proved here: *no mechanism for online organ matching satisfies participation.* Its witness is a three-step instance: two patients join the waiting list first; a kidney with KDPI \(50\) arrives second, followed by one with KDPI \(0\). Each patient has an incentive to delay joining until after the first kidney is allocated.
That is a real formal result about an online allocation mechanism, but it is not a named computational-complexity result: it establishes no polynomial-time algorithm, hardness, or parameterized classification. Under your anchor requirement, the paper therefore has no eligible named computational anchor. MIN and BOX are named mechanisms, but their comparison is supported by experiments, not a numbered complexity result.
The closest continuous question I can defend is Continuous Online Organ Matching Participation Feasibility. Consider a large national or cross-regional transplant pool. Patient types record the features the paper uses: blood type, EPTS, and region; kidney types record blood type, KDPI, and region. At each time step, rational masses of patient types arrive or depart, and rational masses of kidney types arrive. The type set is finite, with many patients per type and far fewer types than patients. A mechanism sees only the arrivals so far and chooses irrevocable match flows \(x_{h,t,k}\): mass of currently waiting type-\(t\) patients assigned at history \(h\) to arriving kidney type \(k\). It must obey compatibility, supply, and waiting-list constraints. Each patient’s utility for a kidney of KDPI \(d\) is \(100-d\); unmatched utility must also be specified.
The question is whether there exists a nonanticipating mechanism that is participation-safe: for every patient, joining at the stated time gives at least as high expected utility as withholding entry until a later time. A solution is such a policy over the possible arrival histories; a negative answer should give a finite arrival scenario and a profitable delay deviation. To make this an actual computational problem, one could give the allowable arrivals as a finite scenario tree and ask for a policy, or a certificate that none exists, with running time measured in the tree size, number of types, and rational input bit length.
This is a plausible population mirror because the society is represented by masses of patients sharing the model’s relevant attributes, while the mechanism still makes online matching decisions. A plausible high-multiplicity setting would need a much larger pool than the paper’s Australian example of roughly \(1{,}200\) wait-listed patients and \(340\) deceased donors a year—perhaps a large cross-regional pool with repeated blood-type, EPTS-band, and region cohorts. Kidney arrivals would also need aggregation into batches large enough for mass flows to be meaningful. That is a constructed regime, and the paper itself does not establish that it is operationally available.
For the outcome, I would expect the participation impossibility to persist if the mirror preserves tagged individual delay deviations: the paper’s short timing witness may be expressible with rational cohort masses. But that is an expectation to prove, not a hardness result. In a strictly atomless fluid model, an individual’s deviation has zero mass and may not affect the allocation; the original witness could disappear unless the model defines how an individual’s expected allocation is recovered from type-level flows.
The proposed mirror covers only Theorem 1. It does not cover the empirical MIN-versus-BOX comparison or turn the paper’s experiments into complexity results. Its weakest point is also decisive under the stated rubric: the anchor is an axiomatic mechanism impossibility, not a computational classification. So this is the paper’s best nearby continuous question, but it does not supply a qualifying ChoCo anchor as specified. Further questions are whether the delay witness survives the fluid limit, whether person-level incentives can be recovered from aggregate flows, and how hard it is to synthesize or verify participation-safe policies on finite scenario trees.
The strongest case against a ChoCo mirror is that the paper has no numbered computational result to mirror. Its sole theorem proves an axiomatic impossibility: no online mechanism satisfies participation. MIN and BOX are fixed mechanisms compared in experiments; the paper gives no complexity classification, reduction, or algorithmic theorem. The proponent’s proposed participation anchor therefore fails the programme’s stated gate.
A finite-tree version of policy synthesis could be a useful computational problem, but it changes the question. The theorem quantifies over all online mechanisms and already gives a small witness to impossibility. Synthesis on a specified tree, perhaps with restrictions on the mechanisms or a welfare objective, asks a new online optimization question. Requiring a positive-mass cohort to delay also changes individual participation into a coalition property. A tagged infinitesimal patient could preserve the individual incentive question alongside a continuous population, so I would not claim the mirror is incoherent; it still would not supply the missing computational result from this paper.
There is a credible adjacent direction: model large pools as masses of patient and kidney types, then study online allocation policies. The paper’s attributes and experiments give that idea a foothold. But that would make this a starting point for a new computational project, rather than a mirror of one of its named results.
So the negative case is strong for rejecting this paper as a ChoCo mirror target under the stated anchor requirement. The broader claim that no continuous question inspired by it could be worthwhile is weaker: population-based online policy optimization could plausibly be worthwhile, and the paper alone cannot rule it out.
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.