| paper | d4-conf-aaai-2018-0406 |
| authors | — |
| venue | ? |
| filed under | unclassified |
| judged by | gpt-6-luna / xhigh (triple__luna__xhigh__c6r1) |
| judge confidence | high |
| authors would recognise it | no |
The atomless question is formally well specified, but one agent's report has zero mass, so pointwise impartiality is automatic and the Borda order follows from sorting aggregate scores. The proposed positive-mass manipulation repair is a coalition question that Theorem 3.2 does not establish, so the mirror changes the individual-impartiality problem the paper studies.
fails bit b-changes — the continuum limit changes the question
A single agent's report cannot affect the aggregate Borda scores in the atomless model, so the paper's individual-impartiality constraint loses force; the proponent notes this but does not show why the authors would recognize the resulting problem as theirs.
fatal: True
The proposed mirror addresses Theorem 3.2 alone; it leaves Theorems 4.1 and 5.1, which give guarantees for k-PARTITE and COMMITTEE, without a continuous analogue.
The strongest anchor is Theorem 3.2, proved in this paper. It says that for every \(n\ge 2\) and every \(\varepsilon>0\), no impartial rank aggregation rule achieves \((1/2-\varepsilon,1/3)\) forward error with respect to Borda. In full: on every input profile, a rule with those parameters would have to place *every* player within \(n/3\) positions of their Borda position with probability at least \(1/2+\varepsilon\); Theorem 3.2 rules this out. The proof constructs a profile where changing one player’s report can put that player at an extreme Borda position, while impartiality requires their output-position distribution to stay fixed. This is a named impossibility result, not an NP-hardness or polynomial-time classification; the paper states no such complexity classification.
A plausible mirror is Atomless Impartial Borda Ordering. Let \(X=[0,1]\) be a population of unit mass, with a measurable assignment \(\theta:X\to T\) to a finite set of ranking types. The mass of type \(t\) is \(\mu_t=\lambda(\{x:\theta(x)=t\})\), where the \(\mu_t\) are rational. Each type \(t\) has a complete ranking of the population, represented by a rank-quantile function \(\rho_t:X\to[0,1]\), with smaller values meaning higher rank. For a computational version, take the type assignment and ranking functions to have explicit finite, piecewise-rational encodings.
The continuous Borda score of candidate \(x\) is \(b_\mu(x)=\sum_{t\in T}\mu_t(1-\rho_t(x))\). Let \(Q_\mu(x)\) be their position in the order by decreasing \(b_\mu\), with ties broken by a fixed public ordering of identities. A solution is a rule that, given the encoded society, outputs a measurable ranking \(R:X\to[0,1]\). It is impartial if changing the reported type of any one point \(x\), with the rest of the society fixed, leaves the distribution of \(R(x)\) unchanged. The question is whether one can compute an impartial rule that returns \(R(x)=Q_\mu(x)\) for every \(x\), or, more strongly, characterize the best achievable forward-error guarantees.
For this model, I expect Class A: the aggregate Borda scores are computable from the type masses and ranking functions, and sorting them gives the target order. A single point has zero mass, so changing its report does not change those aggregate scores. With finite piecewise-rational inputs, computing the score functions and their induced order is a finite calculation. The ranking remains the social outcome; its position coordinate is simply normalized to describe an atomless population.
The regime is narrow but recognizable: a very large peer-assessment cohort in a standardized market, where many participants use one of a small menu of complete evaluation templates. Then \(n\) can be enormous while the number of distinct ranking types \(\tau\) is much smaller. For particularly homogeneous cohorts, templates might rank profile archetypes, with a public tie-break among candidates in each archetype. That coarsening is a modeling choice to make repeated types plausible; it should be disclosed rather than presented as identical to the paper’s individual rankings.
This is a mirror of the paper’s question about impartial Borda ranking, but the expected answer differs: the finite impossibility need not survive when each individual’s vote has zero population mass. In that sense, Theorem 3.2 is a useful anchor for asking whether the obstruction dissolves under high multiplicity. The mirror covers that theorem only; it does not claim to continuize the paper’s whole algorithmic contribution.
The weak point is also the central reason the mirror may be valuable: atomless impartiality can become easy precisely because unilateral influence vanishes. The finite theorem exploits a single report breaking a tie and moving its owner across the ranking. In the continuum, that report has zero weight, and tie-breaking must be specified carefully. The authors might regard the result as a meaningful population limit, or might say that the distinctive strategic problem has disappeared. Natural follow-ups are to study deviations by a positive mass of agents, quantify how finite-\(n\) guarantees approach the continuum result, and test whether similar dissolutions occur for other pairwise rules.
The best objection is that the proposed atomless model makes Theorem 3.2’s incentive condition vacuous. Its proof relies on one player changing their report and thereby moving their own Borda position. In the proposed society, changing a single point’s type leaves every type mass \(\mu_t\), and thus every aggregate Borda score, unchanged. Sorting those scores with a public tie-break gives the exact target order and is impartial at the point level. Under the stated finite piecewise-rational encoding, this looks like a finite sorting task, not a new complexity question. The theorem’s impossibility has vanished because its individual-deviation test has vanished.
The obvious repair is to allow a coalition of positive mass to change reports and ask how much mass is needed to move its members’ positions. That is a plausible continuous research question, but the paper’s theorem does not establish it: a fixed mass models coalition manipulation, while shrinking the mass to \(1/n\) restores the finite electorate’s granularity and makes the answer sensitive to small score margins. I would treat that as a new coalition or robustness problem, not a mirror of Theorem 3.2.
So this anchor makes a weak case for a worthwhile ChoCo mirror: its clean atomless version is easy precisely where the paper’s strategic difficulty matters. I cannot honestly rule out a useful mass-manipulation extension; that is where the negative case is weakest.
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.