| paper | d4-conf-atal-2010-0171 |
| authors | — |
| venue | ? |
| filed under | unclassified |
| judged by | gpt-6-luna / xhigh (triple__luna__xhigh__c6r1) |
| judge confidence | high |
| authors would recognise it | no |
The paper has no numbered theorem, lemma, corollary, or proposition asserting an algorithmic or complexity result; its numbered items are definitions, and its experiments do not provide a qualifying anchor. The proponent also acknowledges that the cohort objective and query protocol would be a prospective new model, so the paper yields no qualifying continuous mirror.
fails bit a — no named computational result to mirror
No named computational result is mirrored; the paper's query method, heuristics, and simulation findings remain individual-user results.
The best positive angle is a new batch version of the paper’s elicitation problem. Imagine a hospital decision service supporting many patients who face the same finite menu of risky options. A type would be a complete CPT response model for that menu, including its utility and probability-weighting functions; \(\mu_t\) would be the fraction of patients of type \(t\). A system could choose prospect-comparison queries for these cohorts and terminal decisions to minimize mass-weighted worst-case regret under a query budget. That makes the population itself continuous; the lotteries remain part of the original decision problem.
But this is only a prospective extension. The paper has no numbered theorem, lemma, corollary, or proposition asserting a computational complexity result. Its numbered items are definitions, and its results are a querying method, heuristics, and experiments. So there is no qualifying anchor to restate, and no per-anchor continuous problem or Class A/B/C prediction that can honestly be attributed to the paper.
The main obstacle is that Hines and Larson study elicitation for an individual with an unknown preference model. Their mention of a probability distribution over utility functions concerns uncertainty about that user’s preferences; it does not make society a distribution of agents. In the proposed cohort version, the aggregate regret objective and shared-query protocol are new. The authors might find them interesting, but the paper gives no basis to say they would recognize them as the continuous form of their problem. If patients truly share the same complete type, one patient’s answers may also stand in for the whole cohort, leaving little population-driven computation to study.
There is no named computational result in this paper to anchor a continuous mirror. Its numbered statements are definitions; the contributions are a query method, two heuristics, and finite simulations. The simulations test elicitation for one user with a specified utility and weighting function. They do not establish a complexity result about a population. That is the strongest reason against treating this paper as a ChoCo case.
The most promising extension is the cohort version the other analyst proposes: let \(\mu_\theta\) be the share of patients with complete CPT model \(\theta\), then allocate a query budget to reduce regret across patients. If patients receive individual recommendations and exact answers, elicitation for each distinct \(\theta\) remains its own problem. Population mass weights those problems; the relevant scale is the number of distinct types \(\tau\), not the number of patients. A budget that couples query allocation could make a new optimization problem, but the paper neither formulates nor analyzes it.
A shared decision for everyone creates a more substantial collective problem. It needs a social objective—perhaps average regret, \(\sum_\theta \mu_\theta R(d,\theta)\), or worst-case regret across types—and a protocol for learning about the population’s preferences. Those choices define a new group elicitation model. Average regret changes the decision criterion; worst-case regret can ignore masses except for whether a type is present. If responses are sampled or noisy, the model also needs a sampling process that the paper does not study. Such a project could be worthwhile, but its value would come from the newly specified collective problem, not from continuizing one of the paper’s results.
The paper’s discussion of a probability distribution over utility functions is uncertainty about one user’s preferences. A distribution over complete user types would instead describe a population, so that passage does not supply the missing population model.
The negative case is therefore strong against a direct mirror of this paper: there is no computational theorem to mirror, and its elicitation queries address private preferences one user at a time. It is weaker against the universal claim that no scenario could yield a worthwhile project. A carefully defined cohort elicitation problem may be worth studying; the paper alone gives little basis for claiming that it is the continuous mirror of these authors’ problem.
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.