| paper | d4-conf-aaai-2006-0116 |
| authors | — |
| venue | ? |
| filed under | unclassified |
| judged by | gpt-6-luna / xhigh (triple__luna__xhigh__c6r1) |
| judge confidence | high |
| authors would recognise it | no |
Theorem 2 is the closest computational candidate because its proof explicitly constructs a ranking rule, although it gives no runtime bound. But the proposed mass-change question replaces a unilateral link deviation with coordinated manipulation by positive mass, and link-set proportions alone lose the identities of the vertices being linked to. A role-based network model would be a re-modeling, not a surviving mirror of a result in this paper.
fails bit b-changes — the continuum limit changes the question
The proposed type masses do not preserve which named vertices links target, while positive-mass changes replace individual deviations with coalition manipulation.
fatal: True
No result has a surviving mirror; the closest proposed extension concerns Theorem 2's constructed rule and incentive guarantees, leaving the other bounds and the three-agent axiomatization without a mirror.
I can’t make a case that meets the requested anchor rule. The paper has no numbered result asserting a computational complexity classification—for example, that a problem is in P, NP-hard, FPT, or W[1]-hard.
Theorem 1 and Proposition 1 give bounds on incentive compatibility under strong monotonicity and for Approval Voting. Proposition 2 gives lower bounds for PageRank. Theorem 2 constructs a non-imposing system with bounded manipulation, and Theorem 3 axiomatizes three-agent systems. These are substantial results, but they characterize ranking systems and their incentive properties; they do not classify the complexity of computing something.
One could define a continuous-population version of the incentive measure—for example, let types be outgoing-link sets, let \(\mu_t\) be the population fraction of each type, and ask for the maximum change in a type’s expected rank when some of its mass changes reports. But that would be a new optimization question, not a mirror anchored in a qualifying result of this paper. It also risks replacing the paper’s axiomatic questions with a different computational problem.
So the strongest honest positive case is limited: the paper’s measures could inspire a continuous-population research question, but this paper supplies no named computational result to anchor it, and its stated contribution is principally axiomatic.
The proponent’s concession is decisive under ChoCo’s gate: the paper has no named computational result to mirror. Theorem 1 and Propositions 1–2 bound incentive measures; Theorem 2 constructs a ranking system with specified incentive properties; Theorem 3 gives an axiomatization for three agents. These are substantive results, but none gives a complexity classification or an algorithm for a computational problem. Even if computing \(\delta_F^G(v)\)\) is a natural optimization question, the paper bounds that quantity rather than studying its computational complexity. A continuous optimization inspired by those bounds would be a new question, not a mirror of a result in this paper.
The proposed type space of outgoing-link sets does not repair that gap. In the paper, an outgoing set names particular vertices, and those vertices are also the alternatives being ranked. A distribution over link sets alone does not say which named vertex has which set; that assignment can change the graph and its ranking. Preserving the identities avoids that loss, but then the model has not really replaced the finite society with an anonymous population of repeated types.
A stronger formulation could use a finite set of network roles, assign mass to each role, and specify links between roles with a kernel. That is coherent, but it changes the incentive question. In an atomless population, one agent’s link change has zero effect on the aggregate network and typically on population ranks. Allowing a positive mass of a type to change links can avoid that degeneration, but it studies coordinated, mass-based manipulation rather than the paper’s unilateral deviations. That could be a worthwhile project; the paper supplies no computational result anchoring it.
So the negative case is strong for the screening decision: this paper offers no qualifying ChoCo mirror. The broader claim that no worthwhile continuous research could ever grow from it is weaker. A role-based network model might be interesting on its own, but presenting it as a continuous mirror of these results would overstate what the paper establishes.
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.