| paper | d4-conf-aaai-2000-0223 |
| authors | — |
| venue | ? |
| filed under | unclassified |
| judged by | gpt-6-luna / xhigh (triple__luna__xhigh__c6r1) |
| judge confidence | high |
| authors would recognise it | no |
The named theorems establish logical consequences of mutual beliefs, commitments, and declarations; none asserts that a computational problem can be solved or is hard. The type-mass optimization is a new problem proposed by the analyst, not a continuous analogue anchored in a computational result of this paper.
fails bit a — no named computational result to mirror
Theorems 1–3 are logical validity results, and the paper gives no named computational result to mirror. The proposed type-mass vote optimization adds an objective and model semantics not established by those results.
The paper has named results, but no named computational result under your anchor rule. Theorem 1 says that mutual belief in the trigger and the specified conditional commitments entails the method’s FSP requirements, and that adding the method’s validity requirements entails the full FSP. Theorem 2 says that an announcement of group acceptance, in the context of an invoked voting mechanism, establishes mutual belief that the group accepted. Theorem 3 says that, for the AdoptMSP(A) decision type, an invoked mechanism and an accepted announcement entail \(MSP(GR,A)\). These are logical validity results; none classifies a problem as tractable or hard. So there is no eligible anchor, and no honest way to give the requested per-anchor continuous problems or expected complexity classes.
There is a plausible, but unanchored, extension. Imagine a large fleet of replicated agents using a few operational templates. A type would have to include each agent’s relevant beliefs, intentions, capabilities, and permitted declarations; \(\mu_t\) would be the fraction of agents of type \(t\). One could define Continuous SharedPlan Adoption with Type-Mass Votes: given rational masses over those types, a proposal to adopt \(MSP(A)\), a specified acceptance quota, per-unit costs for eliciting declarations, and explicit type-level interpretations of the required FSP clauses, find the least-cost allocation of type mass to declarations and commitments that reaches the quota and satisfies those clauses for almost every agent. The allocation would be the solution. Theorem 3 motivates the goal—realizing the premises that yield MSP—but does not establish this optimization problem or its complexity. The quota, costs, and continuum semantics would all need to be added.
My strongest positive case is that the paper’s general voting mechanism is a natural starting point for studying decisions in large populations of replicated agents. Its weakest point is that SharedPlans concern mutual beliefs and commitments among a specified group, often a small team; the paper gives no account of how those group-level notions behave for a distribution of types. That makes the extension conceivable, but not an anchored mirror of a computational result in this paper.
The strongest objection is decisive for this paper: it has no named computational result to mirror. Theorem 1 derives the FSP requirements from mutual belief in a trigger and conditional commitments. Theorem 2 derives mutual belief in group acceptance from the announcement rules. Theorem 3 applies those logical rules to adoption of a minimal SharedPlan. None studies an algorithmic problem, its complexity, or an optimization objective. The proponent is right that these are logical validity results, but that means they supply no eligible anchor.
The proposed “Continuous SharedPlan Adoption with Type-Mass Votes” adds the computational problem it then proposes to study. The paper gives no elicitation costs, acceptance quota, or objective of minimizing declarations; nor does it define acceptance as a fraction of agents. Adding those features could produce a research question, but it would not be the continuous version of Theorem 3. A stronger attempt might give each agent a complete type containing their beliefs, intentions, capabilities, and permitted speech acts, then optimize over masses of types. But the paper’s conclusions concern mutual belief within a specified group: a nested relation among its members, not an average of their individual states. Masses of types and votes alone do not say whether that group mutually believes a proposition. Keeping the epistemic structure would require modeling the relations among agents as well, leaving the paper’s mechanism with no evident aggregate computational problem.
This also defeats Theorems 1 and 2 as alternate anchors: their premises and conclusions are about those same group-level beliefs, conditional commitments, and declarations. Replacing them with quotas or type-level constraints would change what they establish. It could be a sensible new model for some replicated-agent setting, but any resulting complexity question would come from the added model, not from a computational result in this paper.
The negative case is strongest on the paper’s eligibility for the programme, and weaker as a universal claim about every possible extension. Large populations of replicated agents might motivate a new computational study of decision processes. The paper, however, provides no computational result that makes that study its continuous mirror; the proposed extension would need to supply the central question as well as its continuous semantics.
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.