| paper | d4-conf-atal-2018-0212 |
| authors | — |
| venue | ? |
| filed under | unclassified |
| judged by | gpt-6-luna / xhigh (triple__luna__xhigh__c6r1) |
| judge confidence | medium |
| authors would recognise it | yes |
other→P discrete result → its continuous mirror
Answer: The stated continuous opinion-shaping decision problem is in $\mathrm{P}$. · discrete side: other · adjudication: proved
full proof & adversarial review →
Lemma 3.6
statement extracted from the paper’s text layer
Given a finite type set \(T\) with rational masses \(\mu_t\), a fixed set of control types \(I\), stable per-capita organic posting rates \(\lambda_s\), baseline opinions \(\alpha_t\), block influence coefficients \(W_{ts}\), costs \(c_s\), and budget \(C\), choose control-post rates \(h_s^+,h_s^-\ge 0\) for \(s\in I\) and steady-state means \(x_t\) for \(t\notin I\) to maximize \(-\sum_{t\notin I}\mu_t x_t\), subject to \(\sum_{s\in I}\mu_s c_s(h_s^++h_s^-)\le C\) and \(0=\omega(\alpha_t-x_t)+\sum_{s\notin I}W_{ts}\mu_s\lambda_s x_s+\sum_{s\in I}W_{ts}\mu_s(h_s^+-h_s^-)\) for each \(t\notin I\).
An exact block-structured population with finitely many types, type masses \(\mu_t\), and per-capita influence and posting parameters. The decision variables are type-level control rates and uncontrolled-type mean opinions; the objective is the mass-weighted average opinion under a mass-weighted budget.
The paper does not establish that its individual-node Hawkes dynamics aggregate exactly under type grouping and per-capita scaling; this confines the mirror to exact block-structured regimes, which the proponent only partly specifies.
fatal: False
Covers Lemma 3.6’s steady-state mean computation and the fixed-controller rate-selection task for the AOSH-1 objective. It leaves joint selection of individual control users, covariance-robust shaping, other objectives, and the empirical results outside the mirror.
I don’t see a qualifying named computational result in this paper, so I can’t make an anchor-compliant case under your rules. The numbered results—Proposition 3.1, Theorems 3.2–3.3, and Lemmas 3.4–3.8—establish stochastic dynamics, stability, and steady-state formulas. The SmartShape optimization programs and the claims that they are convex and efficiently solvable appear in the numbered equations and prose, not in a numbered theorem or proposition asserting a computational result.
The closest result is Lemma 3.6, proved in the paper: if the mean dynamics are stable, the non-control users’ steady-state expected opinions satisfy \((A_1\Lambda_{I_c}-\omega I)x_{I_c}+A_2(\eta_I^+-\eta_I^-)+\omega\alpha_{I_c}=0\). The paper then uses this steady-state relation in its rate-optimization program for a preselected set of control users. That makes it a useful computational foundation, but it is a formula for the dynamics, not a named complexity or tractability result. I would not present it as an eligible anchor.
The strongest *unanchored* positive candidate is the preselected-controller rate problem. Take a large online population divided into finitely many behavioral and network-role types \(T\), with rational masses \(p_t\) summing to one. A type records the baseline opinion, organic posting rate, influence role, and posting cost; \(W_{ts}\) gives the influence on type \(t\) from a unit of type-\(s\) mass. Fix a set \(I\) of control types. Choose per-person positive and negative control-post rates \(h_s^+,h_s^-\ge 0\) for \(s\in I\), subject to \(\sum_{s\in I}p_sc_s(h_s^++h_s^-)\le C\). For each uncontrolled type \(t\notin I\), its steady-state mean opinion \(x_t\) satisfies \(\omega(x_t-\alpha_t)=\sum_{s\notin I}W_{ts}p_s\lambda_sx_s+\sum_{s\in I}W_{ts}p_s(h_s^+-h_s^-)\), where \(\lambda_s\) is the stable organic posting rate per person. Ask for rates and steady-state opinions maximizing \(-\sum_{t\notin I}p_tx_t\). With stable dynamics and rational input data, this is a linear optimization problem. It mirrors the paper’s preselected-control rate task and its AOSH-1 objective; the expected direction is Class A.
A plausible regime is a platform-scale campaign with millions of users but comparatively few stable communities or behavioral classes, each with similar posting and influence parameters. The network must admit a meaningful type-to-type mixing description; an arbitrary graph whose every user has a distinct influence profile would not give a small support. The authors would likely recognize the typed problem as their Hawkes-based opinion-shaping task, though this mirror covers rate choice for fixed control types—not SmartShape’s joint selection of individual controllers or its covariance-robust variant. That network aggregation is the candidate’s weakest point.
So there is a credible continuous-population question here, and Lemma 3.6 helps make it computationally concrete. But under the required anchor test, it remains an unanchored candidate, not a complete affirmative case.
The affirmative case has no eligible anchor. Lemma 3.6 gives a steady-state equation for expected opinions; it does not assert the complexity or tractability of a computational problem. The other numbered results concern dynamics, stability, and covariance. The paper’s efficiency claims for SmartShape appear in its formulation and discussion, not as a numbered complexity result. So, under the requested anchor test, nothing has yet been identified that a continuous mirror must answer.
The proposed rate-selection mirror does not make Lemma 3.6 computational. In the paper, the preselected-controller problem already chooses real-valued rates on a fixed, individually indexed graph. To replace users by finitely many masses while preserving that problem, users of each type must have the same relevant parameters and the same total influence into every type. On a sparse social graph, sharing a behavioral label does not usually give users the same influence profile. If type records the full network role needed for exact closure, the support may grow with the number of users. If \(W_{ts}\) instead records average type-to-type influence, it substitutes a mixing model for the paper’s user-level network.
The best repair is an exchangeable-community model: finitely many types with block-level influence, per-capita scaling to preserve stability as the population grows, a per-capita budget, and a controllable fraction of each type. Mean objectives could be population-weighted, and worst-user objectives could use an appropriate mass-level analogue. That is a coherent high-multiplicity opinion-control question, and the authors might recognize it as a version of their task. But it requires new assumptions and scaling in the dynamics; the paper’s stability and covariance formulas do not by themselves establish the resulting model or its computational properties. In an exactly symmetric network, the quotient largely reduces the existing node equations to type equations. With average mixing in a heterogeneous network, it is a new approximate model whose fidelity would itself need study. Neither case supplies a computational result from this paper to mirror.
The fidelity bottleneck is the strongest objection; the absence of a named computational anchor is decisive under the stated rules. I would not claim, though, that no worthwhile continuous research question exists in *any* scenario. The block-community variant remains credible if the programme allows unanchored extensions. The narrower conclusion is that this paper, and the proponent’s Lemma 3.6 anchor in particular, do not establish a worthwhile continuous mirror of one of its named computational results.
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.