| paper | Temporal Fairness in Multiwinner Voting |
| authors | — |
| venue | AAAI 2024 |
| filed under | multiwinner · multiwinner |
| judged by | gpt-5.6-luna / xhigh (triple__luna__xhigh__c2r1) |
| judge confidence | high |
| authors would recognise it | yes |
The paper fails bit (a): it is a framework and survey with no numbered theorem, lemma, corollary, or proposition asserting a computational result of its own. Therefore no result of this paper can supply an anchor under the strict screening rule. The proposed CTAR problem is nevertheless a plausible population-level continuation, so the negative case wins only on the named-result gate.
fails bit a — no named computational result to mirror
No named computational result is covered. The proposed continuation covers static approval types, offline temporal committee selection, candidate availability, usage caps, and frequency-based representation, while leaving the paper's dynamic, online, cardinal, ranked, axiomatic, and literature-summary material aside.
The paper has no qualifying anchor. It contains no numbered Theorem, Lemma, Corollary, or Proposition, and proves no computational result of its own. Its algorithmic statements are literature summaries—for example, that Bulteau et al. (2021) obtain proportionality with a simple greedy algorithm—but those results are neither numbered nor proved here. Thus there is no honest theorem number to quote and no anchor-specific mirror under the programme’s rules. Strictly, this paper fails the named-result gate.
The strongest positive case is therefore prospective rather than anchored. The paper’s framework does admit a natural population mirror. I would propose the following lead problem.
Call it Continuous Temporal Approval Representation, or CTAR\(_\infty\). An instance contains candidates \(P\), horizon \(\ell\), committee size \(k\), availability sets \(P_r\), usage caps \(\alpha_p\), and a finite set \(\mathcal T\subseteq 2^P\) of static approval types. Type \(t\) represents fraction \(\mu_t\) of the population, with rational \(\mu_t\ge 0\) and \(\sum_t\mu_t=1\). A solution is a sequence of committees \(K_1,\ldots,K_\ell\), where \(K_r\subseteq P_r\), \(|K_r|=k\), and each candidate \(p\) is used at most \(\alpha_p\) times.
Fix a mass threshold \(\theta\) and a window length \(\kappa\). A group \(G\subseteq\mathcal T\) is \(\theta\)-large and cohesive if
\[ \sum_{t\in G}\mu_t\ge\theta \quad\text{and}\quad \bigcap_{t\in G}t\neq\varnothing. \]
The question is whether there exists a committee sequence such that, for every such group \(G\), every block of \(\kappa\) consecutive rounds contains a selected candidate approved by every type in \(G\). The output is either such a sequence or NO. Equivalently, one may optimize the largest feasible \(\theta\) for a given \(\kappa\).
This is a faithful continuation of the paper’s static-preference, approval-ballot, offline, sequential-committee, availability/capacity, and frequency-based-fairness dimensions. Time and committees remain discrete; only the population is continuized. The authors should recognise it as their question about whether a sufficiently large group can be guaranteed representation within every bounded temporal window, expressed in population fractions rather than voter counts.
A plausible regime is a large professional society, platform, or public institution with millions of members but only dozens or hundreds of stable approval types—say, combinations of region, discipline, and policy priorities. Members of one type are indistinguishable for this problem, and \(n\gg\tau\). If preferences evolve, the complete temporal approval trajectory becomes the type; the static version is simply the cleanest first case.
I would expect the full CTAR\(_\infty\) problem to be Class B or possibly hard for continuum-specific reasons: the temporal packing and candidate-capacity combinatorics remain, and any discrete hard instance embeds through \(\mu_t=n_t/n\). Restricted cases may admit flow or dynamic-programming algorithms. Natural follow-ups are separation for the exponentially many cohesive-group constraints, parameterized complexity in \(\kappa\) or the number of types, and rounding between rational mass instances and finite electorates.
The weakest point is decisive: CTAR\(_\infty\) is a reasonable new problem suggested by the framework, but it is not a mirror of a named computational result in this paper. The paper is an agenda-setting survey, not an anchor-bearing computational paper. Thus the best positive conclusion is “the framework has a credible continuous population analogue,” not “this paper supplies evidence for a continuous mirror” in the programme’s strict screening sense.
The strongest negative case is decisive only at the screening level: this paper supplies no qualifying anchor. It contains no numbered theorem, lemma, corollary, or proposition establishing the complexity of a computational problem. Statements such as Bulteau et al.’s simple greedy algorithm are literature summaries, not results proved by this paper. The “does time hurt or heal?” discussion is an open agenda, not a computational theorem. CTAR\(_\infty\) would therefore be a new problem inspired by the paper, not a continuous mirror of one of its results.
The proposed mirror also has a defect as written. Feasibility becomes easier as \(\theta\) increases, so “optimizing the largest feasible \(\theta\)” is vacuous: \(\theta=1\) may impose no obligation at all if the whole population is not cohesive. A meaningful version would minimize the threshold that can be guaranteed, or maximize the mass of groups represented in every window.
That repair, however, defeats the stronger negative case rather than rescuing it. The repaired problem is a legitimate high-multiplicity continuation. Static approval types are explicitly within the paper’s framework; a large professional society with millions of members and a small number of stable approval profiles is a plausible regime; and temporal committees, availability, usage caps, and bounded-window representation remain substantive. The objective concerns aggregate representation, not individual identity, and mass does not disappear in the limit.
Nor can one object that the resulting complexity might be unchanged or that the problem might be easy: those are permissible outcomes under ChoCo. Nor is this “already done continuously”; the cited temporal work is discrete or axiomatic, not a computational treatment of a distribution over voter types.
Thus the honest negative verdict is narrower: the paper should be rejected as an anchor-bearing paper because it has no named computational result. But the universal substantive claim—that no worthwhile continuous mirror exists in any scenario—is weak and not credible. CTAR\(_\infty\), after correcting its threshold objective, survives the principal modelling objections; it is simply prospective rather than evidence supplied by this paper.
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.