pith. sign in

arxiv: 2604.25258 · v1 · submitted 2026-04-28 · 🧮 math.OC

Regulation of Rumor Propagation via (Multi-Leader) Stackelberg Graphon Games

Huaning Liu , Gokce Dayanikli This is my paper

Pith reviewed 2026-05-07 16:01 UTC · model grok-4.3

classification 🧮 math.OC
keywords rumor propagationStackelberg gamesgraphon gamesmean-field gamesopinion dynamicsoptimal controlmulti-leader games
0
0 comments X

The pith

A leader can steer rumor spread across large networks by setting incentives in a Stackelberg graphon game that anticipates followers' equilibrium responses.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper models rumor control in very large populations by treating agents as players in a graphon game whose connections are described by a continuous limit rather than an explicit graph. A principal sets incentives to favor preferred news and discourage others, then the followers settle into a Nash equilibrium that the principal anticipates. This Stackelberg graphon game equilibrium is characterized by a system of forward-backward differential equations whose solutions are shown to exist. The authors extend the setup to two competing principals and compute examples showing that rivalry produces sharply divided opinion groups instead of consensus. Readers interested in public opinion or misinformation would see a concrete mathematical route for designing incentives that work at scale.

Core claim

Rumor propagation in large networked populations is regulated by Stackelberg graphon games in which a principal incentivizes the spread of preferred news. The graphon game Nash equilibrium of the follower population is characterized by a forward-backward differential equation system for which existence is established. In the multi-leader extension with two competing principals, numerical computation of the equilibria produces strong opinion divisions across the population.

What carries the argument

The Stackelberg graphon game equilibrium (SGGE), in which the leader chooses incentives first and the followers reach a graphon game Nash equilibrium (GGNE) characterized by a forward-backward differential equation system.

If this is right

  • The followers' Nash equilibrium in the graphon limit is given by solutions to a forward-backward system of differential equations.
  • Existence of equilibria holds for both the single-principal and two-principal models.
  • When two principals compete by pushing opposing news, the population splits into strongly opposed opinion clusters.
  • A bi-level numerical algorithm computes the equilibria for given incentive choices.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same incentive-design approach could be tested on real social-media data to see whether targeted rewards or penalties can measurably slow specific rumors.
  • Introducing deliberate competition between narratives might serve as a deliberate fragmentation strategy, though it trades consensus for polarization.
  • The model could be extended by replacing the deterministic mean-field limit with a stochastic differential equation to check how noise affects the existence of divisions.

Load-bearing premise

The population is large enough that its connections are well approximated by a continuous graphon and the rumor dynamics follow a deterministic mean-field limit without significant random fluctuations or external shocks.

What would settle it

Solve the forward-backward system or run the bi-level algorithm for a concrete finite network of several thousand nodes with two competing principals and check whether the resulting opinion distribution shows the predicted sharp divisions or whether no equilibrium exists.

Figures

Figures reproduced from arXiv: 2604.25258 by Gokce Dayanikli, Huaning Liu.

Figure 1
Figure 1. Figure 1: State Transition Flow for SKIR model Game setup. Let T > 0 be a finite time horizon. We denote the set of R ⊃ A-valued admissible controls by A. 1 In the limiting case where the number of agents goes to infinity, we focus on a continuum of agents that are indexed by x ∈ I := [0, 1]. Consider agent x ∈ I, her control process is denoted by α x := (α x t )t∈[0,T] and represents her communication level while h… view at source ↗
Figure 2
Figure 2. Figure 2: Sampled agents’ densities and aggregates on the power-law graph. The color of the lines get darker with higher agent index. 0 20 40 60 80 0.0 0.2 0.4 0.6 0.8 Uninformed(S) State density high-degree agent (no policy) low-degree agent (no policy) high-degree agent (policy) low-degree agent (policy) 0 20 40 60 80 0.0 0.1 0.2 0.3 0.4 0.5 0.6 Preferred(K) State density high-degree agent (no policy) low-degree a… view at source ↗
Figure 3
Figure 3. Figure 3: Socially (most) active/inactive agent under policy/no policy. be spreading preferred news (K), and 3% spreading non￾preferred news (I). In this experiment we set K = 4 to represent 18-29, 30-49, 50-64, 65+ age groups view at source ↗
Figure 4
Figure 4. Figure 4: Stackelberg Equilibrium Policy (0, solid lines) vs No￾Regulation (1, dashed lines) on age-groups. 0 20 40 60 80 0.0 0.2 0.4 0.6 0.8 Uninformed(S) State density Block 1, Policy 0 Block 1, Policy 1 Block 2, Policy 0 Block 2, Policy 1 Block 3, Policy 0 Block 3, Policy 1 Block 4, Policy 0 Block 4, Policy 1 0 20 40 60 80 0.0 0.1 0.2 0.3 0.4 Preferred(K) State density Block 1, Policy 0 Block 1, Policy 1 Block 2,… view at source ↗
Figure 5
Figure 5. Figure 5: Duo-Principal Stackelberg Equilibrium (0, solid lines) vs No-Regulation (1, dashed lines) on age-groups. using Stackelberg graphon games. We first introduce the graphon model of the individuals in the population under the policies of a single principal (i.e., regulator) who has preferred news. Agents in the population control their jumps between finite number of states that represent their knowl￾edge about… view at source ↗
read the original abstract

We study the control of rumor propagation in large networked populations by using Stackelberg graphon games. We first introduce a principal who wants to incentivize the spread of her preferred news and discourage the spread of non-preferred news. We define the Stackelberg graphon game equilibrium (SGGE), characterize the graphon game Nash equilibrium (GGNE) with a forward-backward differential equation system, and establish existence results. We further formulate a multi-leader model with two competing principals, each incentivizing her own preferred news. Finally, we propose a bi-level algorithm for computing (multi-leader) Stackelberg graphon game equilibria and conclude with numerical experiments where we show that existence of competing principals will result in strong opinion divisions in the population.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

3 major / 2 minor

Summary. The paper studies regulation of rumor propagation in large networks via Stackelberg graphon games. It defines the Stackelberg graphon game equilibrium (SGGE), characterizes the graphon game Nash equilibrium (GGNE) by a forward-backward differential equation system, proves existence, formulates a multi-leader model with two competing principals, proposes a bi-level algorithm, and presents numerical experiments concluding that competing principals produce strong opinion divisions.

Significance. If the results hold, the work supplies a mean-field game-theoretic framework for controlling information spread in infinite-population limits, with the forward-backward characterization and multi-leader extension offering new tools for social dynamics. The bi-level algorithm and experiments on opinion polarization provide concrete computational insights that could inform intervention design. Credit is due for the explicit equilibrium construction and the reproducible numerical pipeline.

major comments (3)
  1. [GGNE characterization section] The characterization of the GGNE via the forward-backward differential equation system (abstract and § on GGNE) requires explicit assumptions on the cost functions and rumor dynamics (e.g., convexity, Lipschitz continuity, boundedness) to support the existence proof; these are not stated, making it impossible to verify that the claimed equilibria are well-defined or that the numerical outputs satisfy the system.
  2. [Numerical experiments] Numerical experiments section: the bi-level algorithm outputs are not cross-validated against solutions of the forward-backward DE system, nor is any sensitivity analysis or finite-network comparison provided; this undermines the claim that competing principals produce strong opinion divisions, as the result may be an artifact of the deterministic graphon limit.
  3. [Multi-leader formulation] Multi-leader model: the extension to two competing principals does not specify the precise Stackelberg hierarchy or simultaneous-move equilibrium concept when leaders incentivize opposing news; without this, the reported opinion-division outcome cannot be rigorously attributed to the multi-leader structure.
minor comments (2)
  1. [Abstract] The abstract claims existence results without even a one-sentence pointer to the key theorem or assumptions; adding this would improve accessibility.
  2. [Introduction] Notation for SGGE versus GGNE should be introduced with a clear table or diagram in the introduction to avoid confusion when moving between single- and multi-leader cases.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the careful reading and constructive comments on our manuscript. We address each major comment point by point below and will revise the paper to incorporate clarifications and additional validation where needed.

read point-by-point responses

  1. Referee: [GGNE characterization section] The characterization of the GGNE via the forward-backward differential equation system (abstract and § on GGNE) requires explicit assumptions on the cost functions and rumor dynamics (e.g., convexity, Lipschitz continuity, boundedness) to support the existence proof; these are not stated, making it impossible to verify that the claimed equilibria are well-defined or that the numerical outputs satisfy the system.

    Authors: We agree that the assumptions supporting the forward-backward characterization and existence result should be stated explicitly. In the revised manuscript we will insert a new paragraph in the GGNE section listing the required conditions (strict convexity and continuous differentiability of the running and terminal costs, uniform Lipschitz continuity and linear growth of the rumor dynamics in the state and control variables, and boundedness of admissible controls). We will also verify that the specific quadratic costs and linear dynamics used in the paper satisfy these hypotheses, thereby confirming that the equilibria are well-defined and that the numerical outputs are consistent with the system. revision: yes

  2. Referee: [Numerical experiments] Numerical experiments section: the bi-level algorithm outputs are not cross-validated against solutions of the forward-backward DE system, nor is any sensitivity analysis or finite-network comparison provided; this undermines the claim that competing principals produce strong opinion divisions, as the result may be an artifact of the deterministic graphon limit.

    Authors: The referee is correct that direct validation is missing. Although the bi-level algorithm is constructed from the equilibrium conditions, we did not compare its outputs to independently solved forward-backward trajectories or perform sensitivity checks. In the revision we will add (i) numerical solutions of the forward-backward system for the same parameter values and a quantitative comparison of the resulting opinion profiles, (ii) a sensitivity analysis with respect to the strength of the competing incentives, and (iii) a brief finite-N simulation on a large random graph to illustrate convergence to the graphon limit. These additions will substantiate that the observed strong opinion divisions arise from the multi-leader structure rather than from algorithmic artifacts. revision: yes

  3. Referee: [Multi-leader formulation] Multi-leader model: the extension to two competing principals does not specify the precise Stackelberg hierarchy or simultaneous-move equilibrium concept when leaders incentivize opposing news; without this, the reported opinion-division outcome cannot be rigorously attributed to the multi-leader structure.

    Authors: We acknowledge that the multi-leader section would benefit from an explicit equilibrium definition. The model treats the two principals as simultaneous-move leaders who each choose an incentive function, after which the continuum of followers plays a Nash equilibrium given the aggregate incentive. In the revision we will add a formal definition of the multi-leader Stackelberg graphon game equilibrium (with simultaneous leader actions and a follower Nash response), state the associated optimality conditions, and explain how opposing incentives produce the polarization observed in the simulations. This will make the attribution to the multi-leader structure precise. revision: yes

Circularity Check

0 steps flagged

No significant circularity in equilibrium definitions or characterizations.

full rationale

The paper defines the Stackelberg graphon game equilibrium (SGGE) as a new construct and derives the characterization of the graphon game Nash equilibrium (GGNE) via a forward-backward differential equation system from standard mean-field game theory constructions, followed by existence proofs and a multi-leader extension. No key result reduces by construction to a fitted parameter, self-referential definition, or load-bearing self-citation chain; the bi-level algorithm and numerical illustrations are downstream applications of the independently derived model rather than tautological predictions. The derivation chain is self-contained against external game-theoretic benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 2 invented entities

The central claims rest on standard existence theorems for mean-field games and the graphon limit approximation; no free parameters or data-fitted quantities are mentioned in the abstract.

axioms (2)
  • standard math Existence of Nash equilibria for the graphon game under the stated cost structure
    Invoked to characterize GGNE via the forward-backward differential equation system.
  • domain assumption The finite network converges to a graphon as population size tends to infinity
    Required for the continuous-limit model to represent large networked populations.
invented entities (2)
  • Stackelberg graphon game equilibrium (SGGE) no independent evidence
    purpose: Equilibrium concept capturing the principal's incentive design and population response
    Newly defined in the paper to formalize the leader-follower rumor control problem.
  • Multi-leader Stackelberg graphon game no independent evidence
    purpose: Extension to competing principals each promoting their own news
    Formulated as a new model variant with associated bi-level algorithm.

pith-pipeline@v0.9.0 · 5425 in / 1682 out tokens · 54117 ms · 2026-05-07T16:01:06.669355+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

40 extracted references

  1. [1]

    Carmona and F

    R. Carmona and F. Delarue, Probabilistic Theory of Mean Field Games with Applications II Mean Field Games with Common Noise and Master Equations , pp. 1–679. Probability Theory and Stochastic Modelling, United States: Springer Nature, 2018. Age β S β K β I µ K µ I η 18-29 0.4 0.5 0.75 0.1 0.1 0 30-49 0.3 0.42 0.62 0.05 0.05 0 50-64 0.3 0.32 0.48 0.05 0.05...

    2018

  2. [2]

    Mean field games,

    J.-M. Lasry and P .-L. Lions, “Mean field games,” Japanese Journal of Mathematics , vol. 2, no. 1, pp. 229–260, 2007

    2007

  3. [3]

    Mean field game model for an advertising competition in a duopoly,

    R. Carmona and G. Dayanıklı, “Mean field game model for an advertising competition in a duopoly,” International Game Theory Review, vol. 23, no. 04, p. 2150024, 2021

    2021

  4. [4]

    Dynam ic mar- keting policies with rating-sensitive consumers: A mean-fi eld games approach,

    R. Salhab, J. Le Ny, R. P . Malham´ e, and G. Zaccour, “Dynam ic mar- keting policies with rating-sensitive consumers: A mean-fi eld games approach,” European Journal of Operational Research, vol. 299, no. 3, pp. 1079–1093, 2022

    2022

  5. [5]

    Mean field games incorp orating carryover effects: optimizing advertising models,

    M. Ricciardi and M. Rosestolato, “Mean field games incorp orating carryover effects: optimizing advertising models,” Decisions in Eco- nomics and Finance , 2024

    2024

  6. [6]

    Fracking, renewables, and mean fie ld games,

    P . Chan and R. Sircar, “Fracking, renewables, and mean fie ld games,” SIAM Review , vol. 59, no. 3, pp. 588–615, 2017

    2017

  7. [7]

    Mean–field moral hazard for optimal energy demand response management ,

    R. ´Elie, E. Hubert, T. Mastrolia, and D. Possama¨ ı, “Mean–field moral hazard for optimal energy demand response management ,” Mathematical Finance, vol. 31, no. 1, pp. 399–473, 2021

    2021

  8. [8]

    An extended m ean field game for storage in smart grids,

    C. Alasseur, I. Ben Taher, and A. Matoussi, “An extended m ean field game for storage in smart grids,” Journal of Optimization Theory and Applications, vol. 184, pp. 644–670, 2020

    2020

  9. [9]

    Multi-population mean fi eld games with multiple major players: Application to carbon emissio n regula- tions,

    G. Dayanikli and M. Lauri` ere, “Multi-population mean fi eld games with multiple major players: Application to carbon emissio n regula- tions,” in 2024 American Control Conference , pp. 5075–5081, 2024

    2024

  10. [10]

    Mean field m odels to regulate carbon emissions in electricity production,

    R. Carmona, G. Dayanıklı, and M. Lauri` ere, “Mean field m odels to regulate carbon emissions in electricity production,” Dynamic Games and Applications , 2022

    2022

  11. [11]

    A stackelb erg mean field game for green regulator with a large number of prosumers,

    M. Bichuch, G. Dayanıklı, and M. Lauri` ere, “A stackelb erg mean field game for green regulator with a large number of prosumers,” i n 2025 American Control Conference (ACC) , pp. 4129–4134, 2025

    2025

  12. [12]

    Mean field games and sy stemic risk,

    R. Carmona, J. Fouque, and L. Sun, “Mean field games and sy stemic risk,” Communications in Mathematical Sciences , vol. 13, no. 4, pp. 911–933, 2015. Publisher Copyright: © 2015 Internation al Press

    2015

  13. [13]

    Graphon games: A statistica l framework for network games and interventions,

    F. Parise and A. Ozdaglar, “Graphon games: A statistica l framework for network games and interventions,” Econometrica, vol. 91, no. 1, pp. 191–225, 2023

    2023

  14. [14]

    Graphon mean field games and th eir equations,

    P . E. Caines and M. Huang, “Graphon mean field games and th eir equations,” SIAM Journal on Control and Optimization , vol. 59, no. 6, pp. 4373–4399, 2021

    2021

  15. [15]

    Stochastic graphon games: I. the static case,

    R. Carmona, D. B. Cooney, C. V . Graves, and M. Lauri` ere, “Stochastic graphon games: I. the static case,” Mathematics of Operations Re- search, vol. 47, no. 1, pp. 750–778, 2022

    2022

  16. [16]

    Stochastic gra phon games: Ii. the linear-quadratic case,

    A. Aurell, R. Carmona, and M. Lauriere, “Stochastic gra phon games: Ii. the linear-quadratic case,” Applied Mathematics & Optimization , vol. 85, no. 3, p. 39, 2022

    2022

  17. [17]

    Modeling of rumor propagation in large populations with network via graphon games,

    H. Liu and G. Dayanıklı, “Modeling of rumor propagation in large populations with network via graphon games,” in 2025 American Control Conference (ACC) , pp. 816–821, 2025

    2025

  18. [18]

    Finite state graphon games with applications to epidemics,

    A. Aurell, R. Carmona, G. Dayanıklı, and M. Lauri` ere, “ Finite state graphon games with applications to epidemics,” Dynamic Games and Applications, vol. 12, no. 1, pp. 49–81, 2022

    2022

  19. [19]

    Stochastic Rumours,

    D. J. DALEY and D. G. KENDALL, “Stochastic Rumours,” IMA Journal of Applied Mathematics , vol. 1, pp. 42–55, 03 1965

    1965

  20. [20]

    Mathematical models and app lications : with emphasis on the social, life, and management sciences ,

    D. P . Maki and M. Thompson, “Mathematical models and app lications : with emphasis on the social, life, and management sciences ,” 1973

    1973

  21. [21]

    Modeling and analysis of rumor propagation in social networks,

    Z. Y u, S. Lu, D. Wang, and Z. Li, “Modeling and analysis of rumor propagation in social networks,” Information Sciences, vol. 580, pp. 857–873, 2021

    2021

  22. [22]

    Rumor propagation dynamic model based on evolutionary game and an ti- rumor,

    Y . Xiao, D. Chen, S. Wei, Q. Li, H. Wang, and M. Xu, “Rumor propagation dynamic model based on evolutionary game and an ti- rumor,” Nonlinear Dynamics , vol. 95, 01 2019

    2019

  23. [23]

    Opinion formation models base d on game theory,

    A. DI MARE and V . LA TORA, “Opinion formation models base d on game theory,” International Journal of Modern Physics C , vol. 18, no. 09, pp. 1377–1395, 2007

    2007

  24. [24]

    Opinion dynamics i n social networks through mean-field games,

    D. Bauso, H. Tembine, and T. Bas ¸ar, “Opinion dynamics i n social networks through mean-field games,” SIAM Journal on Control and Optimization, vol. 54, no. 6, pp. 3225–3257, 2016

    2016

  25. [25]

    Laffont and D

    J.-J. Laffont and D. Martimort, The Theory of Incentives: The Principal-Agent Model . Princeton University Press, 2002

    2002

  26. [26]

    A theory of incentives in procurement an d regula- tion.,

    M. Armstrong, “A theory of incentives in procurement an d regula- tion.,” The Economic Journal , vol. 105, pp. 193–194, 01 1995

    1995

  27. [27]

    Mean field s tack- elberg games: Aggregation of delayed instructions,

    A. Bensoussan, M. H. M. Chau, and S. C. P . Y am, “Mean field s tack- elberg games: Aggregation of delayed instructions,” SIAM Journal on Control and Optimization , vol. 53, no. 4, pp. 2237–2266, 2015

    2015

  28. [28]

    Linear quadratic mean field stack elberg differential games,

    J. Moon and T. Bas ¸ar, “Linear quadratic mean field stack elberg differential games,” Automatica, vol. 97, pp. 200–213, 2018

    2018

  29. [29]

    Stackelberg mean field games: convergence and existence results to the problem of principal with multiple agents in c ompeti- tion,

    M. F. Djete, “Stackelberg mean field games: convergence and existence results to the problem of principal with multiple agents in c ompeti- tion,” 2023

    2023

  30. [30]

    A machine learning meth od for stackelberg mean field games,

    G. Dayanıklı and M. Lauri` ere, “A machine learning meth od for stackelberg mean field games,” Mathematics of Operations Research , vol. 0, no. 0, p. null, 0

  31. [31]

    Optimal incentives to mitigate epidemics: A stackelberg mean field g ame approach,

    A. Aurell, R. Carmona, G. Dayanikli, and M. Lauri` ere, “ Optimal incentives to mitigate epidemics: A stackelberg mean field g ame approach,” SIAM Journal on Control and Optimization , vol. 60, no. 2, pp. S294–S322, 2022

    2022

  32. [32]

    Belief and opinion evolution in social networks based on a multi-po pulation mean field game approach,

    R. A. Banez, H. Gao, L. Li, C. Y ang, Z. Han, and H. V . Poor, “ Belief and opinion evolution in social networks based on a multi-po pulation mean field game approach,” in ICC 2020 - 2020 IEEE International Conference on Communications (ICC) , pp. 1–6, 2020

    2020

  33. [33]

    Coronavirus: Government paid reality tv stars to prom ote nhs test and trace,

    “Coronavirus: Government paid reality tv stars to prom ote nhs test and trace,” Aug 2020

    2020

  34. [34]

    Colorado pays 126 influencers to advocate for covid vaccinations,

    S. Hindi, “Colorado pays 126 influencers to advocate for covid vaccinations,” Jul 2021

    2021

  35. [35]

    Borah, Facebook Use in the 2012 USA Presidential Campaign , pp

    P . Borah, Facebook Use in the 2012 USA Presidential Campaign , pp. 201–211. Cham: Springer International Publishing, 201 4

    2012

  36. [36]

    China passes sweeping recommendation algori thm regula- tions,

    A. Huld, “China passes sweeping recommendation algori thm regula- tions,” Mar 2022

    2022

  37. [37]

    Media capture and advertising in turkey: T he impact of the state on news,

    S. Y anatma, “Media capture and advertising in turkey: T he impact of the state on news,” tech. rep., Reuters Institute for the S tudy of Journalism, University of Oxford, Jul 2016

    2016

  38. [38]

    Brezis, Functional Analysis, Sobolev Spaces and Partial Differen- tial Equations

    H. Brezis, Functional Analysis, Sobolev Spaces and Partial Differen- tial Equations . Universitext, Springer New Y ork, 2010

    2010

  39. [39]

    A machine learning meth od for stackelberg mean field games,

    G. Dayanıklı and M. Lauri` ere, “A machine learning meth od for stackelberg mean field games,” Math. Oper . Res., Dec. 2024. APPENDIX A. Finite Player F ormulation To better motivate the study of the graphon game frame- work, we present the corresponding finite-player game on weighted network. Consider a population of N ∈ N agents index by [N ] := {1, 2,......

    2024

  40. [40]

    We are ready to introduce the finite-player Nash equilibrium on the networked system

    where we replace the aggregates with the finite-player versions in ( 13). We are ready to introduce the finite-player Nash equilibrium on the networked system. Definition 4: The control profile α N is an N -player Nash equilibrium if it is admissible and no player can gain from a unilateral deviation, i.e., J j,N ( α j; α − j ) ≤ J j,N ( σ ; α − j ) , ∀σ ∈ A,...