| paper | d4-conf-atal-2022-0317 |
| authors | — |
| venue | ? |
| filed under | unclassified |
| judged by | gpt-6-luna / xhigh (triple__luna__xhigh__c6r1) |
| judge confidence | high |
| authors would recognise it | unclear |
The only anchor is the query-network characterization, which asserts no computational result. The cohort proposal would need a new aggregate invitation model and replaces individual path rewards with type-level masses and reward densities. Bit (a) therefore fails.
fails bit a — no named computational result to mirror
No computational result is covered. The query-network characterization and the paper's other formal results establish mechanism properties such as efficiency, diffusion incentive compatibility, and structural fairness.
The strongest plausible population scenario is a large crowdsourced query campaign in which many participants share a role: the same invitation options, depth or permission rule, and contribution to the answer. Represent those roles by a finite set \(T\), with population shares \(\mu_t\). A continuous query-network question could ask for reward densities \(r_t\) that distribute the value of the reached population efficiently, remain positive for roles on the requester-to-answer path, and make each participant prefer to report all available invitations. The action would be an invitation report; a solution would be a reward rule satisfying those conditions.
The closest result in the paper is the Section 4 theorem: it characterizes anonymous, strongly individually rational, efficient query-network rewards as weighted permission Shapley values, with weights depending on the participant’s and answer provider’s depths. That is a useful conceptual starting point for the proposed cohort model. But it is an axiomatic characterization, not a result about tractability, hardness, or algorithms. The other formal results likewise establish efficiency, diffusion incentive compatibility, or structural fairness. None supplies the kind of computational anchor this task requires.
So the paper contains no named computational result of the required kind, and there is no honest per-anchor continuous computational problem to state. The proposed cohort version would also need a new model of how invitations and reachability work between masses of participants; the paper’s individual DAG does not settle that. It may be a reasonable future mechanism-design question, but under ChoCo’s computational scope it is not a strong positive mirror. The supplied text also omits theorem and proposition numbers, so I cannot quote them as printed.
The proponent’s only anchor is the Section 4 characterization of query-network rewards. It is axiomatic: it classifies which allocations meet anonymity, strong individual rationality, and efficiency. It defines no computational problem, input encoding, algorithm, or complexity measure. More specifically, its “only if” direction sets each player’s weight to their already-chosen reward \(x(i)\); the weighted permission value then reproduces that allocation by normalization. That makes it a weak starting point for a computational result, and a mass version would preserve this simple representation.
The best salvage is a crowdsourced campaign with many participants in repeated roles, asking for efficient, diffusion-incentive-compatible reward densities across role masses. But the paper’s query-network theorem concerns a path to a particular answer provider and positive reward for every individual on that path. In an atomless population, a particular provider has zero mass, and positive rewards to every individual cannot be the condition as written under a finite total budget. Replacing individuals with densities and a positive mass of answer providers is sensible, but it changes the object being characterized. For simple cohorts by depth, the resulting allocation can be stated directly as a normalized density; the theorem still supplies no computational challenge.
A richer version could retain the individual invitation graph and ask to compute weighted permission Shapley rewards. That runs into the opposite problem: to be indistinguishable for this model, agents must match in their permissions, position, and role in the value function. In generic networks—especially when one named person holds the answer—those details distinguish agents, so the support can grow with the population. A compressed type-level graph is possible, but then invitations and deviations become aggregate or coordinated actions. That may define a new mechanism-design problem, but it is not a computational mirror of a result in this paper. The paper also leaves the characteristic function \(v\) arbitrary and gives no computational representation for it.
So the negative case is strong for this paper: its formal results establish mechanism properties, not computational claims, and the closest characterization either loses its individual conditions in a continuum or needs a substantial new model. The universal claim is harder to make airtight: a carefully specified cohort model could still motivate useful algorithmic work. But that would be a new question inspired by the paper, rather than a continuous computational mirror of one of its results.
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.