| paper | Interactive Democracy |
| authors | Markus Brill |
| venue | ATAL 2018 |
| filed under | voting · incomplete-info |
| judged by | gpt-5.6-luna / xhigh (triple__luna__xhigh__c6r1) |
| judge confidence | high |
| authors would recognise it | no |
The paper contains no numbered computational result of its own: Section 2.1 poses an open question, while the cited results belong to other works. The proposed continuous ranked-delegation problem is also not a faithful population relaxation of the paper’s identity-sensitive voting-power object, because collapsing named delegates into role-types changes the graph, cycle structure, and quantity computed. Thus the paper fails the computational-result gate.
fails bit a — no named computational result to mirror
The role-based model loses the named-delegate identities that determine delegation cycles and individual voting power, so it studies anonymous aggregate support rather than the paper’s object.
fatal: True
The candidate covers only Section 2.1’s open aggregate-voting-power question; it leaves the paper’s discussions of inconsistent outcomes, strategic manipulation, proportional ranking, clone independence, and preference elicitation untouched.
The strongest honest answer is that this paper has no eligible anchor under your rules. It contains no numbered Theorem, Lemma, Corollary, or Proposition, and proves no named result of the form NP-hardness, polynomial-time solvability, FPT, or similar. The Gibbard–Satterthwaite theorem is mentioned, but it is neither numbered here nor a computational-complexity result. Results cited from [42], [62], and elsewhere are external results, not results of this paper. Thus, strictly speaking, there is no valid anchor and no compliant “one continuous problem per anchor” deliverable. Claiming otherwise would overstate the paper.
The best prospective mirror is nevertheless clear. Section 2.1 explicitly asks whether one can compute the aggregate distribution of voting power in ranked delegation without computing the delegated vote of every individual. I would call the continuous version Continuous Ranked-Delegation Evaluation.
Let \(C\) be the alternatives and \(T\) a finite set of complete voter types. A type \(t\) contains its status—direct voter for \(c\in C\), abstainer, or delegator—together with its ranked list of delegate types. Let \(\mu\in\mathbb{Q}^{T}\) be the population distribution, so \(\mu_t\) is the fraction of society of type \(t\). For a delegating type with list \((u_{t,1},\ldots,u_{t,\ell})\), the delegation walk chooses \(u_{t,j}\) with probability proportional to \(\varepsilon^{j-1}\), with the paper’s restart rule for abstentions. The restart convention is part of the instance, and instances are required to have well-defined eventual resolution.
For every type \(t\), let \(q_{t,c}\) be the probability that its delegation eventually resolves to a direct voter voting for \(c\). The continuous aggregate voting power is
\[ v_\mu(c)=\sum_{t\in T}\mu_t q_{t,c}. \]
The problem is to compute \(v_\mu\) exactly and return the winner set
\[ W(\mu)=\arg\max_{c\in C} v_\mu(c), \]
or, equivalently, decide whether a designated \(c^\star\) belongs to \(W(\mu)\). The action/decision variable is therefore the induced aggregate vote distribution, not an artificial fractional ballot. Abstention remains abstention; \(\mu\) is population mass, not merely randomness over outcomes.
The intended regime is a large online-democracy platform with \(N\gg |T|=\tau\) participants distributed among a small number of recurring role-types: for example, constituents, issue specialists, local representatives, and abstainers, each with the same ranked delegation behaviour and direct voting behaviour. The finite election with \(N_t\) agents of type \(t\) corresponds to \(\mu_t=N_t/N\). This is a genuine high-multiplicity version of the paper’s problem if delegate targets are role-types or interchangeable representative pools; it is not simply continuity in the Markov chain or in the outcome space.
The authors should recognise this as their question: it preserves ranked delegation, delegation cycles, abstentions, fallback choices, and the request for an aggregate power computation. Its likely classification is Class A for fixed rational \(\varepsilon\): construct the type-level transition matrix and solve the corresponding absorption linear system in polynomial time in \(\tau\), \(|C|\), and the encoding length. The paper itself does not prove this; it is a proposed continuization and an inference. Further questions include the \(\varepsilon\to0\) limit, minimum-mass delegation edits needed to make \(c^\star\) win, and robustness of \(W(\mu)\) under perturbations of \(\mu\).
The weakest point is substantial: arbitrary liquid-democracy graphs are relational and identity-sensitive. Replacing named voters by interchangeable types can erase correlations among individual delegation edges and therefore change which cycles exist. The mirror is strongest only for a legitimate role-based high-multiplicity regime in which agents of the same type really are interchangeable. That gives a plausible continuous analogue of Section 2.1, but it cannot serve as a paper-anchored positive case in the strict grading scheme because the article contains no qualifying named computational result.
The negative case is unusually strong at the gatekeeping level: this paper contains no eligible named computational result. It is a blue-sky position paper. It proves no theorem, lemma, corollary, hardness result, tractability result, approximation guarantee, or parameterized algorithm. The Gibbard–Satterthwaite theorem and the results on proportional rankings and liquid democracy are cited external work, not results of this paper. Section 2.1 merely poses an open implementation question. Thus there is no paper result to which a continuous mirror can be anchored.
The proposed ranked-delegation mirror also fails substantively. The paper’s object is the distribution of delegated voting power over individual non-abstaining voters. Delegation lists point to named voters, and cycles and fallback behaviour depend on the exact directed graph. Two voters with the same preferences and status but different delegate targets are not the same type for this problem: their identities are part of the relevant description. If types retain those target identities, the number of types tracks the graph and the supposed high-multiplicity relaxation disappears.
The strongest rescue is a role-based platform with many constituents delegating to a small set of interchangeable representative types. But that changes the object. A type-level transition matrix can compute aggregate candidate support,
\[ v_\mu(c)=\sum_t \mu_t q_{t,c}, \]
by ordinary finite-state Markov-chain algebra. It cannot recover the distribution of voting power among the individual representatives, which is precisely the paper’s concern. If one instead aggregates representative power by role, the concentration, identity, and cycle structure that motivate the question have been discarded. This is a new anonymous delegation model, not the continuous version of the paper’s individual delegation graph.
One could preserve more structure with a continuum of agents and a measurable delegation kernel or graphon. That creates a trilemma: with a finite block representation it reduces to the same finite type-level chain; with enough structure to preserve individual graph effects, the representation scales with the original agents and edges; with an unrestricted kernel, there is no finite encoded input on which an exact computational-complexity question is defined. Adding mass-edit, winner-robustness, or delegation-optimization questions would produce potentially interesting ChoCo problems, but they would be newly invented problems, not mirrors of a computational result in this article.
The honest qualification is that the role-based model is not nonsensical. It could be a worthwhile independent study of anonymous liquid democracy. It simply cannot rescue this paper under the programme’s anchor requirement, and it does not preserve the paper’s identity-sensitive voting-power 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.