Bounded confidence dynamics generates opinion cascades on growing scale-free networks

David Hernandez; Guillaume Deffuant; Yerali Gandica

REVIEW 4 major objections 5 minor 19 references

Bounded confidence dynamics generates opinion cascades on growing scale-free networks

T0 review · 4 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read In growing social networks, bridge agents are what make minority opinions suddenly merge with the majority.

desk verdict New simulation result — cascades on growing scale-free networks with homophily — looks real, but the theoretical fit in §4 rests on an unstated cluster-degree assumption and a same-data consistency check. read the letter →

arxiv 2506.02669 v3 pith:CMF2T3EP submitted 2025-06-03 physics.soc-ph

David Hernandez , Guillaume Deffuant , Yerali Gandica This is my paper

classification physics.soc-ph

keywords boundedconfidenceopiniondynamicsscale-freenetworkscascadeshomophilybridgeagentsgrowing

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

Bounded-confidence opinion dynamics on a growing, scale-free network with homophily can produce 'opinion cascades': small opinion clusters that form on the periphery of opinion space, stay stable for a time, and then abruptly merge with the dominant cluster. The paper establishes that this merging is triggered by newly arriving agents that, by chance, link to both clusters at once; these bridge agents then pull the clusters' opinions together through repeated interactions. The authors derive an approximate formula for the probability that an arriving agent is a bridge, and show that its time evolution explains why merging is fast at moderate homophily and slow at strong homophily. The result matters because it gives a mechanistic explanation for sudden opinion shifts and for the absorption of minority groups into majorities, without relying on contagion or epidemic spreading.

What carries the argument

The central object is the link-choice probability φ_N(i) = k_i exp(−β|θ_N − θ_i|) / Σ_j k_j exp(−β|θ_N − θ_j|), which sets the chance that an arriving agent connects to each pre-existing node as a trade-off between degree k_i and opinion proximity. Using the assumption that each cluster's total degree is 2 m n_i, the paper reduces this to a two-cluster bridge probability P_2(n_1,n_2) = 1/(1 + (n_1/n_2) exp(β(θ_1+θ_2−2x))), then to P_b = 1 − P_{1,m} − P_{2,m}, the probability that a new agent of opinion x links to both the major and the minor cluster. This bridge probability is what explains the cascade: its growth in time, combined with the small number of links m, determines how fast the minor cluster's average opinion is pulled toward the major cluster.

What would settle it

Run the two-cluster experiment with a specified internal topology (e.g., fully connected clusters or Erdős–Rényi with average degree m) and measure the time until the minor cluster's average opinion crosses halfway to the major cluster; if this merging time differs substantially from the one predicted by the 2 m n_i degree assumption, the bridge-probability explanation is falsified. A simpler check is to compute the empirical average degree inside each cluster just before cascade events and see whether it is close to m.

Watch

Extended reading notes

Core claim

The paper claims that in the parameter range where both node degree and opinion proximity influence link formation (intermediate values of the homophily parameter β), the pairwise bounded confidence model on a growing Barabási–Albert-like network generates opinion cascades. Minor opinion clusters continually appear at distances greater than the confidence bound ε from a major cluster, remain stable for some time, and then suddenly merge with the major cluster when an arriving agent links to both clusters. Using a simplified two-cluster setup, the authors derive the probability P_b that a new agent becomes such a bridge, and they validate the qualitative form of this probability in simulations with 10,000 replicas. The comparison shows that for β=15 the bridge probability grows rapidly as clusters approach, making the number of bridges soon exceed the number of agents attached only to the minor cluster, which drives the fast merging observed at β=15; for β=30 the bridge probability stays low, so the minor cluster is slowly absorbed instead. The paper also notes that the cascade mechanism requires the number of new links per agent m to be small, since a large m lets bridge agents interact almost exclusively with the major cluster.

Load-bearing premise

The predicted cascade speed rests on assuming each cluster has total degree 2 m n_i (every agent holding m connections on average) and that all agents in a cluster share one opinion, even though the simulations let clusters develop internal opinion spread and their internal connection structure is never specified.

Editorial extensions

If this is right

For intermediate homophily strengths (β roughly 15–20) the model predicts that minor opinion groups with a confidence bound equal to that of the majority will be absorbed suddenly, whereas for stronger homophily (β > 30) the absorption is slow and gradual.
The mechanism requires a small number of links per newcomer m; when m is large, bridge agents interact mostly with the large cluster and minor clusters stay stable, matching the earlier fully-connected population result.
The theoretical bridge probability, although derived under strong simplifications, correctly reproduces the qualitative difference in merging speed between β=15 and β=30, giving a testable handle on when opinion cascades should occur.
If these cascades are real features of opinion dynamics, then polls or social-media signals should show stable peripheral opinions followed by sudden jumps toward the mainstream opinion, rather than smooth convergence.
The paper's conclusion suggests that minority groups can survive only by lowering their confidence bound, which the authors interpret as radicalisation.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

We extend the paper's bridge mechanism to conjecture that cascade speed is controlled by the ratio n_1/n_2 as well as β, and that the critical bridge probability at which runaway merging begins could be estimated analytically; this is not tested in the paper.
A testable extension: in a version of the model where arriving agents are drawn from a non-uniform opinion distribution (e.g., heavy-tailed), the cascade frequency should shift toward the peak of the arrival distribution, since bridge agents are more likely where arrivals concentrate.
The persistence of Louvain communities in the merged cluster suggests that even after the opinions have converged, the network's latent community structure may preserve a 'memory' of the pre-cascade division; this is our reading of the simulation result, not a claim the paper tests.
The mechanism may generalize to other network-growing rules: any link-formation process that mixes popularity with similarity could produce cascades analogous to these, since the only ingredient needed is the existence of agents that connect two otherwise isolated opinion basins.

Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

The genuinely new thing here is the observation that on growing scale-free networks with bounded-confidence interactions and homophily-based link formation, minor opinion clusters can sit on the periphery for a while and then suddenly merge with a major cluster. That is a clean, reproducible-in-principle simulation result, and it is distinct from the authors' earlier growing-fully-connected work. The paper is well written and the model is described clearly enough to re-implement.

What it does well: the bridge-agent mechanism is plausible, and the comparison between β=15 and β=30 in Figs. 3–5 gives a coherent qualitative explanation for why merging is fast in one case and slow in the other. The paper is honest about the approximations in the theory.

Soft spots: the main one is Eqs. (3)–(4). The paper claims that for a cluster assembled by the same growth process, total oriented degree is 2m n_i. But the initialized clusters in §4.2 are not described as BA-grown; they are just 'connected only to agents of this cluster.' If they are complete graphs, the degree sum is n_i(n_i−1), which changes the bridge probability by a large factor. The fact that the theoretical curves fit 'quite very accurately' at t=0 then becomes less impressive: it could be matching the simulator's hidden construction rather than validating a derivation. Second, the validation is a consistency check: theoretical curves are computed using simulated cluster positions and sizes, then compared to simulated frequencies. That is not an independent prediction. Third, no error bars on the main simulation figures and no systematic parameter sweep; we don't know how robust the cascade window (β ~ 15–20) is. Minor: no code or data released.

Proportion: the central observation — cascades happen — does not depend on the theory. I believe it. The theory is an attempt to explain the speed difference, and the qualitative story (bridges become more probable as clusters approach) is sound even if the quantitative fit is overclaimed. The degree-sum issue is a real gap in the derivation as written, but it is fixable by specifying cluster initialization or deriving the degree sum for whatever initialization is used.

Who this is for: people working on opinion dynamics, network growth, and minority influence. It is a within-subfield contribution, not a paradigm shift.

Recommendation: I would send it to peer review, with the expectation that the authors must clarify the cluster construction, show error bars, and either provide an independent prediction or soften the claim that the theory is validated. The cascade phenomenon itself deserves to be in the literature.

Referee Report

4 major / 5 minor

Summary. The manuscript studies a growing scale-free network model in which new agents attach with probability proportional to degree times an exponential homophily term in opinion distance, while connected agents interact via the bounded-confidence rule. It reports that for intermediate values of the homophily parameter beta, minority opinion clusters form at the periphery of opinion space, persist for a time, and then abruptly merge with the majority cluster, a process the authors call 'opinion cascades.' To explain this, the authors derive an approximate probability that a newly arriving agent connects to both the major and minor clusters (a 'bridge' agent) and validate this formula against simulations initialized with one large and one small cluster. They argue that the faster growth of bridge agents for lower beta triggers the cascade, while for higher beta the bridge population grows too slowly and the merging is smoother.

Significance. If the reported phenomenon is robust, the paper is a useful contribution: it identifies a plausible mechanism (bridge agents) linking network growth, homophily, and bounded-confidence dynamics, with a qualitative prediction about merging speed as a function of beta. The manuscript has clear strengths: a precise model specification with pseudo-code, a large number of replicas (10,000) in the controlled experiments, and an explicit attempt to go beyond purely qualitative simulation results. However, the key quantitative assumption behind the theoretical derivation is currently unverified, and the validation in Figure 3 is partly a consistency check rather than an independent prediction. The central causal claim is therefore not yet established at the level the paper claims, though it is likely fixable with additional specification and testing.

major comments (4)

[§4.1, Eqs. (3)-(4)] The identities sum_{i in C1} k_i = 2m n1 and sum_{i in C2} k_i = 2m n2 are asserted after noting that an arriving agent connected only to one cluster creates 2m oriented connections. This is only correct if each cluster was assembled by the same growth rule, i.e., each added node links to m existing nodes. Section 4.2 does not specify how the initial clusters of 225 and 25 agents are wired internally. If they are fully connected, as the initial network of N0 agents is described in Section 2, the degree sums are n1(n1-1) = 50,400 and n2(n2-1) = 600 rather than 1,350 and 150 for m=3, changing Eq. (7) by a large factor. Because this assumption enters the theoretical curves in Figure 3, the reported fit at t=0 cannot be evaluated without knowing the cluster construction.
[§4.2, Fig. 3] The theoretical probabilities are computed using the average cluster positions and sizes from the same 10,000 replicas that produce the empirical frequencies. This makes Figure 3 a consistency check rather than an independent test of the bridge-agent mechanism. To support the claim that bridge agents are the cause of cascades, the authors should either measure the degree sums directly in the simulations and insert them into Eq. (2), or derive the degree sums from a fully specified construction and then compare the resulting formula with data not used in its calibration.
[§3.1, Fig. 1] The central empirical distinction between sudden cascades for beta in [15,20] and smoother merging for beta > 20 is illustrated by two representative runs (beta=15 and beta=30) without error bars or a systematic parameter sweep. Since the phenomenon is stochastic and the paper emphasizes the shape of the merging process, the reader needs either ensemble-averaged trajectories with confidence bands or a quantitative measure of cascade suddenness across a range of beta values.
[§4.1, Eqs. (9)-(10)] The product formulas for the probability of connecting only to C1 or only to C2 account for the decreasing number of available agents in the chosen cluster but ignore that each accepted link increases the degree, and hence the future attachment weight, of the selected agent. For m=3 this may be a minor correction, but the derivation as written is not the exact probability under the model's sequential attachment rule; the approximation should be stated explicitly.

minor comments (5)

[Algorithm 1, line 8] The condition 'if random(0, 1) < 1/N' appears to omit the parameter delta; the model description and the input list give the probability of adding a new agent as delta/N.
[§4.2] There are several typos and awkward phrasings, including 'quite very accurately,' 'frequence' in §3.1, and 'probability to of arriving bridge agent' in §4.2.
[Fig. 3] The panels would benefit from error bars or shaded confidence intervals, especially because the text claims that the fit is 'quite well' at t=50 and that accuracy 'decreases significantly' at t=100.
[§4.2] Inserting average positions and sizes into a nonlinear formula can bias the theoretical curves; a more direct comparison would compute P_b for each replica and then average the resulting probabilities.
[§2] The effect of the parameters delta and m on the presence and sharpness of cascades is not systematically explored, although Fig. 2 uses delta=6 while Fig. 1 uses delta=3; a sentence on parameter sensitivity would help.

Circularity Check

0 steps flagged · score 1.0 of 10

No significant circularity: the bridge-probability derivation is algebraically self-contained; the in-sample validation and under-specified cluster topology are caveats, not circular reductions.

full rationale

The claimed derivation chain is largely self-contained. Section 4.1 starts from the model's link probability (Eq. 1) and derives the bridge probability via Eqs. (2)-(11); the algebra is explicit and no fitted parameter is introduced. The only non-algebraic input is Eqs. (3)-(4), which assert each cluster's degree sum is 2m n_i; this is a modeling assumption that would hold if clusters were grown by the same m-link process, but Section 4.2 does not specify the internal topology of the initialized clusters. That is a reproducibility and correctness gap, not a circular reduction, because the formula is not defined in terms of the quantity it predicts. The validation in Fig. 3 evaluates the theoretical curves using the measured average cluster positions and sizes from the same 10,000 replicas, so the agreement at t=50 and t=100 is a consistency check rather than an out-of-sample prediction; however, the functional dependence on the arriving agent's opinion x and on beta is a non-trivial test, and the paper explicitly labels the curves as approximations. The causal statement that bridges trigger merging is essentially entailed by the model's link-formation rule (new agents are the only source of new links), but the quantitative characterization of bridge probability and its beta dependence is independent content. Self-citations ([17], [18]) are used for model context and prior results, not to justify the new cascade mechanism. Overall, no step reduces to its own inputs by construction.

Assumptions & free parameters 3 free parameters · 3 assumptions · 0 invented entities

The theoretical explanation depends on hand-set initial cluster sizes and opinion gaps, plus a simplifying assumption about internal degrees that is not guaranteed by the model. These are not fitted to data in the usual sense, but they are free choices that the central bridge-probability calculation rests on.

free parameters (3)

n1 (initial major cluster size) = 225
Hand-chosen in §4.2 for the idealized bridge-agent experiments; the theoretical bridge probability P_b directly depends on n1 via Eq. (7)-(9). Not fitted to data, but a free choice that shapes the theory-experiment comparison.
n2 (initial minor cluster size) = 25
Hand-chosen in §4.2; same rationale as n1.
delta_epsilon (opinion gap offset) = 0.03 (theta2 = 0.83, theta1 = 0.5, epsilon = 0.3)
Hand-set to place the minor cluster just outside the confidence bound of the major cluster in the idealized experiments. This offset is a free choice and appears in the exponential terms of Eq. (7).

assumptions (3)

domain assumption All agents in each cluster share exactly the same opinion and the clusters have no internal opinion spread.
Invoked in §4.1 to derive P2(n1,n2); it is an idealization that becomes less valid as interactions proceed, which the authors acknowledge.
ad hoc to paper For each cluster, the sum of degrees equals 2 m n_i (Eqs. 3-4).
This is not justified from the model: if the cluster were fully connected, the sum would be n_i(n_i-1). It is introduced specifically to simplify Eq. (5) and is load-bearing for all subsequent formulas.
domain assumption Arriving agents connect to m distinct existing agents, drawing without replacement according to the link probability phi_N(i).
Used in Eqs. (9)-(10) to compute probabilities of exclusive connections. This matches the algorithm described in §2.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Bounded confidence dynamics generates opinion cascades on growing scale-free networks." pith.science (2026). https://pith.science/paper/CMF2T3EP

@misc{pith2026250602669,
  author       = {Pith},
  title        = {Pith review of: Bounded confidence dynamics generates opinion cascades on growing scale-free networks},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/CMF2T3EP}},
  note         = {Machine review of arXiv:2506.02669}
}

read the original abstract

We study the pairwise bounded confidence model on scale-free networks where new agents regularly arrive over time. The probability that arriving agents form links to preexisting ones depends on both agent degree and opinion proximity. In parameter value ranges where both factors impact the link choice, a new phenomenon is observed. Minor clusters continuously form on the periphery of the opinion space and remain stable for a time, before suddenly merging with a major cluster in the network. We label these processes as "opinion cascades", and analyse their origin and behavior. They are triggered by the arrival of agents acting as "bridges" between the previously disconnected minor and major clusters. Lastly, we propose theoretical approximations to describe the varying shapes and merging behavior of opinion cascades under different conditions.

Figures

Figures reproduced from arXiv: 2506.02669 by the authors.

**Figure 2.** Simulation result after detecting communities via the Louvain method. Each [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗

**Figure 3.** Theoretical probability distributions and experimental frequencies of an arriving [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗

Figures from the paper (2 more)

**Figure 4.** Figure 4: Average (over the 10,000 replicas of experiments) number of "Bridge" agents [PITH_FULL_IMAGE:figures/full_fig_p009_4.png]

**Figure 5.** Figure 5: Average (over the 10,000 replicas of experiments) opinions of agents by type [PITH_FULL_IMAGE:figures/full_fig_p010_5.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

19 extracted references · 18 canonical work pages

[1]

Atheoryoffads,fashion,custom,andcultural change as informational cascades,

S.Bikhchandani,D.Hirshleifer,andI.Welch,“Atheoryoffads,fashion,custom,andcultural change as informational cascades,”Journal of Political Economy, vol. 100, no. 5, pp. 992– 1026, Oct. 1992

work page 1992
[2]

The spreading of misinformation online,

M. Del Vicario, A. Bessi, F. Zollo, F. Petroni, A. Scala, G. Caldarelli, H. E. Stanley, and W. Quattrociocchi, “The spreading of misinformation online,”Proceedings of the National Academy of Sciences, vol. 113, no. 3, pp. 554–559, Jan. 2016. 12 D. Hernandez et al

work page 2016
[3]

Cognitivecascades:Howtomodel(and potentially counter) the spread of fake news,

N.Rabb,L.Cowen,J.P.deRuiter,andM.Scheutz,“Cognitivecascades:Howtomodel(and potentially counter) the spread of fake news,”PLOS ONE, vol. 17, no. 1, p. e0261811, Jan. 2022

work page 2022
[4]

A simple model of global cascades on random networks,

D. J. Watts, “A simple model of global cascades on random networks,”Proceedings of the National Academy of Sciences, vol. 99, no. 9, pp. 5766–5771, Apr. 2002

work page 2002
[5]

M. E. J. Newman,Networks, second edition ed. Oxford: Oxford University Press, 2018, literaturverzeichnis: Seite 737-750 ; Index: Seite 751-780

work page 2018
[6]

Cascade dynamics of complex propagation,

D. Centola, V. M. Eguíluz, and M. W. Macy, “Cascade dynamics of complex propagation,” Physica A: Statistical Mechanics and its Applications, vol. 374, no. 1, pp. 449–456, Jan. 2007

work page 2007
[7]

Complex propagation on directed small world networks,

Y. Gandica, I. Bonalde, and J. L. Cabrera, “Complex propagation on directed small world networks,”PhysicaA:StatisticalMechanicsanditsApplications ,vol.389,no.19,pp.4172– 4176, Oct. 2010

work page 2010
[8]

Small butslowworld:Hownetworktopologyandburstinessslowdownspreading,

M.Karsai,M.Kivelä,R.K.Pan,K.Kaski,J.Kertész,A.-L.Barabási,andJ.Saramäki,“Small butslowworld:Hownetworktopologyandburstinessslowdownspreading,” PhysicalReview E, vol. 83, no. 2, p. 025102, Feb. 2011

work page 2011

Show all 19 references

[9]

Modelingoftemporalfluctuation scaling in online news network with independent cascade model,

J.Chołoniewski,J.Sienkiewicz,G.Leban,andJ.A.Hołyst,“Modelingoftemporalfluctuation scaling in online news network with independent cascade model,”Physica A: Statistical Mechanics and its Applications, vol. 523, pp. 129–144, Jun. 2019

2019
[10]

Effects of network structure, competition and memory time on social spreading phenomena,

J. P. Gleeson, K. P. O’Sullivan, R. A. Baños, and Y. Moreno, “Effects of network structure, competition and memory time on social spreading phenomena,”Physical Review X, vol. 6, no. 2, p. 021019, May 2016

2016
[11]

Reference to global state and social contagion dynamics,

K. Jin and U. Yu, “Reference to global state and social contagion dynamics,”Frontiers in Physics, vol. 9, May 2021

2021
[12]

Mixing beliefs among interacting agents,

G. Deffuant, D. Neau, F. Amblard, and G. Weisbuch, “Mixing beliefs among interacting agents,”Advances in Complex Systems, vol. 3, pp. 87–98, 2000

2000
[13]

Opiniondynamicsandboundedconfidence:Models,analysis and simulation

R.HegselmannandU.Krause,“Opiniondynamicsandboundedconfidence:Models,analysis and simulation.”Journal of Artificial Societies and Social Simulation, vol. 5, no. 3, 2002

2002
[14]

Hovland and M

C. Hovland and M. Sherif,Social judgment: Assimilation and contrast effects in communi- cation and attitude change.Greenwood, 1980

1980
[15]

Universality of the threshold for complete consensus for the opinion dynamics of deffuantet al

S. FORTUNATO, “Universality of the threshold for complete consensus for the opinion dynamics of deffuantet al.”International Journal of Modern Physics C, vol. 15, no. 09, pp. 1301–1307, Nov. 2004

2004
[16]

Simulation of consensus model of deffuant et al on a barabasi-albert network,

D. Stauffer and H. Meyer-Ortmanns, “Simulation of consensus model of deffuant et al on a barabasi-albert network,”Journal of Modern Physics C, vol. 15, no. 2, pp. 241–246, 2004

2004
[17]

Bounded confidence model on growing populations,

Y. GANDICA and G. DEFFUANT, “Bounded confidence model on growing populations,” Advances in Complex Systems, vol. 27, no. 07n08, Nov. 2024

2024
[18]

The role of homophily in the emergence of opinion controver- sies,

F. Gargiulo and Y. Gandica, “The role of homophily in the emergence of opinion controver- sies,”Journal of Artificial Societies and Social Simulation, vol. 20, no. 3, 2017

2017
[19]

Fast unfolding of communities in large networks,

V. D. Blondel, J.-L. Guillaume, R. Lambiotte, and E. Lefebvre, “Fast unfolding of communities in large networks,” Journal of Statistical Mechanics: Theory and Experiment , vol. 2008, no. 10, p. P10008, Oct. 2008. [Online]. Available: http://dx.doi.org/10.1088/1742-5468/2008/10/P10008

2008 doi

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.

[1] [1]

Atheoryoffads,fashion,custom,andcultural change as informational cascades,

S.Bikhchandani,D.Hirshleifer,andI.Welch,“Atheoryoffads,fashion,custom,andcultural change as informational cascades,”Journal of Political Economy, vol. 100, no. 5, pp. 992– 1026, Oct. 1992

work page 1992

[2] [2]

The spreading of misinformation online,

M. Del Vicario, A. Bessi, F. Zollo, F. Petroni, A. Scala, G. Caldarelli, H. E. Stanley, and W. Quattrociocchi, “The spreading of misinformation online,”Proceedings of the National Academy of Sciences, vol. 113, no. 3, pp. 554–559, Jan. 2016. 12 D. Hernandez et al

work page 2016

[3] [3]

Cognitivecascades:Howtomodel(and potentially counter) the spread of fake news,

N.Rabb,L.Cowen,J.P.deRuiter,andM.Scheutz,“Cognitivecascades:Howtomodel(and potentially counter) the spread of fake news,”PLOS ONE, vol. 17, no. 1, p. e0261811, Jan. 2022

work page 2022

[4] [4]

A simple model of global cascades on random networks,

D. J. Watts, “A simple model of global cascades on random networks,”Proceedings of the National Academy of Sciences, vol. 99, no. 9, pp. 5766–5771, Apr. 2002

work page 2002

[5] [5]

M. E. J. Newman,Networks, second edition ed. Oxford: Oxford University Press, 2018, literaturverzeichnis: Seite 737-750 ; Index: Seite 751-780

work page 2018

[6] [6]

Cascade dynamics of complex propagation,

D. Centola, V. M. Eguíluz, and M. W. Macy, “Cascade dynamics of complex propagation,” Physica A: Statistical Mechanics and its Applications, vol. 374, no. 1, pp. 449–456, Jan. 2007

work page 2007

[7] [7]

Complex propagation on directed small world networks,

Y. Gandica, I. Bonalde, and J. L. Cabrera, “Complex propagation on directed small world networks,”PhysicaA:StatisticalMechanicsanditsApplications ,vol.389,no.19,pp.4172– 4176, Oct. 2010

work page 2010

[8] [8]

Small butslowworld:Hownetworktopologyandburstinessslowdownspreading,

M.Karsai,M.Kivelä,R.K.Pan,K.Kaski,J.Kertész,A.-L.Barabási,andJ.Saramäki,“Small butslowworld:Hownetworktopologyandburstinessslowdownspreading,” PhysicalReview E, vol. 83, no. 2, p. 025102, Feb. 2011

work page 2011