| paper | d4-journals-mss-2019-0013 |
| authors | — |
| venue | ? |
| filed under | unclassified |
| judged by | gpt-6-luna / xhigh (triple__luna__xhigh__c6r1) |
| judge confidence | high |
| authors would recognise it | no |
Theorem 2 has constructive computational character because the paper gives formulas for recovering an individual decision maker’s parameters from small-menu choices. The proposed population version instead asks whether aggregate shares identify a distribution over behavioral types, a different inverse problem that the paper neither studies nor implies. The remaining results are axiomatic or representational, so no continuous population question survives as a mirror of a named result.
fails bit b-changes — the continuum limit changes the question
The nearest candidate concerns Theorem 2’s small-menu identification result; the axiomatic characterizations and random-utility representation results have no surviving computational population mirror.
The strongest positive lead is Theorem 2, proved in this paper: for a standard choice function represented by \((\gamma,\succ)\), the consideration probabilities \(\gamma\) and preference order \(\succ\) are uniquely identified from choices on menus of size two and three. But that is an identification result, not a named computational result: it does not classify an algorithmic problem as tractable or hard.
The other tempting anchors do not fill the gap. Theorem 1, proved here, characterizes independent RCM by four axioms. Theorem 3, proved here, shows that standard independent RCM is independent RUM. Propositions 1–3 and Theorem 4, proved here, give related results for choice with a default. None asserts a complexity classification for a computational problem. So this paper supplies no qualifying named anchor for the requested case.
There is a plausible adjacent population model: let a consumer type be \((\gamma,\succ)\), let \(\mu_t\) be the fraction of consumers of type \(t\), and define the aggregate share choosing \(a\) from menu \(A\) as \(\sum_t\mu_t p_t(a,A)\). One could ask whether observed aggregate choice rates identify \(\mu\), or whether a given rate table has such a representation. Theorem 2 does not establish identification for this heterogeneous population: it concerns one choice function represented by one pair \((\gamma,\succ)\). The paper’s random choice probabilities also cannot by themselves serve as evidence that the authors studied a continuous society.
Thus I can’t honestly give a prover-ready continuous problem *per anchor* or predict its place in the tractable/hardness trichotomy. The weak point is decisive: the paper studies axioms, identification, and random-utility representation for a decision maker’s choice behavior, not computational problems about a population or a collective decision.
The strongest case against a mirror is that none of the paper’s results is a computational result. Theorem 1 and Proposition 1 are axiomatic characterizations; Theorem 2 identifies one decision maker’s parameters; Theorems 3 and 4 and Propositions 2 and 3 are representation and identification results. There is no named problem here whose complexity or algorithms a continuous-population version could classify.
Theorem 2 is the best candidate for repair. A population of consumers with types \((\gamma,\succ)\) is a sensible high-multiplicity model, and its aggregate choice rates would be
\[
\bar p(a,A)=\sum_t \mu_t p_{\gamma_t,\succ_t}(a,A).
\]
One could ask whether those rates identify the mass distribution over types. But Theorem 2 starts with the choice function of a single decision maker and identifies that person’s \((\gamma,\succ)\). Recovering a population mixture from pooled rates is a different inverse problem. In a homogeneous population, aggregation leaves \(p\) unchanged, so the population dimension adds nothing to the theorem. In a heterogeneous population, the new unknown is the distribution of behavioral types, which the paper does not study. That may be a worthwhile econometric question; it is not a computational social choice problem supplied by this result.
The same mismatch holds for the other results. A computational test for whether aggregate rates admit an independent RCM representation would turn Theorem 1 into a model-recognition problem, while an aggregate random-utility test would recast Theorems 3 and 4. Those are possible new questions, but they concern fitting or identifying a stochastic choice model. The paper has no collective rule, winner, or intervention objective to carry into them. Adding a goal such as changing market shares would create a new optimization problem rather than mirror a result in the paper.
This does not show that no useful continuous-population research could be built near the paper. A population of consumers is plausible, and heterogeneity may make it interesting. The negative case is narrower: every proposed route from these named results to computation changes the object from an individual’s stochastic choice behavior to population-level inference or policy. That leaves no strong, paper-anchored continuous mirror for ChoCo’s complexity programme.
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.