Pith. sign in

REVIEW 4 major objections 6 minor 32 references

Emergence of network communities driven by local rules

T0 review · 4 major / 6 minor · reviewed 2026-08-10 · deepseek-v4-flash

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

Pith's one-line read The paper argues that local network-growth rules alone can force communities to appear, without any node heterogeneity, once networks are large enough.

desk verdict A useful formal threshold for when local growth rules yield detectable communities, backed by honest simulation evidence, but the necessity claim and the choice of null both need more care. read the letter →

arxiv 2501.17042 v4 pith:AIXTNJ52 submitted 2025-01-28 physics.soc-ph cond-mat.dis-nncs.DMcs.SImath.CO

Alexei Vazquez This is my paper · ORCID
classification physics.soc-phcond-mat.dis-nncs.DMcs.SImath.CO MSC 05C8005D1091D30 PACS 89.75.Fb89.75.Hc
keywords networkcommunitiesemergentpropertieslocalrulesRamseycommunitynumberstochasticblockmodelInfomapdegree-preservingrewiringgenerativemodels
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

The paper sets out to show that network communities can appear without any pre-existing node types or node heterogeneity: purely local rules of network growth are enough. To make this precise it defines the Ramsey community number, the smallest network size at which a given generative model yields at least two detected communities with probability at least $1-\epsilon$, and the emergent communities property for models where that number exists for every $\epsilon$. Using the degree-corrected stochastic block model and regularized Infomap, the paper reports finite Ramsey numbers for the local search, duplication-split, and Watts-Strogatz models, while degree-preserving randomized versions of these networks show no emergent communities. It conjectures that network communities are an emergent property of networks evolving with local rules, which would mean observed segregation in real networks could reflect how links form rather than hidden node attributes.

What carries the argument

The central object is the Ramsey community number $r_\kappa(f_G, f_\kappa, \epsilon)$: the minimum $n$ such that a graph generative model $f_G$ on $n$ vertices has probability at least $1-\epsilon$ of being assigned two or more communities by a community count function $f_\kappa$. A model has the emergent communities property when $r_\kappa$ exists for all $0<\epsilon<1$. This number carries the argument by converting a qualitative intuition about emergence into a measurable threshold, and the key comparison is between a model and its degree-preserving randomized version: the local-rule models cross the threshold at finite size, while the randomized versions never do. Tunable model families (Watts-Strogatz rewiring, Barabasi-Albert attractiveness, and the bubble model's cycle length and attachment distance) are then used to show that the finite threshold tracks the degree of locality, not degree heterogeneity.

What would settle it

Generate LS($n$, $d=1$), DS($n$, $q=0.5$), and WS($n$, $K=\{1,2\}$, $p=0.1$) networks at large $n$ (say 100,000 nodes) and apply a community detector calibrated to report no communities on configuration-model random graphs; if that detector still reports two or more communities on almost every locally grown instance, the emergent-communities claim survives, while if it reports none, the reported dichotomy is a method artifact. In the other direction, running the degree-corrected stochastic block model on BA($n$, $m=2$) networks up to $n=10^6$ and finding $P_\kappa$ rising back to 1 would falsify the claim that the Barabasi-Albert model lacks the emergent communities property.

Watch

Extended reading notes

Core claim

The core claim is the conjecture that network communities are an emergent property of networks evolving with local rules, stated against the null picture that communities require node heterogeneity. The paper gives the conjecture an operational form: for the local search, duplication-split, and Watts-Strogatz (with at least two neighbor sets) models, the Ramsey community number is finite, whereas for their degree-preserving randomized versions it does not exist within the tested sizes, and the canonical Barabasi-Albert model lacks the emergent communities property. The evidence also shows that implicit locality suffices: Barabasi-Albert models with baseline attractiveness and degree exponent $\gamma<3$ do have finite Ramsey numbers, with $r_\kappa$ falling as the clustering coefficient rises, and the bubble model's Ramsey number grows as the added cycles lengthen and disappears when attachments can reach more than two steps away. Communities, on this picture, are a large-size consequence of local connection rules rather than a trace of node types.

Load-bearing premise

The load-bearing premise is that the two community-detection methods used here, the degree-corrected stochastic block model and regularized Infomap, reveal the true presence or absence of communities; the paper itself notes that there is no gold standard for community detection, so a method artifact could in principle manufacture the local-versus-nonlocal split.

Editorial extensions

If this is right

  • If the conjecture is right, community structure in a single static network cannot be taken as evidence of hidden node types; the network's growth history alone could have produced it.
  • Degree-preserving rewiring becomes a standard control: a generative model whose communities survive rewiring, or whose randomized version also shows communities, is not demonstrating emergent community structure.
  • Scale-free networks generated by preferential attachment with baseline attractiveness and degree exponent below 3 are predicted to show communities at finite size, while the canonical Barabasi-Albert model should not.
  • The absence of the property for the one-neighbor Watts-Strogatz ring and its presence for the two-neighbor ring delimits what 'local' must mean: enough local connections beyond immediate neighbors are required.
  • For real networks, the framework suggests that observed communities may be explained by local dynamics, making node attributes a possible consequence of community membership rather than its cause.

Reading between the lines

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

  • The Ramsey community number could double as a benchmark for community detection methods: a method that assigns communities to degree-preserving random graphs at finite size is unlikely to be a reliable probe of emergence.
  • A testable extension would classify real networks by whether their communities persist under degree-preserving rewiring; networks whose communities persist would be candidates for growth driven by global rather than local link selection.
  • The apparent divergence of $r_\kappa$ as the clustering coefficient approaches zero suggests a possible phase transition in community emergence; scanning the Barabasi-Albert attractiveness parameter near the boundary could locate a critical clustering value.
  • Because the numerical values of $r_\kappa$ depend on the detection method, a natural strengthening would define the Ramsey number against a consensus of several regularized methods or against planted ground truth, making the threshold less method-dependent.
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.

Referee Report

4 major / 6 minor

Summary. The paper proposes a Ramsey-theoretic criterion for the emergence of network communities: the Ramsey community number r_kappa is the smallest graph size n at which a generative model produces graphs with two or more detected communities with probability at least 1-epsilon, and a model has the emergent communities property if r_kappa exists for all epsilon. Using numerical simulations of four generative model families (local search, duplication-split, Watts-Strogatz, and Barabasi-Albert, plus variants), the author reports that the three models with explicit local rules have finite r_kappa, while their degree-preserving randomized versions do not, and that the canonical BA model does not have the emergent communities property. The paper concludes by conjecturing that network communities are an emergent property of networks evolving with local rules, and that node heterogeneity is not a necessary condition.

Significance. The paper introduces a crisp, potentially reusable definition (the Ramsey community number) and supports it with a reproducible computational pipeline (code at github.com/av2atgh/ramsey netcom) and substantial simulations: 1000 instances per condition, two independent community detection methods (degree-corrected SBM and regularized Infomap), and degree-preserving randomization controls. The randomized baselines are a genuine strength and correctly show that simple degree heterogeneity is not the explanation. If the central conjecture survives the robustness checks below, the paper would make a useful conceptual contribution to the community-detection literature by shifting attention from node attributes to network-growth dynamics. However, the paper is fully simulation-based, contains no proofs, and its central dichotomy is method-dependent; those limitations currently prevent the strong necessary-condition claim in Sec. VIII from being established.

major comments (4)
  1. [Sec. IV.B and Definition 2] The emergent communities property is defined relative to a community-count function f_kappa, and all conclusions depend on the two chosen algorithms. The paper acknowledges in Sec. IV.B that 'there is no gold standard,' yet the Infomap regularization strength is fixed at 1.0 precisely because it makes random networks appear community-free. This risks circularity: the detection method is tuned to reproduce the desired local-vs-randomized dichotomy. I request a robustness analysis across a range of regularization_strength values, and ideally a third detection method whose null model preserves local correlations (for example, triangle-preserving or 2K-preserving rewiring) to show that the finite r_kappa values are not artifacts of the configuration-model null implicit in SBM-DC and Infomap.
  2. [Sec. V and Table I] The Ramsey community numbers in Table I are reported as point estimates without confidence intervals or error bars. The text says 'variations in the last digit due to the stochastic nature' but does not quantify this, and the binary-search procedure does not state how many instances are generated at each candidate n (Sec. II specifies 1000 for the P_kappa curves, but the search points are not documented). Since the paper draws quantitative comparisons across models and parameters (for example, LS d=1 vs d=2, DS q=0.3 vs q=0.5, and WS p=0.1 vs p=0.2), the reader cannot assess whether the reported differences are statistically meaningful.
  3. [Sec. VI and Table I] The finite r_kappa values for BA(n,m,a) with a<m complicate the abstract's claim that local rules are necessary. The paper's response is to reinterpret locality as 'implicit formation of local structures' through clustering, but this shifts the operative concept from local rules to local structure. The necessary-condition statement in Sec. VIII should be formulated and tested in terms of local structure (for example, finite clustering or the presence of short cycles) rather than inherited from the local-rule models; otherwise the claim is not supported by the BA(n,m,a) results.
  4. [Sec. III.D and Fig. 5] The evidence that BA(n,m=2) lacks the emergent communities property rests on a non-monotonic P_kappa curve that peaks at about 0.5 near n ~ 1000 and then declines with increasing n, a behavior the paper states it cannot explain. Because the nonexistence of r_kappa is inferred from this decline, the authors should either provide a mechanistic explanation for the non-monotonicity or demonstrate that the decline is not a community-detection artifact, for example by showing that the detected communities near the peak are stable under algorithmic variations and that the decline persists under alternative detection settings.
minor comments (6)
  1. [Abstract] The word 'warranties' in 'the minimum graph size that warranties the emergence' should be 'warrants'; the same typo appears in Sec. I.
  2. [Table I caption] The abbreviation 'SMB' should be 'SBM' (stochastic block model); the same typo appears in Sec. V.
  3. [Sec. IV.B] The sentence 'there is no gold standard, we cannot tell which methods is right' has a subject-verb agreement error and should be rephrased, e.g., 'we cannot tell which method is right.'
  4. [Sec. VII.A] The sentence 'I obtain the Ramsey community numbers r_kappa(L=1,W=1)=108 and r_kappa(L=1,W=1)=1090' reports the same condition twice; the second value is presumably for W=2 and should be corrected.
  5. [Sec. V, Definition 1] The notation 'fG(n) :→ SG' is malformed; it should define fG as a map from a size parameter to the set of graphs, and the signature of P_kappa (a function on graphs, but written 'P_kappa(fG)') should be clarified.
  6. [Sec. II] The Greek word for community is rendered as garbled text ('kappaoinu ´o t eta'); it should be 'koinotita' or the proper Greek 'κοινότητα'.

Circularity Check

0 steps flagged · score 1.0 of 10

No significant circularity: simulations are parameter scans, and community-detection choices are null-calibrated rather than fitted to the target dichotomy.

full rationale

The derivation chain is: define generative models, generate graphs, apply community detection, measure P_kappa, and define the Ramsey community number. No parameter of any generative model is fitted to make r_kappa finite; model parameters (m, a, q, p, L, K) are scanned inputs, and community detection is applied to generated graphs rather than designed to produce the stated thresholds. The r_kappa values are measurements, not derivations that reduce to assumed constants. The choice of the degree-corrected stochastic block model and regularized Infomap is justified by their null behavior on degree-randomized graphs; this is a detector calibration, not a fit to the target local-vs-random dichotomy, since the local models must still cross P_kappa = 1 under the same detector. The randomized versions are, by construction, the null model of these detection methods, so their lack of communities is a sanity check rather than a derived prediction. The self-citations (Refs. 14, 21, 26) supply models and motivation, but the models are fully defined in the text, so the results do not reduce to those citations. The acknowledged absence of a gold standard (Sec. IV.B) is a methodological limitation, not circularity.

Assumptions & free parameters 2 free parameters · 5 assumptions · 0 invented entities

The model parameters (q, p, m, a, L, W) are scanned inputs, not fitted to the target outcome, so they are not listed as free parameters. The two hand-chosen thresholds below define the reported r_kappa values rather than being fitted to force the result. The paper introduces no new physical or mechanistic entities; the Ramsey community number and the emergent communities property are mathematical definitions used to quantify the simulations.

free parameters (2)
  • epsilon (error rate) = 0.05
    The Ramsey community number is defined for a fixed error rate; the paper chooses epsilon = 0.05 (95% confidence). This is a conventional threshold chosen by hand and affects the reported r_kappa values, though the qualitative findings are robust across the tested methods.
  • Infomap regularization_strength = 1.0
    Set explicitly in Section IV.B to make Infomap classify random networks as single-community. This choice affects r_kappa(IM) values and is not fitted to data, but it is a hand-picked sensitivity setting.
assumptions (5)
  • domain assumption The degree-corrected stochastic block model as implemented in graph-tool (minimize_blockmodel_dl with default parameters) returns a valid community count for these networks.
    Invoked in Section II as the primary community detection method; the paper validates it indirectly by comparing with Infomap and rewired baselines, but no formal guarantee is known.
  • domain assumption Regularized Infomap with regularization_strength=1.0 correctly labels degree-randomized networks as having no communities.
    Section IV.B relies on this property to corroborate the SBM results; the parameter is set by hand based on prior literature.
  • domain assumption The configuration model random rewiring that preserves degrees is the appropriate null model for 'no local structure'.
    Section II uses graph_tool.random_rewire; the paper's conclusion that randomized versions do not have communities depends on this rewiring preserving the degree sequence while destroying correlations.
  • domain assumption Finite-sample estimates from 1000 instances, up to n=100,000, justify statements that r_kappa exists or does not exist.
    The emergent communities property is defined for all n, but the paper checks finite sizes; for n=100,000 only 10 instances are used. The extrapolation to 'almost certainty' is an assumption.
  • ad hoc to paper The finite family of generative models tested is representative enough to support the 'necessary condition' statement.
    Section V concludes a necessary condition from a dichotomy observed across the studied models; this is an inductive step, not a mathematical proof.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Emergence of network communities driven by local rules." pith.science (2026). https://pith.science/paper/AIXTNJ52

@misc{pith2026250117042,
  author       = {Pith},
  title        = {Pith review of: Emergence of network communities driven by local rules},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/AIXTNJ52}},
  note         = {Machine review of arXiv:2501.17042}
}
abstract

Natural systems are modeled by networks with nodes and links. Often the nodes are segregated into communities with different connectivity patterns. Node heterogeneity such as political affiliation in social networks or biological function in gene networks are highlighted as key factors driving the segregation of nodes into communities. Here, by means of numerical simulations, I show that node heterogeneity is not a necessary requirement. To this end I introduce the Ramsey community number, $r_ \kappa$, the minimum graph size that warranties the emergence of network communities with almost certainty. Using the stochastic block model and Infomap methods for community detection, I show that networks generated by local rules have finite $r_ \kappa$ values while their randomized versions do not have emergent communities. I conjecture that network communities are an emergent property of networks evolving with local rules.

Figures

Figures reproduced from arXiv: 2501.17042 by the authors.

Figure 1
Figure 1. shows an instance of LS(n = 50, d = 1). The stochastic block model with degree sequence correction detected two communities that we can corroborate from [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. FIG. 2. a) Probability of having 2 or more communities and [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗
Figure 3
Figure 3. shows an instance of DS(n = 50, q = 0.3). The stochastic block model with degree sequence correc￾tion identified two communities. As I mentioned, the du￾plication rule does not make triangles and the split rule breaks triangles if they would exist. You can confirm that by looking and the example in [PITH_FULL_IMAGE:figures/full_fig_p003_3.png] view at source ↗
Figures from the paper (6 more)
Figure 4
Figure 4. Figure 4: shows an instance of WS(n = 50, K = {1, 2}, p = 0.1). This instance has two communities that are verified by visual inspection. The communities retain part of the starting ring arrangement, giving them a different look than the previous examples. The Pκ vs n plot for t…
Figure 5
Figure 5. Figure 5: FIG. 5. Probability of having 2 or more communities as a [PITH_FULL_IMAGE:figures/full_fig_p004_5.png]
Figure 6
Figure 6. Figure 6: FIG. 6. Probability of having 2 or more communities ( [PITH_FULL_IMAGE:figures/full_fig_p005_6.png]
Figure 7
Figure 7. Figure 7: FIG. 7. Probability of having 2 or more communities ( [PITH_FULL_IMAGE:figures/full_fig_p005_7.png]
Figure 10
Figure 10. Figure 10: FIG. 10. An instance of BB( [PITH_FULL_IMAGE:figures/full_fig_p007_10.png]
Figure 11
Figure 11. Figure 11: FIG. 11. Ramsey community number as a function of [PITH_FULL_IMAGE:figures/full_fig_p008_11.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

32 extracted references · 22 canonical work pages

  1. [1]

    M. E. J. Newman, Proceedings of the Na- tional Academy of Sciences 103, 8577 (2006), https://www.pnas.org/doi/pdf/10.1073/pnas.0601602103. 9

    doi:10.1073/pnas.0601602103 2006

  2. [2]

    Radicchi, C

    F. Radicchi, C. Castellano, F. Cecconi, V. Loreto, and D. Parisi, Proceedings of the National Academy of Sciences 101, 2658 (2004), https://www.pnas.org/doi/pdf/10.1073/pnas.0400054101

    doi:10.1073/pnas.0400054101 2004

  3. [3]

    Fortunato and M

    S. Fortunato and M. Barth´ elemy, Proceedings of the National Academy of Sciences 104, 36 (2007), https://www.pnas.org/doi/pdf/10.1073/pnas.0605965104

    doi:10.1073/pnas.0605965104 2007

  4. [4]

    J. M. Hofman and C. H. Wiggins, Phys. Rev. Lett. 100, 258701 (2008)

    work page 2008

  5. [5]

    Karrer and M

    B. Karrer and M. E. J. Newman, Phys. Rev. E83, 016107 (2011)

    2011

  6. [6]

    Network reconstruction via the minimum description length principle,

    T. P. Peixoto, “Network reconstruction via the minimum description length principle,” (2024), arXiv:2405.01015 [stat.ML]

    arXiv 2024

  7. [7]

    Vazquez, J

    A. Vazquez, J. Stat. Mech.: Theory Exp. 2009, P07006 (2009)

    work page 2009

  8. [8]

    Contisciani, F

    M. Contisciani, F. Battiston, and C. De Bacco, Nat Commun 13, 7229 (2022), publisher: Nature Publishing Group

    work page 2022

Show all 32 references
  1. [9]

    E. M. Jin, M. Girvan, and M. E. J. Newman, Phys. Rev. E 64, 046132 (2001)

    2001

  2. [10]

    R. V. Sol´ e and S. Valverde, Journal of The Royal Society Interface 5, 129 (2007), publisher: Royal Society

    2007

  3. [11]

    Bianconi, R

    G. Bianconi, R. K. Darst, J. Iacovacci, and S. Fortunato, Phys. Rev. E 90, 042806 (2014)

    2014

  4. [12]

    Wharrie, L

    S. Wharrie, L. Azizi, and E. G. Altmann, Phys. Rev. E 100, 022315 (2019)

    2019

  5. [13]

    Graham, B

    R. Graham, B. Rothschild, and J. Spencer, Ramsey The- ory, A Wiley-Interscience Publication (Wiley, 1980)

    1980

  6. [14]

    Vazquez, Phys

    A. Vazquez, Phys. Rev. E 67, 056104 (2003)

    2003

  7. [15]

    T. P. Peixoto, figshare (2014), 10.6084/m9.figshare.1164194

    2014 doi

  8. [16]

    T. P. Peixoto, Phys. Rev. X 12, 011004 (2022)

    2022

  9. [17]

    V. D. Blondel, J.-L. Guillaume, R. Lambiotte, and E. Lefebvre, Journal of Statistical Mechanics: Theory and Experiment 2008, P10008 (2008)

    2008

  10. [18]

    Hagberg, P

    A. Hagberg, P. J. Swart, and D. A. Schult (Los Alamos National Laboratory (LANL), Los Alamos, NM (United States), 2008)

    2008

  11. [19]

    The MapE- quation software package,

    D. Edler, A. Holmgren, and M. Rosvall, “The MapE- quation software package,” https://mapequation.org (2025)

    2025

  12. [20]

    Park and M

    J. Park and M. E. J. Newman, Phys. Rev. E 70, 066117 (2004)

    2004

  13. [21]

    Vazquez, Europh

    A. Vazquez, Europh. Lett. 54, 430 (2001)

    2001

  14. [22]

    V´ azquez, A

    A. V´ azquez, A. Flammini, A. Maritan, and A. Vespig- nani, Complexus 1, 38 (2003)

    2003

  15. [23]

    Chung, L

    F. Chung, L. Lu, T. G. Dewey, and D. J. Galas, J. Comp. Biol. 10, 677 (2003)

    2003

  16. [24]

    Pastor-Satorras, E

    R. Pastor-Satorras, E. Smith, and R. V. Sol´ e, J. Theor. Biol. 222, 199 (2003)

    2003

  17. [25]

    P. L. Krapivsky and S. Redner, Phys. Rev. E 71, 036118 (2005)

    2005

  18. [26]

    Vazquez, I

    A. Vazquez, I. Pozzana, G. Kalogridis, and C. Ellinas, Sci Rep 13, 509 (2023)

    2023

  19. [27]

    D. J. Watts and S. H. Strogatz, Nature 393, 440 (1998)

    1998

  20. [28]

    Barab´ asi and R

    A.-L. Barab´ asi and R. Albert, Science286, 509 (1999)

    1999

  21. [29]

    Rosvall and C

    M. Rosvall and C. T. Bergstrom, Proceedings of the National Academy of Sciences 105, 1118 (2008), https://www.pnas.org/doi/pdf/10.1073/pnas.0706851105

    2008 doi

  22. [30]

    Smiljani´ c, D

    J. Smiljani´ c, D. Edler, and M. Rosvall, Phys. Rev. E 102, 012302 (2020)

    2020

  23. [31]

    S. N. Dorogovtsev, J. F. F. Mendes, and A. N. Samukhin, Phys. Rev. Lett. 85, 4633 (2000)

    2000

  24. [32]

    Klemm and V

    K. Klemm and V. M. Egu ´ ıluz, Phys. Rev. E65, 057102 (2002)

    2002

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.