| paper | d4-journals-mss-2015-0049 |
| authors | — |
| venue | ? |
| filed under | unclassified |
| judged by | gpt-6-luna / xhigh (triple__luna__xhigh__c6r1) |
| judge confidence | high |
| authors would recognise it | no |
Theorem 5.1 has constructive computational content because it gives a recurrence for the selected outcome. But the proposed continuous population consists of independent two-player games: it batches instances without making the paper’s two negotiators a population whose bargaining outcome is being computed. The continuous-population gate therefore fails for the sole anchor.
fails bit b-none — no population with multiplicity to relax
The rejected candidate relies on Theorem 5.1’s multi-round outcome recurrence and its large-horizon limit; it leaves Theorem 4.1 and the discussion of general commitment-cost functions unmirrored.
The strongest honest positive case is narrow: the paper studies two-player bargaining, so the plausible population is a large collection of bargaining episodes, not a continuum inside one negotiation.
My anchor is Theorem 5.1, proved in this paper. For a game with \(N\) potential rounds, commitment cost parameter \(\lambda<1-\delta\), and first-round proposer \(p\) and responder \(r\), it says common belief in future rationality selects a unique outcome: \(p\) commits to \(0\), \(r\) commits to \(x_r^N\), \(p\) proposes \((x_p^N,x_r^N)\), and \(r\) accepts. The amounts are given recursively by
\[
x_p^1=\delta y_p,\qquad x_r^1=1-\delta y_p,
\]
and, for \(N\ge 2\),
\[
x_p^N=\delta(1-\lambda)x_r^{N-1},\qquad
x_r^N=1-\delta(1-\lambda)x_r^{N-1},
\]
using the paper’s convention for the discrete share grid. Here \((y_p,y_r)\) is the terminal division if all offers are rejected. Since every subgame-perfect equilibrium satisfies common belief in future rationality, the theorem also implies that the subgame-perfect equilibrium outcome is unique. The paper further describes the limit as the horizon grows, the grid becomes fine, and commitment costs vanish: the shares approach the reversed Rubinstein outcome, \(\bigl(\delta/(1+\delta),1/(1+\delta)\bigr)\).
Lead mirror: High-Multiplicity Commitment-Bargaining Outcome. An instance is a finite support \(T\) of bargaining-environment types, with rational masses \(\mu_t\ge 0\) summing to \(1\). A type \(t\) specifies a complete bargaining episode: the two roles, their discount factor \(\delta_t\), commitment cost \(\lambda_t\), horizon \(N_t\), share grid, terminal division, and first proposer. Each type obeys the paper’s assumptions, including \(\lambda_t<1-\delta_t\). Mass means the fraction of a large set of independent bargaining episodes having that complete environment.
The task is to compute, for every supported type, the selected first-round commitments, proposal, and accepted division, and then the mass-weighted aggregate shares and net payoffs. A solution is the type-by-type outcome together with those aggregates; each type’s answer must satisfy the paper’s common-belief-in-future-rationality prediction. There is no separate welfare-maximization objective: the question is to compute the selected outcomes. This keeps the negotiation protocol and its discrete share grid intact; only the population of episodes becomes a distribution.
The natural regime is a large batch of standardized negotiations—say, many contract renewals or procurement negotiations grouped by horizon, patience, commitment costs, and fallback terms. There may be a very large number of episodes but only \(\tau\) distinct environment types. For a discrete batch with \(n_t\) episodes of each type, \(\mu_t=n_t/\sum_s n_s\); the continuous instance is its high-multiplicity representation. The model is sensible if these episodes really share all the parameters the game uses.
The expected direction is tractable. Theorem 5.1 supplies a recurrence for each type, followed by weighted summation. With \(N_t\) counted as a parameter, the computation takes \(O(\sum_t N_t)\) recurrence steps plus aggregation, with exact rational arithmetic. It yields a clean bridge from many repeated finite games to aggregate outcome statistics. Further questions include whether the same aggregation remains tractable for richer type spaces, such as continuous distributions over \((\delta,\lambda)\), and what happens when episodes share a budget that couples their commitments.
The weakness is that this is an aggregation of independent games. It creates no strategic interaction between masses, and Theorem 5.1 already solves each episode’s equilibrium outcome. The authors might recognize the task as high-multiplicity computation for their bargaining model, while seeing it as an application of their result rather than a new bargaining question. Also, the paper contains no named theorem establishing a complexity classification such as membership in P or NP-hardness. Theorem 5.1 is an exact equilibrium characterization that induces a computational task; it should not be presented as a complexity result proved by the paper.
The decisive weakness is that this paper proves no computational-complexity result. Theorem 5.1 characterizes a selected outcome under a solution concept; it does not classify an algorithmic problem. The proposed mirror creates a task by asking us to compute those outcomes for many bargaining episodes and average them. That is a possible computation, but it is not a complexity question the paper poses.
The batch scenario is the strongest version of the mirror, and it is a sensible high-multiplicity regime: many standardized negotiations can share a complete environment type. But the negotiations are independent. For each supported type, Theorem 5.1 already supplies a recurrence for the outcome; then one takes a mass-weighted sum. The number of episodes disappears, and the remaining work is per-type evaluation plus arithmetic aggregation. That makes a useful application or implementation of the theorem, but offers little of the computational landscape ChoCo is meant to chart.
A richer mirror could ask how to change the aggregate division by shifting mass among episode types. But that adds a planner, a goal, and costs for changing the population—none of which the paper’s bargaining procedure specifies. Alternatively, putting a continuum inside a single negotiation would require a new collective bargaining protocol: this paper has one proposer and one responder, and gives no way to turn either role into a population whose mass can make offers or commitments. Those extensions may be worthwhile research, but they would not be continuizations of Theorem 5.1.
The negative case is not airtight. The batch of repeated negotiations is a plausible continuous population, and the authors might recognize aggregate outcome computation as useful. What is missing is a substantive computational problem beyond evaluating their already-characterized independent games and aggregating the results.
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.