| paper | d4-conf-aaai-1996-0085 |
| authors | — |
| venue | ? |
| filed under | unclassified |
| judged by | gpt-6-luna / xhigh (triple__luna__xhigh__c6r1) |
| judge confidence | high |
| authors would recognise it | no |
The numbered theorems establish deal existence and payoff properties, but none states an algorithm, complexity bound, hardness result, or computational formulation used by the protocol. A population of repeated negotiations would require a new aggregate problem, so it cannot supply the missing computational result to continuize.
fails bit a — no named computational result to mirror
Theorems 2.1, 2.2, and 3.1–3.4 concern deal enablement and Pareto improvement; none yields a continuous computational mirror.
This paper contains no named computational-complexity result of the required kind. Its named Theorems 2.1 and 3.1/3.3 establish that leveled-commitment contracts can enable individually rational deals unavailable under full commitment; Theorems 2.2 and 3.2/3.4 establish Pareto improvements under stated conditions. None classifies an underlying computational problem as tractable or hard, so none can serve as the required anchor.
The paper’s probability distributions over future outside offers are uncertainty about what may happen to two named agents. They are not a continuous population. One could imagine a large platform handling many repeated contractor–contractee negotiations, grouped by bilateral game type, but optimizing contracts across that population would introduce a new aggregate problem. The paper’s theorems do not establish its computational properties, and it would be a stretch to present it as a mirror of their results.
So I can’t give a compliant continuous problem per anchor here. The paper studies automated contracting, but it offers no named computational result for this programme to continuize.
The negative case starts with the paper’s actual results. Theorems 2.1 and 3.1/3.3 show that leveled commitment can enable individually rational deals; Theorems 2.2 and 3.2/3.4 show Pareto improvements under stated conditions. None classifies a computational problem. They analyze two named agents’ equilibrium choices under uncertainty, using distributions over outside offers. Those distributions make uncertainty part of the model; they do not make the society continuous.
A stronger computational question could ask whether, given finite rational distributions of outside offers, there are penalties and a price that yield an individually rational equilibrium—or which terms maximize expected welfare. That would be a reasonable algorithmic extension of the paper. But it keeps the society bilateral and finite. Treating the offer realizations as voter-like types would confuse possible states of an agent’s future with distinct agents in a population.
A population version has a dilemma. If a platform runs many independent negotiations, grouping identical bilateral games by type only repeats the paper’s analysis: without shared constraints, solve each game and aggregate the payoffs. If negotiations are coupled by matching, scarce resources, or chains of interdependent contracts, the mass-allocation problem could become nontrivial. But those couplings supply the new computational problem; the paper’s theorems do not formulate or analyze it. Its discussion of contract webs does not fix this gap: there, who is connected to whom matters, and type masses alone do not describe the network.
So the strongest case against a worthwhile mirror is that the paper offers no computational result to continuize, while the most plausible population extensions either aggregate independent copies or introduce a different market-design problem. That is a strong case against a mirror of this paper, though not proof that no valuable continuous bargaining problem could be built from its ideas.
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.