| paper | Negotiable Votes |
| authors | Umberto Grandi, Davide Grossi, P. Turrini |
| venue | JAIR 2019 |
| filed under | voting · manipulation |
| judged by | gpt-5.6-luna / xhigh (triple__luna__xhigh__c6r1) |
| judge confidence | high |
| authors would recognise it | unclear |
Propositions 19(2) and 19(3) clearly satisfy the computational-result gate, but the proposed continuous questions do not settle what repeated types may control. Common type ballots preserve the agenda hardness only by turning each type into a strategic bloc; independent atomless copies lose pivotality and do not preserve the paper’s Nash-and-transfer game. Whether the bloc interpretation is a legitimate, authors-recognisable high-multiplicity version is therefore unresolved.
fails bit b-changes — the continuum limit changes the question
The proposed mirror must choose between strategic blocs, which collapse the population to finitely many negotiators, and independent atomless voters, who are non-pivotal.
fatal: True
It is undecided whether type-level common ballots and positive-mass deviations are an acceptable high-multiplicity interpretation of the paper’s individual Nash negotiation, rather than a finite bloc game. A formal discrete-to-continuous dictionary fixing agency, transfers, and deviations would settle the grade.
The proposed mirror covers Propositions 19(2) and 19(3); it leaves the equilibrium-selection theorems and structural results without continuous formulations.
The strongest honest case is a qualified yes, led by Proposition 19(2). The right mirror is not merely to replace a finite voter count by a real number. It is to study pre-vote negotiation among a large population represented by finitely many repeated stakeholder types.
Let \(\mathcal T\) be a finite set of types, with mass distribution \(\mu\in\mathbb Q_{\ge 0}^{\mathcal T}\), \(\sum_{t\in\mathcal T}\mu_t=1\). A type \(t\) includes the complete information relevant to the game: an arbitrary goal formula \(\gamma_t\), an outcome utility \(u_t\), and its voting influence. The regime is a large referendum or federation with \(N\gg |\mathcal T|\): many members share the same goals, utilities, and institutional role. For rational \(\mu_t=a_t/A\), this is the high-multiplicity society obtained by repeating type \(t\) \(a_t\) times.
There are \(m\) binary issues and ballots \(b_t\in\{0,1\}^m\). A natural continuous aggregator is a weighted quota rule,
\[ F_{\mu}(b)(j)=1 \quad\Longleftrightarrow\quad \sum_{t\in\mathcal T}\lambda_{j,t}\mu_t b_t(j)\ge q_j. \]
With \(\lambda_{j,t}=1\), this is ordinary mass-based quota aggregation; nonuniform \(\lambda_{j,t}\) allows the role-weighted and dictatorial aggregators already admitted by the paper. A negotiation schedule specifies nonnegative transfers \(\Theta_{t,u}(b)\) between types at every type-level ballot profile. The adjusted per-capita payoff of type \(t\) is
\[ u_t(F_\mu(b))+ \frac{1}{\mu_t} \left( \sum_{r\ne t}\Theta_{r,t}(b) - \sum_{u\ne t}\Theta_{t,u}(b) \right). \]
A type-level Nash equilibrium is a profile \(b\) at which no positive-mass type can change its common ballot and obtain a lexicographically preferred result: satisfying \(\gamma_t\) is preferred to not satisfying it, and adjusted utility breaks ties. This is the necessary population-level co-modification: literal atomless individual Nash would make each voter non-pivotal and largely destroy the paper’s strategic phenomenon. The transfer schedule should be represented succinctly, for example by rational arithmetic circuits, so the question remains computationally well-defined.
The paper’s strongest computational anchor is Proposition 19(2), stated here and justified by a translation from Wooldridge et al. (2013), Proposition 6. It says that, for aggregation games with arbitrary goals, deciding whether there exists a transfer function \(\tau\) and a ballot profile \(B'\) such that \(B'\) is a Nash equilibrium of \(\tau(A)\) and the collective outcome satisfies every player’s goal, \(F(B')\models\bigwedge_i\gamma_i\), is \(\Sigma^p_2\)-hard.
Its continuous mirror is:
Continuous Good-Equilibrium Existence. Given \((\mathcal T,\mu)\), the type goals \(\gamma_t\), utilities \(u_t\), the weighted-quota aggregator \(F_\mu\), and a succinct transfer-schedule representation, does there exist a schedule \(\Theta\) and a type ballot profile \(b\) such that \(b\) is a type-level Nash equilibrium under \(\Theta\) and
\[ F_\mu(b)\models\bigwedge_{t:\mu_t>0}\gamma_t? \]
A yes-certificate is precisely the schedule \(\Theta\) together with the equilibrium profile \(b\). There is no optimisation objective in the anchored decision problem; a natural extension would minimise total transfer mass subject to the same equilibrium and efficiency conditions.
I would expect this problem to be Class B: hardness should transfer. The Boolean-game reduction underlying Proposition 19 can be embedded by giving each Boolean-game player a positive-mass role type and using weighted quotas to make each issue controlled by the appropriate type. Replicating every role many times changes \(N\) but not the type-level problem. The combinatorics live in the issue agenda and arbitrary goal formulas, not in the number of individually named voters. Thus continuization removes population multiplicity without removing the source of hardness.
A second worthwhile anchor is Proposition 19(3), also stated in this paper and inherited from Wooldridge et al. (2013), Proposition 14. It says that deciding whether there exists a transfer function \(\tau\) such that every ballot profile \(B'\) that is a Nash equilibrium of \(\tau(A)\) satisfies all players’ goals is \(\Sigma^p_2\)-hard.
The corresponding problem is:
Continuous Robust Negotiation. Given the same population-level instance, does there exist a transfer schedule \(\Theta\) such that every type-level Nash equilibrium \(b\) under \(\Theta\) satisfies
\[ F_\mu(b)\models\bigwedge_{t:\mu_t>0}\gamma_t? \]
This is the continuous version of asking whether negotiation can eliminate every bad equilibrium, not merely select one good equilibrium. In a nonvacuous variant, one additionally requires that at least one type-level equilibrium exist; the paper’s literal universal statement does not explicitly impose that condition.
The pure type-profile version should again be Class B, because the universal quantification ranges over the same agenda-driven equilibrium structure as in the finite problem. If one instead allows each type’s mass to split continuously across ballots, the equilibrium set becomes genuinely geometric; that richer version could become Class C, with hardness caused by the continuum rather than merely inherited from Boolean games.
The paper’s authors should recognise these as their problem. The mirror retains binary issues, ballot goals, outcome utilities, the same lexicographic preferences, non-manipulable or quota aggregation, pre-vote side-payments, and Nash equilibrium. It changes only the population representation and makes transfers and influence per-capita or mass-weighted. It is not outcome-space continuity, smoothed analysis, or an axiomatic profile simplex. Nor is it a novelty collision: the paper studies finite players, and Proposition 19’s Boolean-game translation is evidence for a high-multiplicity analogue rather than an existing continuous-population treatment.
The weakest point is the equilibrium notion. A literal continuum of individually acting voters is non-pivotal, so the mirror must use type-level or positive-mass deviations. A referee may regard that as changing the game rather than continuizing it. The defence is that large federations, parties, unions, and institutional stakeholder groups genuinely negotiate through type-level mandates, and the paper already assumes a binding authority enforcing transfers. Still, the positive case should not claim that the paper’s \(\Sigma^p_2\)-hardness automatically transfers: a new reduction is needed, especially if one insists on anonymous quotas and independently acting members. That is precisely the substantive continuous-computation question this mirror would generate.
The negative case is that Proposition 19 is not really a population result. It is a result about named strategic agents: who can change a ballot, who can pay whom, and which continuation each individual selects.
This is especially clear in the pre-vote phase. A transfer is indexed by a payer, a recipient, and a complete ballot profile, \( \tau_i(B,j) \). Even when the underlying game is uniform and anonymous, negotiation is not: the identity of the payer and recipient is part of the mechanism. High multiplicity can identify agents with identical preferences and costs, but it does not turn those agents into one decision-maker.
That defeats the proposed mirror of Proposition 19(2). If all copies of type \(t\) are required to choose one common ballot \(b_t\), then a deviation by \(t\) changes a positive mass of voters at once. This is a coalition or bloc deviation, not the unilateral deviation in the paper. Weighted quotas and coefficients \( \lambda_{j,t} \) do not repair this; they simply encode each type as a weighted super-voter. The Boolean-game reduction is preserved only by granting each type collective control over its entire mass. The continuous population has then disappeared as a strategic population: there are merely \( |\mathcal T| \) finite negotiators.
If, instead, the copies act individually, the natural atomless limit has the opposite problem. A single voter has measure zero and cannot change any quota outcome. Under an anonymous transfer rule depending only on aggregate type masses, that voter cannot change the transfer state either. Hence the central strategic fact used by Proposition 19—the ability of one player to alter the resulting game or outcome—vanishes. Allowing positive-mass deviations would produce a strong-equilibrium or coalition-deviation model, not a continuous version of the paper’s Nash equilibrium. Allowing action-contingent individual payments retains the missing identities and yields a mechanism-design problem over a continuum of agents, not the proposed finite type-level game.
The same dichotomy defeats Proposition 19(3), more sharply. In an atomless individual model, many different microscopic ballot assignments generate the same aggregate society and are indistinguishable to the aggregator. A universal claim over every Nash equilibrium therefore quantifies over arbitrary microstates rather than over the finite equilibrium structure of the paper. If one quotients those microstates away, one obtains the finite weighted game of type blocs. Requiring at least one equilibrium only removes vacuity; it does not restore the lost individual strategic structure.
The proponent’s circuit representation for \( \Theta \) addresses encoding, but not this semantic problem. There is no canonical way to decide whether transfers are made per person or per mass, whether a type may coordinate, who has authority to commit the mass, or whether a positive-mass deviation is permitted. Those are not technical consequences of continuization; they define a different negotiation game.
A federation of caucuses, unions, or institutional representatives could certainly be modelled with mass-weighted voting and type-level transfers. But that would be a new finite-agent model of bargaining among blocs. It may be interesting, yet it is not a continuous mirror of *Negotiable Votes*: the population mass supplies voting weights while all meaningful agency remains concentrated in finitely many representatives. Conversely, preserving individual agency makes the atomless limit non-pivotal and destroys the paper’s negotiation mechanism.
So neither Proposition 19(2) nor Proposition 19(3) survives as a genuine population continuization. The negative case is not a proof that no related model could ever be invented; its point is that every plausible repair must either collapse the population into strategic blocs or replace the paper’s Nash-and-transfer model with a new coalitional or nonatomic mechanism.
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.