| paper | d4-conf-ijcai-1991-0190 |
| authors | — |
| venue | ? |
| filed under | unclassified |
| judged by | gpt-6-luna / xhigh (triple__luna__xhigh__c6r1) |
| judge confidence | high |
| authors would recognise it | no |
The paper gives rules for sequential and parallel propagation and says that combinations preserve certain properties, but it contains no numbered result asserting an algorithm, complexity bound, or other computational result. Without an eligible result to anchor, no continuous mirror survives; the full paper confirms this focus. [Paper](https://www.ijcai.org/Proceedings/91-2/Papers/086.pdf)
fails bit a — no named computational result to mirror
There is no computational result to cover; the paper presents propagation rules and claims that combinations preserve properties of the input relations.
I can’t make a result-anchored positive case for this paper. Its contribution is to describe sequential and parallel propagation of qualitative preference relations across finite frames, using compatibility relations. The conclusion says that certain properties of the inputs are preserved by combination, but the paper states no numbered Theorem, Lemma, Corollary, or Proposition asserting a computational result—no tractability, hardness, or parameterized-complexity result to anchor a mirror. The belief-function representation result it mentions is attributed to Wong et al. (1990), not proved here. [The IJCAI paper](https://www.ijcai.org/Proceedings/91-2/Papers/086.pdf)
A plausible *new* direction would model many evidence sources drawn from a small set of types, with each type specifying a preference relation and compatibility relation, and mass recording its share of the sources. But the paper defines no mass-weighted parallel-combination rule. Adding one would supply new semantics rather than continuize a problem the paper analyzes, and may conflict with its motivation for qualitative input. I therefore have no eligible anchor and no honest prover-ready continuous problem to offer for this paper.
The proponent raises no result-anchored continuous question, so there is no anchor to defeat by silence. That is the strongest case against a mirror: this paper states no computational problem and proves no complexity or algorithmic result. Its contribution is rules for propagating and combining qualitative preference relations, plus claims about which properties those combinations preserve. The belief-function representation result is attributed to earlier work. There is no existing computational result here for a continuous population to mirror.
The best population scenario is a large community of evidence sources, many sharing the same complete preference relation and compatibility data. That could make a high-multiplicity regime sensible. One might ask how much source mass must change to make a target frame prefer one proposition over another. But that adds an intervention, costs, and a target condition that this paper never studies. It is a new control problem built on the paper’s combination rules, not a continuous version of one of its results. Alternatively, one could define a mass-weighted parallel-combination rule, but that adds new semantics before there is a computational question to analyze.
There is a further change in what the model represents. The paper combines evidence inputs to describe one body of uncertain knowledge; it does not aggregate a population’s choices into a social outcome. Repeated source types may supply multiplicity, but they do not by themselves supply the social-choice problem that the programme continuizes.
I would not claim that no useful continuous problem could ever be built around evidence sources. A population-control question might be worthwhile on its own. The honest negative case is narrower: this paper offers no computational result to anchor such a mirror, and the strongest proposed replacement is a new problem rather than a continuization of the paper’s contribution.
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.