| paper | d4-journals-geb-2021-0003 |
| authors | — |
| venue | ? |
| filed under | unclassified |
| judged by | gpt-6-luna / xhigh (triple__luna__xhigh__c6r1) |
| judge confidence | high |
| authors would recognise it | no |
The propositions establish equilibrium existence and welfare comparisons, while the numbered experimental results report observed outcomes; none asserts an algorithm, complexity classification, or computational formulation used to compute. Thus bit (a) fails, regardless of whether a continuous analogue of the strategic setting can be described. The proposed nonatomic election is a plausible population model, but it removes individual pivotality and changes the strategic claim.
fails bit a — no named computational result to mirror
No named result yields a continuous computational mirror: Propositions 1–3 concern finite-election equilibria, and Results 1–4 report laboratory findings.
The strongest positive case is a plausible population model. Take a large electorate facing states \(a\) and \(b\), with a common-value majority of share \(\alpha\), where \(1/2<\alpha<2/3\), and a partisan bloc of share \(1-\alpha\) that supports \(C\). The independent voters receive signals \(s_a\) or \(s_b\). Conditional on state \(a\), their type masses are \(\alpha q_a\) and \(\alpha(1-q_a)\); conditional on state \(b\), they are \(\alpha q_b\) and \(\alpha(1-q_b)\). The partisan type has mass \(1-\alpha\). A strategy assigns each independent signal type a distribution over first-round votes; the runoff rule then determines the finalist and second-round vote. The objective is the independent voters’ expected utility: \(H\) for the correct majority candidate, \(L\) for the incorrect one, and \(0\) for \(C\).
A large referendum electorate with common interests, a substantial partisan bloc, and a shared signal technology is a credible high-multiplicity scenario: millions of voters could fall into just these few payoff-and-information types. That makes the population description natural. The article’s numbered theoretical results also give it a clear analytic point of contact. Proposition 3, proved in this paper, says that for every admissible information structure and electorate size, runoff has an informative equilibrium giving the independent voters strictly higher expected utility than *any* plurality equilibrium. Proposition 1, also proved here, gives generic information environments in which both rules admit undominated Duvergerian equilibria. Proposition 2 says both rules always admit informative equilibria in undominated strategies; its plurality part follows from Bouton et al. (2016), while the runoff part is established here using Proposition 3.
But none of these is a qualifying computational anchor. They assert equilibrium existence and welfare comparisons, not complexity or algorithms for a computational problem. The paper’s experimental Results 1–4 report observed performance and voting behavior; they do not supply such an anchor either. So I would not present Propositions 1–3 as results about problems being in P, NP-hard, FPT, or similar. The paper contains no named computational result of the kind this task requires. Consequently, there is no valid continuous computational question to state per anchor.
The weakest point in the population analogy is also central to whether it can become a ChoCo mirror. The experiment itself uses groups of five voters, and the theoretical results concern finite electorates. In a literal nonatomic model, an individual voter has no effect on the aggregate vote, so the pivotal incentives that drive the paper’s strategic analysis can disappear. Treating each type’s mass as a coordinated voting bloc would restore consequential choices only by changing the game. Thus the type distribution is a convincing way to continue the paper’s *setting*, but a continuum equilibrium analysis risks changing the authors’ strategic question—and, even if accepted, it would be analytic rather than computational.
A large divided-majority electorate is a plausible population regime. The problem is that this paper supplies no computational result to mirror: Propositions 1–3 concern equilibrium existence and welfare comparisons, while Results 1–4 report experimental behavior. None asks for an algorithm or establishes a complexity classification. Turning one into an algorithmic question would add a new research target rather than continuize a computational result in the paper.
The strongest anchors also lose their strategic content in the natural nonatomic limit. Conditional on the state, a continuum electorate has fixed masses of each signal type. A single voter has measure zero, so changing their ballot cannot change the aggregate vote or the winner. Every strategy profile is then a Nash equilibrium: an individual deviation leaves their payoff unchanged. The distinction between informative and Duvergerian play, and the pivotal incentives behind the paper’s coordination problem, no longer constrain individual behavior.
That defeats the three propositions in different ways. Proposition 1’s undominated Duvergerian equilibria become uninformative as a prediction: with no unilateral effect on the outcome, the equilibrium condition does not select Duvergerian behavior. Proposition 2’s informative-equilibrium existence likewise becomes vacuous; informative and uninformative profiles both qualify. Proposition 3’s welfare comparison becomes an optimization over aggregate voting behavior, unconstrained by the strategic incentives that the proposition compares. One could study that aggregate optimization, but it would not be a computational version of the paper’s equilibrium claim.
Making the model more elaborate does not plainly repair this. A positive-mass bloc that votes as one strategic player changes the agents into coalitions. A single designated voter has vanishing mass as the electorate grows. Keeping each voter pivotal requires retaining finite vote granularity; that yields a high-multiplicity finite game, not the nonatomic strategic model. Computing equilibria in such games could be a worthwhile separate project, but the paper does not pose that algorithmic problem, and its finite equilibrium results are not themselves complexity results.
The experimental findings offer no stronger anchor. They come from fixed groups of three human voters interacting over repeated periods, where coordination and learning matter. Aggregating their votes into masses would discard those group interactions; replicating the lab many times would increase the number of groups, not the population within each election.
So the negative case does not depend on denying that the population setting is sensible. It is that the paper’s substantive object is strategic equilibrium in finite elections, while its continuous limit removes individual pivotality. The remaining plausible computational questions require introducing a new finite-game algorithmic agenda. I cannot rule out that agenda being worthwhile, but it would not be a continuous mirror of any result this paper establishes.
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.