Expanding the Reach of Social Choice Theory

The decisive objection is provenance. This is an overview article, not a paper presenting numbered computational theorems. Its claims that connected fair allocation is \(NP\)-hard, that connected-cake decisions have complexity results, and that fair-allocation reachability has varying complexity are reports of Goldberg et al. [2020], Igarashi et al. [2024], and other cited work. A continuous version of those problems would be a mirror of the cited papers, not of this paper. Under the stated anchor rule, the proponent is right that the anchor set is empty.

Their wording that the paper contains “only axiomatic” material is too strong: it does contain computational assertions. But none is a named computational result of this paper in the required sense. That distinction is enough to defeat the proposed case.

The conditional connected-allocation mirror also fails as a direct rescue. With finitely many indivisible line items and an atomless population, at most finitely many agents can receive nonempty bundles, hence almost all agents receive nothing. Positive proportionality or meaningful envy guarantees then collapse. Allowing type-level fractions of an item changes the resource into a divisible one; scaling the number of items with the population requires a second limit over item positions and turns the problem into a new transport or density model. Preserving contiguity for each individual still requires information about each agent’s interval, which an agent-type mass vector does not contain. That two-sided model might be interesting, but it is no longer a population-only mirror of the reported result.

The same obstruction is sharper for reachability. The input consists of allocations to named agents, a named target allocation, and sequential exchanges. Quotienting agents by type forgets which agent owns which bundle, although that can change whether a target is reachable. Retaining those labels destroys multiplicity. Moreover, one exchange has zero mass in an atomless population: replacing it by a positive-mass transfer changes sequential exchange into bulk reallocation. Thus the best type-level reformulation is a different reachability problem.

The tournament material has no population axis at all. Margin of victory concerns perturbations of a tournament relation, while seeded knockout fixing concerns finitely many named players and bracket positions. A continuum of spectators or voters is irrelevant; a continuum of players has no finite bracket or single knockout winner. A distribution of player strengths could define a new probabilistic tournament model, but not a continuous mirror of the paper’s computational objects.

The remaining candidates do not repair this. Weighted picking sequences depend on discrete turns and indivisible items; a continuum Nash product would require redefining the objective as something like an integral of log utility; differential privacy is based on changing one individual, whose mass becomes zero; externalities encode identity and social-network structure. Cake sharing and budget aggregation already continuize outcomes, not populations, and the paper’s results there are primarily axiomatic rather than computational.

So the negative case is strong under the programme’s rules: this overview supplies no eligible computational anchor, and its nearest computational passages either belong to cited papers or lose their meaning under a population-only continuum. The universal claim that no conceivable high-multiplicity fair-division model could ever be worthwhile would be too strong—especially if one allowed a carefully designed two-sided scaling of agents and goods—but that concession cannot turn this paper into a worthwhile continuous-mirror target.