| paper | d4-conf-ijcai-2021-0616 |
| authors | — |
| venue | ? |
| filed under | unclassified |
| judged by | gpt-6-luna / xhigh (triple__luna__xhigh__c6r1) |
| judge confidence | high |
| authors would recognise it | no |
The numbered results establish whether manipulation is possible or ruled out; they do not give an algorithm, computational bound, or computational formulation for finding a manipulation. Both cases identify no qualifying computational anchor, and the proposed cohort model changes the strategic actor from one agent to a group.
fails bit a — no named computational result to mirror
No computational result is covered. Proposition 1 and Theorems 1–5 classify strategic susceptibility or resistance, while the paper leaves the complexity of finding a manipulation for future work.
The qualifying anchor set is empty. The paper contains no numbered theorem, lemma, corollary, or proposition classifying a problem as NP-hard, polynomial-time solvable, W[1]-hard, FPT, or similar. Proposition 1 proves that Max-Util admits a strict-improvement manipulation by removing edges in an undirected network. Theorems 1–4 establish other resistance and susceptibility results. Those are strategic properties, not computational complexity classifications. The paper itself leaves the complexity of finding a manipulation as future work.
There is a plausible follow-on mirror suggested by Proposition 1, though it cannot serve as a required anchor. Let types \(T\) have rational masses \(\mu_t\), and let a symmetric matrix \(W\) give contact densities between types. An assignment \(y_{tq}\) gives the fraction of type \(t\) placed in coalition \(q\), with \(\sum_q y_{tq}=1\). A member of type \(t\) assigned to \(q\) has utility
\[
U_t(y;W)=\sum_{u\in T}\mu_u W_{tu}y_{uq}.
\]
The organizer chooses \(y\) to maximize
\[
\sum_{t,u\in T}\sum_{q=1}^k \mu_t\mu_u W_{tu}y_{tq}y_{uq}.
\]
A precise follow-on question is: given \((T,\mu,W,k)\) and a distinguished type \(r\), does there exist a symmetric report \(W'\) that only deletes contacts incident to \(r\), such that every Max-Util-optimal assignment under \(W'\) gives the \(r\)-cohort greater utility evaluated at the true \(W\) than every truthful optimum does? A solution would be such a report \(W'\), with the strict-improvement condition verified. A plausible regime is large school or workplace cohorts whose members share a contact profile; type mass represents the fraction of people in each cohort.
The soft spot is substantial: this makes the manipulator a positive-mass, coordinated cohort. A single infinitesimal agent’s effect on normalized utilitarian welfare vanishes in the limit. The cohort version is a sensible high-multiplicity research question, but it changes the paper’s one-agent manipulation problem, and I have no sound A/B/C prediction for it from this paper.
The anchor set is empty, and that is the strongest ground against a ChoCo mirror here. Proposition 1 and Theorems 1–4 classify when manipulation succeeds or is impossible; none classifies the complexity of finding a manipulation. The paper explicitly leaves that complexity question for future work. So there is no computational result in this paper whose population-continuous version the programme could chart.
The best candidate is a stronger version of the proposed Max-Util follow-on: represent a large society by blocks of agents with the same contact profile, use a symmetric contact-density matrix \(W\), and ask whether a positive-mass block can report fewer contacts so that every resulting welfare-optimal coalition assignment improves its members’ true utility. That is coherent for homogeneous cohorts, and the block’s positive mass gives its report a real effect on aggregate welfare.
Its strategic actor is a cohort, though. In the paper, one named agent changes their own reported edges. In an atomless continuum, that agent has measure zero: changing their row of the contact kernel does not change the organizer’s aggregate objective, and their assignment is not defined by an allocation understood up to changes on null sets. Giving the manipulator positive mass restores influence by making a whole group act together. That is a new collective-manipulation problem, rather than a mirror of Proposition 1.
The same issue reaches the paper’s other results. Replacing the manipulator with a positive-mass group changes the strategic unit; replacing “every agent” in At-least-1 with an almost-everywhere condition changes the guarantee. A fine-grained graphon could preserve individual network structure, but it would not give the proposed small-support society model or a clear high-multiplicity compression. The block model supplies that compression by imposing structure the paper’s arbitrary networks do not require. It also cannot directly preserve the distance-1 and distance-2 information conditions from the paper.
High-multiplicity work supports the plausibility of cohort regimes; it does not supply the missing computational results here. My negative case is therefore strong for the paper as a source of continuous mirrors: its results are strategic classifications, and preserving the manipulator’s influence requires changing the strategic unit. I can’t honestly rule out the cohort problem as worthwhile research in its own right. I would classify it as a new follow-on question, not a continuous mirror of 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.