| paper | d4-conf-ijcai-2017-0741 |
| authors | — |
| venue | ? |
| filed under | unclassified |
| judged by | gpt-6-luna / xhigh (triple__luna__xhigh__c6r1) |
| judge confidence | medium |
| authors would recognise it | no |
Theorem 2 supplies a computational anchor, and CP-PSAT is a precise, coherent reinterpretation. But every society has a support-one representative with the same majority valuation, so its masses do no work; the case gives no paper-grounded reason the authors would recognize this as their population problem.
fails bit b-none — no population with multiplicity to relax
The proposed majority map makes population mass irrelevant, and the case supplies no link from this new interpretation to the paper’s game-form problem.
fatal: True
The candidate covers the CP satisfiability hardness consequence associated with Theorem 2; it does not mirror the other embedding results as separate population problems.
The strongest mirror I see is a population semantics for CP satisfiability. It is a plausible extension of the paper’s logic, though a weak fit to the programme’s central optimization examples.
My anchor is Theorem 2, proved in this paper. For an individual STIT formula \(\varphi\) of length \(m\), it says three things are equivalent: \(\varphi\) is STIT-satisfiable; it is satisfiable in a STIT model where each agent has at most \(2^m\) actions; and \(\mathrm{GRID}_m \land \mathrm{tr}_1(\varphi)\) is CP-satisfiable. The theorem itself is an embedding result, not a standalone complexity classification. The paper uses it, together with the cited NEXPTIME-completeness of individual STIT satisfiability, to obtain NEXPTIME-hardness for CP satisfiability. That is the hardness consequence I would mirror.
Here is a concrete question: Ceteris-Paribus Population Satisfiability (CP-PSAT). An input is a CP formula \(\psi\) over a finite proposition set \(P\). A type is a complete Boolean profile \(t\in\{0,1\}^{P}\). A society is a distribution \(\mu\) over these types; \(\mu_t\) is the fraction of the population with profile \(t\). A population model has a finite, nonempty set \(W\) of such societies. At society \(\mu\), proposition \(p\) holds when more than half the population has \(p=1\). Define \(\mu\sim_X\nu\) when the two societies agree on that majority truth value for every \(p\in X\). Then \(\langle X\rangle\psi\) holds at \(\mu\) if some \(\nu\in W\) satisfies \(\mu\sim_X\nu\) and satisfies \(\psi\). The question is whether there is such a model and a society \(\mu\in W\) where \(\psi\) holds. A solution is a finite set of societies, with their type masses, and a designated satisfying society. Rational masses suffice.
The scenario is a large deliberative population in which people share a relatively small number of stance profiles. A society’s state is the distribution across those profiles; propositions describe aggregate facts, such as whether a majority supports a claim. CP’s “all else equal” condition then has a direct population reading: look for another possible society with the same aggregate truth values on \(X\). This mirrors CP satisfiability itself; it does not claim that the paper’s named agents already form a continuum.
I expect the hardness to transfer. Any ordinary CP state with valuation \(v\) can be represented by the homogeneous society \(\delta_v\), with all mass on type \(v\). Conversely, a population state can be mapped to its majority valuation. Since the modal relation only observes those truth values, the population model and its induced CP model agree on formula satisfaction. So this candidate problem is equivalent to CP satisfiability and inherits its NEXPTIME-hardness; with the paper’s stated upper bound, it has the same NEXPTIME-complete complexity. The hard structure lies in the propositions and agency logic, not in population multiplicity.
The mirror covers the CP-satisfiability result and the individual-STIT hardness transfer through Theorem 2. It does not mirror the paper’s other embedding theorems as separate problems. It also generates a more substantive follow-up: restrict feasible societies using rational type quotas or other linear constraints, then ask whether a CP formula is satisfiable within that population polytope. That would make mass composition affect feasibility and invite the continuous-optimization questions ChoCo is designed to study, but it is an extension rather than the anchor.
The weak point is substantial: CP-PSAT’s semantics collapses every society to its majority truth vector. Every such vector has a homogeneous, support-one representative, so fractional masses add no computational leverage or difficulty. Theorem 2 is about finite-agent logic, and the paper does not study population distributions. I think CP’s authors could recognize this as a coherent population interpretation of their logic, but they might not regard it as a meaningful new computational problem until quotas or other mass-sensitive constraints are added.
Theorem 2 is a real computational anchor: through the cited complexity of individual STIT satisfiability, it gives a lower bound for CP satisfiability. The proposed population interpretation is also formally coherent. Its weakness is that the population contributes no information the logic can use.
In CP, a state is a world with a valuation of propositions. The modalities compare worlds according to which propositions agree. In the proposed mirror, each society is first reduced to its majority truth values. Those values are exactly a CP valuation: every valuation has a representative society \(\delta_v\), and replacing any society by that representative preserves every formula’s truth. Thus the continuous masses can be removed without changing the problem. The society is not an input population whose composition an algorithm must reason about; the existentially chosen set of societies simply re-encodes CP’s set of worlds.
A more population-sensitive version could constrain feasible compositions—for example with quotas—or ask for the least mass transfer that makes a formula true. But those constraints or intervention costs would supply the substantive computational problem. They are not consequences of Theorem 2 or of CP’s semantics, which gives no population, cost, or way to change one. They might support a new project, but they do not rescue this mirror as a continuization of the paper’s result.
So the negative case is not that CP satisfiability is uncomputational or that its complexity would survive the move. It is that the proposed move leaves CP satisfiability intact while giving “population” no role beyond a redundant encoding. This is a strong objection to this anchor, though not an airtight universal one: a convincing, mass-sensitive population model for agency could make a worthwhile extension, but the paper gives little basis for identifying one as its problem.
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.