| paper | d4-journals-jet-2019-0032 |
| authors | — |
| venue | ? |
| filed under | unclassified |
| judged by | gpt-6-luna / xhigh (triple__luna__xhigh__c6r1) |
| judge confidence | medium |
| authors would recognise it | yes |
other→P discrete result → its continuous mirror
Answer: Polynomial-time solvable: every valid instance admits an implementing regime. · discrete side: other · adjudication: proved
full proof & adversarial review →
Theorem 2
statement extracted from the paper’s text layer
Given finite alternatives \(A\) and states \(\Theta\), rational masses \(\mu_t\) over constituency types \(t\), per-capita utilities \(u_t(a,\theta)\), state probabilities \(p\), discount factor \(\delta\), and a fixed tie-breaking rule, let \(f_\mu(\theta)\) maximize \(\sum_t \mu_t u_t(a,\theta)\). With one coordinating organization per type observing each state, decide whether there is a regime of repeated, possibly stochastic mechanisms with a nonempty Nash equilibrium set such that every equilibrium selects \(f_\mu(\theta)\) in every period and state.
A society is a distribution \(\mu\) over constituency types, each represented by one strategic organization whose per-capita utility is \(u_t\). The masses determine the desired alternative \(f_\mu(\theta)\); organizations play the repeated implementation game, and the designer’s regime must implement that target in every Nash equilibrium.
With one representative per constituency type, positive mass scaling does not change that representative’s incentives, so mass affects the target \(f_\mu\) but not the repeated implementation game’s incentive structure.
fatal: False
The mirror covers Theorem 2’s repeated Nash implementation characterization and Theorem 1’s dynamic monotonicity test for mass-weighted targets. It leaves aside the two-agent extensions, alternative equilibrium concepts, deterministic-stage-mechanism results, and Proposition 2’s outcome-space continuity.
There is a defensible positive case if the society is made up of large constituencies whose members are represented by strategic organizations. The key modeling step is that an organization, rather than each member individually, sends the messages and chooses strategies. That fits settings such as repeated allocation among unions, provider associations, or regional stakeholder groups. A type is the complete state-dependent preference and information profile shared by a constituency; its mass \(\mu_t\) is its fraction of the represented population. A plausible regime has \(N\) members but only \(\tau\) types and organizations, with \(N\gg\tau\) and \(\tau\ge 3\). Keep the alternatives finite and time discrete: the population is the continuous object.
For a concrete target, let \(f_\mu(\theta)\) select an alternative maximizing \(\sum_t\mu_t u_t(a,\theta)\), with a fixed tie-breaking rule. All organizations observe the period’s state, which is drawn i.i.d., but the designer does not. The organizations receive their members’ per-capita utility \(u_t\); the masses weight the social objective. The designer chooses repeated, possibly stochastic mechanisms based on public histories. This is a co-modification: the organizations act as type-level players. It is plausible when constituencies can coordinate through representatives, but it is a substantive condition on the scenario.
The lead anchor is Theorem 2, proved in this paper. In the paper’s model, with more than two agents and the stated assumptions—including strict preferences, pure-strategy Nash implementation, and stochastic stage mechanisms—a social choice function \(f\) is repeatedly implementable if and only if it is dynamically monotonic with respect to some collection \(C=(C_i(\theta))_{i,\theta}\). Dynamic monotonicity rules out any deception that changes the prescribed outcome while leaving every agent without an incentive to whistle-blow, including deceptions that continue over many periods.
Its mirror is Mass-Weighted Repeated Implementation. Given finite \(A\) and \(\Theta\), rational type masses \(\mu\), type utilities \(u_t\), state distribution \(p\), discount factor \(\delta\), and target \(f_\mu\), decide whether there is a regime of repeated mechanisms for the type-level organizations such that an equilibrium exists and every equilibrium implements \(f_\mu(\theta)\) in every period. A solution is a regime description with that guarantee; a negative answer rules out every such regime. This asks whether the paper’s implementation problem remains implementable when the social objective is defined over a distribution of constituency types. It does not turn the problem into one about fractional outcomes.
The supporting anchor is Theorem 1, also proved here. For a collection of nonempty closed sets \(C_i(\theta)\subseteq L_i((f(\theta),v^f),\theta)\), it states that \(f\) is dynamically monotonic with respect to \(C\) if and only if (1) \(f\) is Maskin monotonic* with respect to \(C\), and (2) \(v^f\) is the unique efficient equilibrium payoff vector of the associated repeated game \(G_C\). The paper notes that this characterization makes DM practically testable: MM* requires finitely many incentive checks, while the equilibrium payoff set can be approximated numerically.
Its mirror is a more focused computational problem. Supply the same population data, a target \(f_\mu\), and explicit rational-polytope sets \(C_t(\theta)\) describing the alternative and continuation-payoff pairs each type organization can secure by deviating. Decide whether \(f_\mu\) is dynamically monotonic with respect to those sets. Equivalently, check MM* and whether \(v^{f_\mu}\) is the unique efficient equilibrium payoff vector in the associated game with one player per type. A numerical version asks for an \(\varepsilon\)-approximation of that game’s equilibrium payoff set, then uses Theorem 1 to assess DM when the relevant payoff frontier is separated from \(v^{f_\mu}\) by a known margin.
I would tentatively expect the approximation version to be a Class A candidate when \(\tau\), \(|A|\), and \(|\Theta|\) are moderate: the paper gives a finite MM* check and points to repeated-game numerical methods for the payoff set. That is a research direction, not a polynomial-time result. Neither theorem proves a complexity bound, and the paper gives no formal guarantee that the numerical approximation settles instances near the efficient frontier. There is no NP-hardness, P, or parameterized-complexity theorem here to serve as a conventional complexity anchor; the case rests on these two computationally actionable characterizations.
The mirror covers repeated implementation and the numerical test for dynamic monotonicity, not all of the paper. In particular, I would not use Proposition 2 as an anchor: its outcome set is \(A=[0,1]\), so its continuity is in the outcome space, which is outside this programme’s scope. The proposed mirror keeps \(A\) finite and puts the distribution \(\mu\) over the population’s preference types.
The weakest point is the move from many individuals to one strategic player per type. In an ordinary nonatomic population, a single person’s deviation may not change aggregate reports or the outcome, so the paper’s individual Nash incentives can disappear. The mirror therefore needs a credible institution: type cohorts must coordinate through representatives who act for them. If the authors or a referee reject that as changing the strategic problem rather than adapting it, this positive case fails. The next questions are whether the finite-population model with \(N_t/N=\mu_t\) has a useful two-way relation to the organization model, and whether the approximation can be turned into exact complexity bounds in \((|A|,\tau,|\Theta|,L)\).
The cleanest objection is that neither anchor is a computational result of the kind ChoCo is meant to continuize. Theorem 2 characterizes when a social choice function is repeatedly implementable; Theorem 1 characterizes dynamic monotonicity through MM* and an equilibrium payoff set. The paper calls the latter practically testable, but gives no exact, approximation, or parameterized complexity bound. Turning either characterization into a complexity problem could be worthwhile, but it would be a new computational question. The question is whether the proposed population continuum gives that question useful content.
The strongest version of the proposed mirror still faces a basic modeling split. If the many members of each type are ordinary nonatomic agents, an individual’s deviation has measure zero and does not change an outcome determined by aggregate reports. The paper’s whistle-blower incentives then disappear. Allowing a whole type cohort to deviate together restores them, but changes Nash implementation into a problem with coordinated or coalitional deviations. If instead each cohort has a binding representative who chooses for everyone, the game has one strategic player per type. That may describe unions or provider associations, but those organizations are the players; their members are not a high-multiplicity population of the paper’s agents.
In that representative model, a type’s positive mass does not itself affect its incentive constraints. Multiplying the representative’s utility by \(\mu_t>0\) leaves its best responses unchanged; the dynamic game and its equilibrium structure are unchanged up to this rescaling. Mass can affect the target \(f_\mu\) if the designer imposes a mass-weighted objective, as the proponent proposes. But then the continuum’s role is to select a target among finitely many possible tables \(f:\Theta\to A\). With finite \(A\) and \(\Theta\), varying \(\mu\) changes that table across finitely many regions of the mass simplex; it does not turn the repeated implementation game into a high-multiplicity problem. This could be parametric implementation by organizations, but the computational work remains in repeated-game incentives and payoffs.
Theorem 1’s numerical angle does not repair that mismatch. Its proposed input must include the deviation sets \(C_i(\theta)\), or a way to derive them, in addition to the population data. Given explicit rational polytopes, approximating the associated repeated game’s payoff set is a coherent computational problem. But the paper gives no complexity guarantee for it, and a finite-precision approximation cannot settle uniqueness of an efficient payoff vector when the frontier touches the relevant boundary. A promised separation gap could make a useful approximation problem; its hardness or tractability would still come from computing the repeated-game payoff set, not from population mass.
The strongest version of Theorem 2’s mirror would ask for the complexity of deciding implementability for mass-weighted constituency targets, perhaps by using Theorem 1 as a test. That is better posed than asking whether the theorem itself is “continuous.” Yet if mass affects only \(f_\mu\), the population axis just chooses among target functions; if mass is meant to affect strategic behavior, the model must add an aggregation or coalition rule that the paper’s Nash game does not provide. Either choice leaves the proposed mirror without a demonstrated computational gain from continuization.
There is a plausible constituency scenario, so I cannot honestly rule out every worthwhile application. The negative case is strongest on fit: binding organizations may be sensible players, but that makes this a repeated implementation problem for organizations, not a continuous-population version of the paper’s individual-agent problem. The authors might welcome that as a new domain. On the evidence here, though, neither theorem supplies a complexity result for ChoCo to mirror, and making the population continuous either removes the paper’s individual incentives or replaces its players with collective actors.
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.