| paper | d4-conf-aaai-2015-0397 |
| authors | — |
| venue | ? |
| filed under | unclassified |
| judged by | gpt-6-luna / xhigh (triple__luna__xhigh__c6r1) |
| judge confidence | high |
| authors would recognise it | no |
Proposition 1 states a cited equivalence between extensions and labellings; it does not assert that an object can be computed, give a computational guarantee, or establish hardness. The algorithms and speedup findings are the paper's computational contributions, but they are not stated as a qualifying numbered result, so bit (a) fails.
fails bit a — no named computational result to mirror
The attempted type-role mirror targets preferred-extension enumeration; it does not model the paper's parallel decomposition or performance study as population questions.
The strongest honest positive case is narrow. Proposition 1, cited to Caminada (2006), says that a labelling is complete, grounded, or preferred exactly when it is obtained by \(\mathrm{Ext2Lab}\) from an extension of the corresponding semantics; the proposition itself is not proved in this paper. The paper computes preferred labellings, so this is a representation result it uses to compute the underlying problem: enumerating preferred extensions. The paper has no numbered complexity result for that problem. Its introduction says, without a numbered result, that preferred-extension enumeration lies at the second level of the polynomial hierarchy. That informs the expected direction, but it does not meet the requested named complexity-anchor criterion. If only such complexity results count as anchors, this paper supplies none.
A concrete mirror would take argument occurrences as the mass-bearing units. Let \(T\) be a finite set of argument types, let each type \(t\) have rational mass \(\mu_t>0\) with \(\sum_{t\in T}\mu_t=1\), and let \(R_T\subseteq T\times T\) be the attack relation. A type is a complete attack role: every occurrence of type \(t\) attacks every occurrence of type \(u\) exactly when \((t,u)\in R_T\). The continuous society is a set of argument occurrences of total mass one, partitioned into type classes of masses \(\mu_t\).
The problem is: enumerate all preferred extensions of this continuous argumentation framework. A solution can be given as a labelling \(\ell:T\to\{\mathrm{in},\mathrm{out},\mathrm{undec}\}\). It is complete when, for each \(t\), \(\ell(t)=\mathrm{in}\) exactly when all attacking types are out, and \(\ell(t)=\mathrm{out}\) exactly when some attacking type is in; otherwise \(t\) is undec. It is preferred when its in-set is inclusion-maximal among complete labellings. The corresponding extension has mass \(\mu_t\) for each in-type and zero for the other types. There is no optimization objective: the task is to output the whole family, as in the paper.
A plausible regime is a large structured consultation where many argument occurrences instantiate a small number of recurring attack roles: the number of occurrences \(n\) is large compared with the number of types \(\tau=|T|\). For unrestricted \(\tau\), I would expect Class B: a finite AF embeds by giving each argument its own type and preserving its attacks, so the continuous formulation need not remove the combinatorics. The useful further questions are whether enumeration admits output-sensitive or polynomial-delay algorithms in \(\tau\), and how complexity changes in the genuinely small-support regime.
The weakness is substantial: in Dung’s semantics, identical attack roles receive the same status, so mass does not create fractional extension choices or otherwise affect the answer beyond identifying which types are present. Also, repeated argument occurrences are a plausible “population” only in structured debate settings; many AFs model individual arguments whose attack profiles are all distinct. This mirror covers the paper’s preferred-extension enumeration problem, not its parallel speedup result as a separate social-choice objective. It is a defensible, limited mirror, but the paper offers no named hardness anchor to make the positive case strong.
The proposed mirror has one basic defect: mass does not enter preferred semantics. Proposition 1 is a representation equivalence cited from prior work, not this paper’s computational result. The paper’s contribution is an enumeration algorithm and an empirical parallel-speedup study; it states the enumeration problem’s complexity level in the introduction, but proves no numbered complexity result to mirror.
In the proposed model, the semantics depend on the attack relation among types, not on their masses. As long as \(\mu_t>0\), changing \(\mu_t\) cannot change a type’s status. Reporting that the accepted types have total mass \(\sum_{t\text{ in}}\mu_t\) adds a statistic, but does not make the enumeration question depend on a continuous society.
The stronger case for a mirror is a large consultation with many argument occurrences in recurring attack roles. But if occurrences of each role really are indistinguishable in every way the semantics uses, the role graph already captures the problem: preferred extensions select whole roles. In a corresponding blow-up with identical attack neighborhoods, preferred maximality rules out selecting only some mutually nonattacking copies of a role. The mass-bearing formulation therefore adds no population-level choice to the computation. If occurrences have different attacks, they need different types, and the supposed high-multiplicity regime disappears.
One might instead ask how much mass must be removed from a role to change a preferred-status query. Under these semantics, removing part of a role leaves its status unchanged while any positive mass remains; changing the support requires removing the entire role. A genuinely gradual response would require mass-sensitive attacks or defenses. That could be a worthwhile new argumentation model, but it changes the semantics the paper computes rather than continuizing its problem.
So the proposed mirror is not a computational question about a continuous population in the programme’s sense: its continuous coordinates are semantically inert, and making them matter requires replacing the paper’s semantics. The negative case is strongest for the paper’s actual contribution—parallel enumeration—and weaker as a universal claim about argumentation: a particular application could motivate a new mass-sensitive semantics, but that would be a different target.
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.