| paper | d4-journals-geb-2008-0077 |
| authors | — |
| venue | ? |
| filed under | unclassified |
| judged by | gpt-6-luna / xhigh (triple__luna__xhigh__c6r1) |
| judge confidence | high |
| authors would recognise it | yes |
The paper’s named results characterize diffusion thresholds and stability, but none states an algorithmic or computational result. Minimizing initial seed mass is motivated by the paper’s own question, yet it would be a new optimization problem rather than a computational result of this paper.
fails bit a — no named computational result to mirror
The proposed seed question draws on the positive- and unique-diffusion analyses, while leaving the comparative network-threshold results, continuity and hysteresis analyses, and finite-network simulations unmirrored.
The strongest positive case is a continuous campaign-control problem around a question the paper explicitly raises: how many initial adopters are needed to make diffusion persist? In a product launch, the agents could be customers in a large market grouped by degree and by shared response and campaign-cost parameters. Type \(t\) has degree \(k_t\); its mass \(\mu_t\) is the fraction of the population in that cohort. The launch decision is a seed mass \(x_t\in[0,\mu_t]\), and the cost is \(\sum_t c_t x_t\). Starting from \(\rho_t(0)=x_t/\mu_t\), the population follows the paper’s mean-field dynamics, with adopter mass summarized by \(\theta=\sum_t k_t\mu_t\rho_t/\sum_t k_t\mu_t\).
I would call the resulting problem Minimum Seed Cost for Persistence: given finite-support rational type masses, degrees, response functions, campaign costs, and spreading rate, find the least-cost seed profile whose mean-field trajectory enters the basin of a locally asymptotically stable state with positive adopter mass. This is a credible analogue of the paper’s “how many initial adopters” question: the action is an intervention on population mass, and the objective is campaign cost. A large customer base with relatively few degree and response cohorts is a plausible high-multiplicity regime. The authors would likely recognize the question, though type-targeted seeding extends their homogeneous initial-shock analysis. A solution would give an \(\varepsilon\)-optimal seed profile and the positive stable state it reaches.
The closest named result is Theorem 5, proved in this paper. It says that for connectivity distribution \(P(k)\) and diffusion function \(f(k,a)\) satisfying (A-1), positive diffusion occurs if and only if \(\lambda>\lambda_p\), where
\[
\lambda_p=\frac{\langle k\rangle}{\sum_k k^2P(k)f(k,1)}.
\]
Proposition 8, also proved here, says that if \(f(k,a)\) additionally satisfies the concavity condition (A-2), then unique diffusion occurs if and only if \(\lambda>\lambda_u\), with \(\lambda_u=\lambda_p\); otherwise the dynamics converge to the state with no active consumers. These results support the proposed campaign question by showing that the basin of persistent diffusion depends on the spreading rate and type composition. But neither is a computational-complexity result: the paper contains no named result classifying a problem as polynomial-time, NP-hard, or similar. So neither can honestly serve as the required computational anchor.
My tentative expectation for Minimum Seed Cost for Persistence is Class A, if the basin boundary can be characterized through the scalar mean-field statistic \(\theta\); then the boundary calculation and the choice of which types to seed may reduce to root-finding plus a linear optimization problem. The paper does not prove that reduction. If it fails, the geometry of the basin could make the problem continuum-specifically hard. Further questions include whether type-targeted seeding beats uniform seeding, and whether the mean-field optimum predicts success in the finite stochastic network.
The weakest point is also the key boundary of the case: the paper already takes a mean-field limit and studies densities, so simply restating its diffusion threshold would collide with existing continuous-population analysis. The defensible positive case is the new computational intervention problem motivated by that analysis. It is plausible and concrete, but it is not anchored by a named computational result of this paper, as the requested standard requires.
The central problem is that there is no computational result here to mirror. Theorem 5 gives a diffusion threshold, and Proposition 8 gives a uniqueness condition, but both are analytic claims about the mean-field dynamics. The paper states no computational problem and proves no complexity or algorithmic result. That matters especially for this paper: it already passes to population densities \(\rho_k\) and a degree distribution \(P(k)\). Its main object is a continuum description of aggregate behavior, not a finite-voter problem awaiting continuization.
The proposed *Minimum Seed Cost for Persistence* is the strongest follow-on. The paper explicitly asks how many initial adopters are needed, so a minimum-seed question is recognizable and deserves more credit than a wholly imported objective. But adding type-specific campaign prices and optimizing a seed profile creates a new intervention problem. The paper neither models those prices nor gives a computational result about choosing interventions. Removing the prices and asking for the minimum seed mass brings the question closer to the paper, but then it asks for a basin boundary in a model whose state variables and dynamics are already continuous. That could be a useful study of the mean-field model; it is not a high-multiplicity mirror of one of this paper’s computational results.
A better attempt to make the seed question algorithmic would specify a finite input: rational masses for a finite set of degree and response types, an encoding of their response functions, and a precise seed-cost objective. That would repair the representation problem—the paper allows general \(P(k)\) and \(f(k,a)\), without a finite encoding for them. Yet this repair underscores the gap: the resulting optimization task is a new formalization, not a computational problem the paper formulated or classified. Theorem 5 and Proposition 8 can motivate it, but they do not supply computational anchors.
I would not claim that no useful paper could be built around targeted seeding. The explicit “how many initial adopters?” question gives that idea a genuine foothold, and a finite-cohort formulation might be worth studying on its own. The negative case is weaker if “mirror” includes any new computational intervention problem suggested by the paper. Under the programme’s narrower standard—continuizing a population to chart the complexity of a computational problem in the paper—these anchors do not establish such a mirror: the paper’s results are analytic, and its population is already represented through mean-field densities.
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.