| paper | d4-journals-jet-2006-0029 |
| authors | — |
| venue | ? |
| filed under | unclassified |
| judged by | gpt-6-luna / xhigh (triple__luna__xhigh__c6r1) |
| judge confidence | high |
| authors would recognise it | yes |
The paper’s numbered Results 1–7 report experimental or simulation findings; none is a numbered Theorem, Lemma, Corollary, or Proposition asserting an algorithm, guarantee, or hardness result. A large-district fixed-report allocation limit is a coherent adjacent question, but it cannot supply the missing computational anchor, and it leaves the behavioral response central to the observed efficiency comparison unspecified.
fails bit a — no named computational result to mirror
The proposed limit addresses the allocation and welfare comparison in RESULT 6 conditional on a supplied report distribution; it does not model the preference manipulation and behavioral findings in Results 1–5 or the behavior-calibrated simulations in RESULT 7.
The honest positive case is unanchored: this paper has no numbered Theorem, Lemma, Corollary, or Proposition asserting a computational complexity result. Its numbered Results 1–7 report experiments and simulations. The closest is Result 6: in the designed environment, efficiency ranks \(GS>TTC>BOS\); in the random environment, \(GS\sim BOS>TTC\); and all three mechanisms outperform No Choice in both environments. That is an empirical finding, not a result about an underlying problem being in P or hard. The strategy-proofness and efficiency properties discussed in the paper are attributed to earlier work. So there are no eligible anchors to mirror under the stated rule.
There is still a plausible, narrower continuous question in the paper’s subject matter: Limit School-Choice Outcome. Consider a large district assigning a cohort to schools \(S\), with per-capita school capacities \(q_s\), where \(\sum_s q_s=1\). Each type \(t\) has mass \(\mu_t\), with \(\sum_t\mu_t=1\), and records its true ranking, submitted ranking, district school, and utility \(u_t(s)\). For a mechanism \(M\in\{BOS,GS,TTC\}\), realize each rational instance as finite populations of size \(N\) with counts \(N\mu_t\) and capacities \(Nq_s\), run the mechanism with the paper’s lottery rules, and take the expected assigned mass \(x^N_{t,s}\) over lotteries.
The question is whether \(x_{t,s}=\lim_{N\to\infty}x^N_{t,s}\) exists and can be computed: given \(\varepsilon>0\), return an assignment within \(\varepsilon\) of \(x\), and its expected true welfare \(W=\sum_{t,s}x_{t,s}u_t(s)\). Here \(x_{t,s}\) is the mass of type \(t\) assigned to school \(s\); each student still receives one whole seat, while a type’s aggregate assignment can be fractional. This co-modification of capacities and type-level assignments is part of the population mirror.
A credible regime is a large urban district’s annual admissions cohort: \(N\) is large, while the number of schools, catchment areas, rankings, and modeled utility categories is bounded, so many students share a complete type. The paper’s 36-student laboratory sessions are not such a regime; its distinct payoff tables make that clear. For the fixed-report outcome question, I would expect tractability (Class A), though the paper proves nothing about this limit. Further questions include how quickly finite-population outcomes approach the limit and whether a continuum model of strategic reporting can reproduce the paper’s manipulation patterns.
The scope of this candidate is aggregate allocation and welfare under specified reports, especially the outcomes measured in Result 6. It does not explain why subjects manipulate, or establish a continuum analogue of their behavior. That is its weakest point: it is a sensible high-multiplicity school-assignment question, but it is not a continuous version of a named computational result in this paper.
The proponent’s strongest point is also the limit of the case: the paper has no numbered computational result to mirror. Results 1–7 are experimental and simulation findings; the mechanism properties come from cited theory. Result 6 is a real finding, but it reports efficiency rankings under subjects’ choices. It does not pose an input-output problem or analyze its computational complexity.
The proposed fixed-report limit is a coherent new problem. A large school district is a credible high-multiplicity setting, and grouping students by complete preference and payoff types is legitimate. I would not reject the mirror because school assignments are indivisible, capacities must scale, or lotteries need a continuous treatment. Those are natural co-modifications. Nor does the fact that the paper’s experiments used only 36 students dispose of the annual-cohort scenario.
The harder objection is that this limit computes allocations conditional on a specified distribution of reports. It does not continue Result 6’s central comparison of mechanisms *as implemented among boundedly rational people*. To model that comparison, one needs a rule predicting reports from private preferences, beliefs, and understanding of each mechanism. A nonatomic model can preserve an individual’s incentives: a student’s report may change their own assignment probability even when it has no effect on aggregate mass. But the paper supplies no behavioral response model for those incentives. Assuming rational equilibrium would add a different behavioral premise from the one the experiment investigates; reproducing its observed departures from truth-telling would require further assumptions about bounded rationality or education. The resulting predictions would depend on those additions, not on continuizing the paper’s findings.
One could make a worthwhile ChoCo project of computing expected mass assignments under Boston, Gale–Shapley, and TTC for a supplied type distribution, or of building and analyzing a continuum model of strategic reporting. But the first is an adjacent algorithmic study of these mechanisms, not a computational mirror of the paper’s empirical result. The second adds a behavioral theory the paper does not give. My negative case is therefore strongest on the narrow claim that this paper supplies a computational anchor; it is weak as a claim that no worthwhile continuous school-choice question can be built from its subject matter.
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.