| paper | d4-journals-jet-2005-0068 |
| authors | — |
| venue | ? |
| filed under | unclassified |
| judged by | gpt-6-luna / xhigh (triple__luna__xhigh__c6r1) |
| judge confidence | high |
| authors would recognise it | no |
Theorem 3 is a partial exception to the proponent’s claim that none of the results is computational: its proof explicitly gives an equilibrium strategy profile. But the proponent states no continuous computational question for it, and a graphon-style limit would change the pairwise incentive model. No anchor survives both requirements.
fails bit b-changes — the continuum limit changes the question
No result has a surviving mirror. Theorem 3 supplies a strategy construction; Theorems 1, 2, and 4 concern equilibrium existence or welfare and absorption. All remain claims about finite-agent network formation.
The strongest honest case is very limited: this paper has no named computational result that qualifies as an anchor under your criteria. Its numbered theorems concern equilibrium existence and welfare properties, not the complexity or algorithms for a computational problem. I can’t turn them into \(P\), hardness, or parameterized-complexity results without changing what they prove.
For completeness, Theorem 1, proved here, says a mixed-strategy equilibrium in bilateral and unilateral actions always exists. Theorem 2, also proved here, says that with an anonymous allocation rule allowing limited transfers, there is a worth function and a patience threshold such that, above that threshold, no efficient graph is strongly absorbing at any pure-strategy equilibrium. Theorem 3 says link monotonicity guarantees some equilibrium in which the complete graph is strongly absorbing, for every discount factor \(\delta\in(0,1)\). Theorem 4 says that under increasing returns to link creation and component-wise egalitarian allocation, some equilibrium makes the complete graph strongly absorbing for sufficiently patient players. None is a complexity or algorithmic claim; the existence proof uses Kakutani’s theorem, and the paper gives no computational input encoding or running-time result.
There is a plausible population setting nearby: many similar communities or firms, grouped by payoff-relevant primitives, with mass representing their population share and link decisions shaping discounted payoffs. But a distribution over those types does not capture the endogenous pairwise network. A faithful continuum model would also need to represent the network state and specify how random pair activation and bilateral link formation scale. That looks like a mean-field extension of the paper’s equilibrium analysis, rather than a computational mirror of one of its named results—and the brief explicitly places such analytic continuization outside the programme’s target.
So I have no continuous problem to state as an anchor-based deliverable. The main weakness is not merely that the paper omits a complexity classification: its central moves are made by individual pairs, and preserving those moves in a continuum population requires additional modeling choices that may change the problem. The paper offers an interesting analytic neighbor, but no defensible positive case for a ChoCo computational mirror under the stated anchor rule.
The proponent’s central point holds: none of Theorems 1–4 states a computational problem with an input, output, and complexity claim. Theorem 1 proves equilibrium existence; Theorems 2–4 establish welfare and absorption properties. Their proofs may involve finite strategy spaces, but that does not make them complexity results.
The tempting retrofit is to ask for an algorithm that finds an equilibrium. The paper gives no input encoding for that task. Its worth functions and allocation rules can range over network states whose number grows exponentially with the number of possible links, and an equilibrium specifies actions at those states. A continuous distribution over agent types does not compress that network state. Restricting payoffs to a succinct family could make equilibrium computation a well-posed research problem, but the restriction and the computational target would have to be added.
The four theorem candidates each have a further obstacle. Theorem 1’s equilibrium existence is a result for each finite game; replacing the population by a continuum would require a new account of how pair activation and a tagged agent’s deviations work. In a graphon limit, a single agent’s links have zero mass, so the original incentive test disappears unless the model adds a tagged-agent or comparable construction. That points toward a mean-field equilibrium analysis, not a computational mirror of the theorem. Theorem 2’s impossibility construction uses three players and an existential choice of worth function; it supplies no population parameter whose scaling could yield a complexity question. Theorems 3 and 4 concern finite-\(n\) strategies that make the complete graph absorbing. Asking whether a complete graphon is absorbing would require specifying new continuum dynamics and incentives, while the paper offers no algorithmic problem for those claims.
This is not an argument that repeated communities or firms are implausible; the paper itself gives identical rural communities as an example. Nor does it rely on existing high-multiplicity work as a collision. Such settings could motivate a worthwhile study of continuum network formation. The negative case is narrower: that study would need to choose a new network limit, preserve individual incentives, and formulate a new computational task. The paper’s named results do not provide the computational anchor ChoCo is meant to continuize.
So the case against a *mirror of this paper* is strong. The universal claim that no scenario could produce worthwhile work is weaker: a researcher could define a computational problem for a restricted family of large network games. The paper alone gives little reason to count that as its continuous mirror.
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.