| paper | d4-conf-ijcai-1993-0059 |
| authors | — |
| venue | ? |
| filed under | unclassified |
| judged by | gpt-6-luna / xhigh (triple__luna__xhigh__c6r1) |
| judge confidence | high |
| authors would recognise it | yes |
Theorem 2 establishes truthfulness, but does not assert that a computational task can be solved or is hard. The proposed mass-weighted planning problem echoes the paper’s welfare objective, yet remains an unanchored follow-up and leaves out its dynamic aggregation procedure. The paper therefore has no qualifying computational result to mirror.
fails bit a — no named computational result to mirror
No computational result is covered; the proposed mass-weighted planning objective echoes the paper’s welfare goal but leaves out the dynamic constraint aggregation procedure and Theorem 2’s truthfulness result.
I don’t see a qualifying computational-complexity anchor in the supplied text. The closest numbered result is Theorem 2, proved in the paper’s Section 5: at every step \(k\), agent \(i\)’s best strategy is to vote according to their true preferences over that step’s alternatives. That is an incentive result for the iterative Clarke tax mechanism, not a result classifying a computational problem as tractable or hard. The paper’s welfare-maximizing planning objective is described without a numbered complexity result; it also says complexity is not its primary concern. So I can’t honestly give the requested anchor-backed case or assign its anchors to the A/B/C trichotomy.
There is a plausible *unanchored* follow-up question. In a large fleet or service system, agents might share a finite set of complete types—goal, utility, capabilities, and costs—so the number of agents \(n\) is much larger than the number of supported types \(\tau\). Let \(S_0\) and the planning model describe the shared world, let \(\mu_t\) be each type’s rational population fraction, and let \(u_t(s)\) be that type’s utility for terminal state \(s\). Continuous Welfare-Optimal Planning asks for an executable plan \(\pi\) and resulting state \(s\) maximizing \(\sum_{t\in T}\mu_t u_t(s)-C(\pi)\). Its solution is the maximizing plan and state. This keeps the population as the continuous object and uses a social-welfare form the paper permits.
I would expect hardness to persist if the planning model retains the discrete combinatorics of ordinary planning, but the paper does not establish the embedding needed to make that a result about its own problem. The weak point is that this direct optimization question abstracts away the paper’s dynamic constraint aggregation and its truthfulness guarantee. It is a credible research direction, but it is not a mirror anchored by a named complexity result in this paper.
The strongest negative case is that this paper offers no named computational result to mirror. Theorem 2 proves a truthfulness property of the iterative Clarke tax mechanism; it does not classify the complexity of a computational problem. The paper discusses planning and describes a search procedure, but explicitly says complexity is not its primary concern. So there is no result here to place in ChoCo’s A/B/C landscape.
The best repair is the proposed welfare-optimal planning problem: give a large fleet or service system a finite set of complete agent types, with rational population shares \(\mu_t\), and seek an executable plan \(\pi\) maximizing \(\sum_t \mu_t u_t(s(\pi))-C(\pi)\). That is a plausible high-multiplicity scenario. But it is a new problem specification, not a mirror of a result in this paper. The paper does not fix an input language for plans, a plan encoding, or a cost and evaluation model precise enough to define a canonical complexity question. After those choices are made, the problem may be worth studying, but the computational landscape will depend heavily on them; the paper supplies neither a result nor a settled model to anchor that work.
Trying to mirror Theorem 2 directly fares no better for ChoCo’s computational scope. In an atomless population, an individual has zero mass, so an individual deviation cannot affect the outcome and the Clarke tax’s pivotality disappears. Treating a whole type cohort as a strategic voter could preserve influence, but that changes the mechanism’s strategic unit, and the resulting question would still be about truthfulness rather than computational complexity.
I would not claim that no plausible continuous planning question could ever be worthwhile. A finite-type fleet acting in a shared world is credible, and a carefully specified mass-weighted planning problem could become a useful project in its own right. The defensible negative conclusion is narrower: this paper does not establish a computational result that makes such a project its continuous mirror. The case against a mirror is strong on that ground; the broader claim that no scenario could justify one is not.
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.