| paper | d4-journals-geb-2016-0075 |
| authors | — |
| venue | ? |
| filed under | unclassified |
| judged by | gpt-6-luna / xhigh (triple__luna__xhigh__c6r1) |
| judge confidence | medium |
| authors would recognise it | yes |
other→P discrete result → its continuous mirror
Answer: The stated mass procedure always terminates with a fair-with-diversity flow maximal under \(>_q\). Computing such a flow is polynomial-time solvable, although the stated procedure can require exponentially many rounds. · discrete side: other · adjudication: proved_with_minor_gaps
full proof & adversarial review →
Theorem 1
statement extracted from the paper’s text layer
Given finite schools \(C\), finite student types \(T\), rational type masses \(\mu_t\), per-capita capacities and diversity targets, and school priorities over tiers, does a school-proposing deferred-acceptance procedure on mass—allowing a type’s mass to split at a priority-tier cutoff—compute an assignment flow \(x_{tc}\) (including an outside option) that is fair-with-diversity and maximal under the mass version of \(>_q\)?
A unit-mass population with finite support over types that include demographic category, school preferences, and priority tiers; schools have per-capita capacities and reserve targets. The decision is a rational assignment flow \(x_{tc}\) with each type’s mass fully assigned, no capacity exceeded, no positive mass able to block under the tiered reserve choice rule, and maximal diversity among such flows.
Strict rankings over named students make complete priority profiles unique, and the case does not establish that the authors would accept the priority-tier choice rule needed to aggregate students.
fatal: False
The mirror covers Theorem 1’s fair-with-maximal-diversity guarantee; it leaves Theorem 2’s segregation result and the DAMR comparisons unaddressed.
On a strict reading of your anchor rule, this paper has no qualifying computational result. The closest is Theorem 1, proved here: the SPDiv mechanism produces an assignment that is fair-with-diversity and has maximal diversity among such assignments. Theorem 2, also proved here, says that SPDiv minimizes segregation when the diversity objectives mirror the population distribution. Both are mechanism guarantees, not results classifying a computational problem as tractable or hard. The paper gives an explicit deferred-acceptance procedure, but it does not establish a complexity result for computing its outcome. So I can’t honestly present either theorem as a compliant anchor.
The strongest positive case I can make is a candidate mirror of the problem behind Theorem 1. A plausible regime is a large district assigning an annual cohort: many students, but far fewer realized combinations of relevant attributes. A type would include a student’s demographic category, school preferences, and school-priority characteristics. The mass \(\mu_t\) would be that type’s fraction of the cohort. Capacities and diversity targets would be expressed as fractions of the cohort as well.
The continuous computation question would be: given finite schools and types, rational type masses, per-capita capacities and diversity targets, and reserve-based school choice rules, compute a rational assignment flow \(x_{tc}\), including an outside option, such that \(\sum_c x_{tc}=\mu_t\), no school exceeds capacity, and the assignment is fair-with-diversity and maximal with respect to the paper’s diversity partial order. Here, fairness means that no positive mass of students prefers a school and would be admitted under its reserve choice rule. Maximal diversity means that no other fair-with-diversity flow can increase mass in a currently under-target type–school pair without reducing mass in any such pair. This retains the paper’s central question: how to balance school priorities and diversity objectives in the selected assignment.
I would expect this finite-support version to be tractable through a compressed school-proposing deferred-acceptance procedure, but that is a conjecture for the mirror, not a conclusion of Theorem 1. It also raises useful follow-up questions: can the computation be polynomial in the number of schools, support size, and input bit length; how closely can its flow be rounded to a finite assignment; and does the segregation guarantee of Theorem 2 survive that rounding?
The weak point is substantial: the paper assumes strict priority rankings over individual students, whereas the mirror groups students with identical relevant priority characteristics and permits fractional mass to be accepted. That type-level choice rule needs to be defined and analyzed; the paper does not do so. The proposed problem is recognizably adjacent to SPDiv, but it is a new computational question rather than a continuous counterpart to a named complexity result in this paper.
The decisive weakness is that there is no named computational result here to mirror. Theorem 1 says that SPDiv’s outcome is fair-with-maximal-diversity; Theorem 2 gives a segregation guarantee. Both are claims about the properties of assignments, not about the complexity of a computational problem. The paper does present a deferred-acceptance procedure, but its result about that procedure is not a complexity classification. The proposed flow problem would be a new computational question built around Theorem 1, rather than a continuous counterpart to a computational result the paper establishes.
The strongest version of that proposed mirror still runs into a problem with the paper’s model. Schools have strict rankings over named students, and those rankings determine who fills reserved seats. Students who differ in their place in a school’s ranking can be treated differently, so a complete type must include that priority information. Under strict rankings, each student has a distinct place at every school. Keeping those priorities therefore leaves no repeated complete types to aggregate; dropping them requires replacing strict priorities with priority tiers or another coarser rule. That may be a sensible school-choice model, but the paper does not supply it, and it changes the priority structure its reserve choice function uses.
The proposer’s alternative—compute any fractional assignment that is fair-with-diversity and maximal—makes the issue clearer. It is a plausible new problem, but it need not compute the continuous counterpart of SPDiv’s output: it asks for any assignment satisfying the translated properties. A closer counterpart would define school-proposing deferred acceptance over masses, including how a school’s strict priority order selects part of a type. The paper gives no basis for that fractional selection rule.
This is a real case against a mirror of this paper’s results, but the broader claim that no worthwhile continuous school-choice problem exists in any scenario is weak. A district with many students in shared preference and priority tiers is a plausible high-multiplicity regime, and the resulting fractional mechanism may be worth studying if the authors would recognize the tiered-priority model as theirs. The paper does not establish that connection, though, and the proponent’s flow formulation cannot fill the gap by itself.
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.