| paper | d4-conf-aaai-2017-0852 |
| authors | — |
| venue | ? |
| filed under | unclassified |
| judged by | gpt-6-luna / xhigh (triple__luna__xhigh__c6r1) |
| judge confidence | high |
| authors would recognise it | yes |
The named results are distortion bounds or normative and incentive properties, none of which meets the stated computational-result criterion. A mass-distribution analysis of 2-Agree is statable, but it does not supply a qualifying named result to mirror.
fails bit a — no named computational result to mirror
A distributional extension could revisit the distortion analyses of Theorems 4 and 6; it would leave the lower bounds of Theorems 1 and 5 and the normative and incentive results of Theorems 8 and 9 without a computational mirror.
The paper has no named computational result of the required kind. Theorem 1 gives a distortion lower bound; Theorem 4 bounds 2-Agree’s distortion; Theorems 5 and 6 give distortion bounds for decisive voters; Theorem 8 proves a participation property; and Theorem 9 bounds strategic gains. None classifies an underlying computational problem as tractable, hard, or parameterized.
A plausible continuous analogue would put mass on voter types, sample types until two share a top alternative, and study the resulting distortion. But that would mirror the paper’s mechanism analysis, not a named computational result of the kind required here. I therefore can’t give it as a qualifying anchor or claim a continuous complexity question on this paper’s behalf.
The proponent’s strongest point is that this paper gives no algorithmic complexity classification for a problem whose input is a society and whose output is an optimal intervention. Theorem 1 and Theorems 4–6 bound distortion; Theorems 8–9 concern participation and incentives. The paper does report sample and communication costs, so “no computational result” is a little too broad. But those are resource guarantees for running a randomized mechanism, not complexity results for solving an input problem of the kind ChoCo studies.
The best proposed mirror would give society a distribution \(\mu\) over complete voter types, including preferences and metric costs, then analyze 2-Agree on samples from \(\mu\). That is a sensible model of a large electorate. Yet the mechanism’s behavior is already determined by the top-choice masses \(p_a\): it samples voters with replacement until a top choice repeats. If \(q_a(p)\) is the probability it selects alternative \(a\), and \(S_a=\int d(i,a)\,d\mu(i)\) is that alternative’s per-capita social cost, its distortion is governed by
\[
\frac{\sum_a q_a(p)S_a}{\min_b S_b}.
\]
So the population enters through a finite collection of aggregate quantities. The continuous formulation does not expose a new separation problem, algorithmic bottleneck, or dependence on support size and encoding length. The sample bound also remains a function of \(m\), independent of population size.
The large-electorate scenario is credible, and I would not use high multiplicity as an objection. The problem is fit: the continuous version restates the mechanism’s existing distributional analysis rather than opening a computational question about the population. Studying new sampling policies or learning an unknown distribution might be worthwhile, but that would launch a different research problem rather than mirror a named result here.
That leaves the negative case strong on ChoCo’s computational remit, but weak as a universal claim that no useful continuous-population study could come from the paper. Its natural mass-distribution model is easy to formulate; what is missing is a compelling new complexity question for it.
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.