arxiv: 2604.22677 · v2 · submitted 2026-04-24 · ⚛️ physics.soc-ph

Recognition: no theorem link

Self-Similarity in Online Networks During Social Movements

M. \'Angeles Serrano, Manuel Su\'arez, Y\'erali Gandica

Authors on Pith no claims yet

Pith reviewed 2026-05-12 01:46 UTC · model grok-4.3

classification ⚛️ physics.soc-ph

keywords self-similaritysocial movementshashtag co-occurrencecollective actiononline networksnetwork renormalisationscale invarianceTwitter data

0 comments

The pith

Self-similarity emerges in online networks exactly at the peak of social movements.

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

The paper examines hashtag co-occurrence networks built from Twitter data of three large, distinct social movements and applies a degree-thresholding renormalisation step to detect scale-invariant patterns. It shows that at the moments of highest activity these networks undergo a shift from modular to nested organisation, user participation peaks, and clustering measures become consistent across scales. A sympathetic reader would care because the pattern suggests a built-in way for information and coordination to amplify efficiently from small groups to large ones, independent of the specific geography or politics involved. The consistency across three unrelated events points to self-similarity as a possible common organising feature that enables successful collective action.

Core claim

By constructing co-occurrence networks from Twitter hashtag data and applying a degree-thresholding renormalisation procedure, we demonstrate that these highly correlated social phenomena exhibit clear signatures of self-similarity at peak mobilisation times. These critical points are characterised by modular-to-nested transitions, both in the co-occurrence networks and the bi-partite ones, maxima in user participation, and clustering spectrum collapse across multiple network scales. Despite their geographical and sociopolitical diversity, all three movements display remarkably analogous self-similar properties. Furthermore, the results hint at the emergence of a latent metric structure that

What carries the argument

Degree-thresholding renormalisation applied to hashtag co-occurrence networks, which exposes self-similar structure by collapsing data across resolution scales.

If this is right

Modular-to-nested transitions appear in both co-occurrence and bipartite networks precisely at peak mobilisation.
User participation maxima coincide with the onset of self-similar signatures.
Clustering spectra collapse to a common form across multiple network scales.
A latent metric structure emerges that permits hyperbolic embedding and an estimate of effective social distance.

Where Pith is reading between the lines

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

Real-time detection of self-similarity in streaming hashtag networks could serve as an observable marker of mobilisation intensity.
The pattern's independence from specific political content suggests it may apply to coordination on other platforms or even non-digital networks.
If the renormalisation step is robust, the same procedure could be used to compare mobilisation dynamics across historical events with available digital traces.

Load-bearing premise

That hashtag co-occurrence networks after degree-thresholding renormalisation capture genuine social coordination dynamics rather than platform algorithms, sampling choices, or the specific threshold value.

What would settle it

Repeating the same renormalisation analysis on hashtag data from another major mobilisation and finding neither modular-to-nested transitions nor clustering-spectrum collapse at the reported peak times.

Figures

Figures reproduced from arXiv: 2604.22677 by M. \'Angeles Serrano, Manuel Su\'arez, Y\'erali Gandica.

**Figure 1.** Figure 1: FIG. 1 view at source ↗

**Figure 2.** Figure 2: FIG. 2 view at source ↗

**Figure 3.** Figure 3: FIG. 3 view at source ↗

**Figure 4.** Figure 4: displays the hyperbolic embeddings of the CTW networks for the three social movements alongside the comparison of their empirical and inferred corresponding connection probability curves. The upper row shows the two-dimensional hyperbolic embeddings, where node size represents degree and spatial proximity in the hyperbolic plane reflects the likelihood of connection. In each embedding, the seed hashtags—… view at source ↗

read the original abstract

Online platforms provide an infrastructure for social movements, leaving digital traces that can be modelled as networks to quantify how information, participation, and coordination emerge during episodes of collective action and evolve over time. In this work, we unveil the emergence of scale-invariant online interaction patterns in social movements through network analysis of three geographically and sociopolitically distinct massive mobilisation events. By constructing co-occurrence networks from Twitter (now X) hashtag data and applying a degree-thresholding renormalisation procedure, we demonstrate that these highly correlated social phenomena exhibit clear signatures of self-similarity at peak mobilisation times. These critical points are characterised by modular-to-nested transitions, both in the co-occurrence networks and the bi-partite ones, maxima in user participation, and clustering spectrum collapse across multiple network scales. Despite their geographical and sociopolitical diversity, all three movements display remarkably analogous self-similar properties. Furthermore, the results hint at the emergence of a latent metric structure that supports successful hyperbolic embedding, providing an estimate of effective social distance. Together, these findings suggest that self-similarity may constitute a universal organising principle of social movements during peak mobilisation phases, as it lays the groundwork for the rapid amplification of information across scales that is necessary for the successful coordination of collective action.

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.

Desk Editor's Note private letter to a colleague

The paper applies degree-threshold renormalization to hashtag networks from three movements and reports self-similarity plus modular-to-nested shifts at peaks, but the absence of null models and threshold checks makes it unclear whether the patterns are intrinsic or induced by the procedure.

read the letter

The main observation is that three distinct social-movement hashtag networks, after degree-threshold renormalization, display self-similar structure at their peak mobilization times, with a shift from modular to nested organization and a collapse in the clustering spectrum. The consistency across the cases is the clearest positive finding, and the side note on possible hyperbolic embedding adds a concrete geometric angle that could be followed up elsewhere. The work takes established renormalization ideas from network physics and puts them on real, time-resolved Twitter data, which is a straightforward empirical step worth having in the literature. The cross-event comparison at least shows the pattern is not limited to one context. The soft spots sit in the methods. The abstract and description give no error bars on the collapse, no statistical test for the scale invariance, and no comparison to configuration-model rewirings or shuffled temporal controls that would keep the degree sequence while destroying the temporal structure. The choice of degree threshold and the definition of peak times also look post-hoc; if modest changes to either move the signatures, the claim that self-similarity is a universal coordination principle loses force. The weakest link is the assumption that the filtered co-occurrence networks mainly capture social coordination rather than platform recommendation effects or the filtering step itself. Without those checks the interpretation stays provisional. This paper is for readers who track scaling phenomena in digital social data or who want empirical examples of renormalization applied outside physics. A network scientist or computational social scientist could extract the figures and the three-event comparison even if the causal story needs tightening. It deserves peer review so that referees can request the missing robustness tests and null models; the core descriptive result is clear enough to be worth that scrutiny.

Referee Report

3 major / 2 minor

Summary. The manuscript constructs hashtag co-occurrence networks from Twitter data for three distinct social movements, applies a degree-thresholding renormalization procedure, and reports signatures of self-similarity (modular-to-nested transitions in both co-occurrence and bipartite networks, maxima in user participation, and clustering spectrum collapse across scales) precisely at peak mobilization times. These patterns are described as analogous across events despite their geographic and sociopolitical differences, with an additional suggestion of an emergent latent metric structure supporting hyperbolic embedding; the authors interpret this self-similarity as a potential universal organizing principle enabling cross-scale information amplification for collective action coordination.

Significance. If the self-similarity signatures prove robust, the work would offer a data-driven contribution to complex-systems approaches in social physics by linking scale-invariant network structure to mobilization dynamics across multiple real-world cases. The multi-event design and focus on peak-phase transitions provide a concrete basis for testing universality claims, though the current lack of statistical validation and null-model controls limits immediate impact.

major comments (3)

[Methods (network construction and renormalization procedure)] Methods (network construction and renormalization procedure): the degree-thresholding renormalization is central to producing the reported modular-to-nested transitions and clustering collapse, yet no robustness analysis across a range of threshold values is described. Without such checks, it remains possible that the self-similarity signatures are induced by the specific threshold choice rather than reflecting intrinsic coordination dynamics.
[Results (clustering spectrum collapse and transitions at peak times)] Results (clustering spectrum collapse and transitions at peak times): the abstract and results claim clear signatures of self-similarity and spectrum collapse, but supply no quantitative error bars, statistical significance tests for the collapse, or explicit comparisons to null models such as configuration-model rewirings that preserve the degree sequence. These omissions make it impossible to determine whether the observed patterns exceed what would be expected from sampling biases or platform effects alone.
[Discussion (interpretation as universal organizing principle)] Discussion (interpretation as universal organizing principle): the claim that self-similarity at peak mobilization constitutes a universal principle enabling cross-scale amplification rests on the assumption that the renormalized networks faithfully encode social coordination. The manuscript would need temporal or shuffled baselines demonstrating that the modular-to-nested transitions and participation maxima are specific to the identified peak phases rather than generic features of the data.

minor comments (2)

[Abstract] The abstract states that the three movements display 'remarkably analogous self-similar properties' but does not specify the quantitative similarity metric or distance measure used to establish this analogy.
[Throughout] Notation for the clustering spectrum, network scales, and the effective social distance derived from hyperbolic embedding should be introduced with explicit definitions in the main text to support reproducibility.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for their constructive and detailed comments on our manuscript. We address each major point below, clarifying our approach where possible and outlining revisions to strengthen the statistical and robustness aspects of the work.

read point-by-point responses

Referee: Methods (network construction and renormalization procedure): the degree-thresholding renormalization is central to producing the reported modular-to-nested transitions and clustering collapse, yet no robustness analysis across a range of threshold values is described. Without such checks, it remains possible that the self-similarity signatures are induced by the specific threshold choice rather than reflecting intrinsic coordination dynamics.

Authors: We selected the degree threshold to ensure the emergence of clear modular structure while preserving the core co-occurrence patterns observed in the raw data. However, we agree that explicit robustness checks are necessary to rule out threshold-specific artifacts. In the revised manuscript, we will add a supplementary analysis varying the threshold by approximately +/-20% around the chosen value and demonstrate that the modular-to-nested transitions, participation maxima, and clustering spectrum collapse persist consistently across this range. revision: yes
Referee: Results (clustering spectrum collapse and transitions at peak times): the abstract and results claim clear signatures of self-similarity and spectrum collapse, but supply no quantitative error bars, statistical significance tests for the collapse, or explicit comparisons to null models such as configuration-model rewirings that preserve the degree sequence. These omissions make it impossible to determine whether the observed patterns exceed what would be expected from sampling biases or platform effects alone.

Authors: We acknowledge that the original presentation relies primarily on visual and qualitative consistency across the three events. To address this, the revision will include error bars on key quantities such as participation maxima and clustering coefficients. We will also add explicit comparisons against configuration-model null networks that preserve the degree sequence, allowing us to assess whether the observed transitions and spectrum collapse exceed expectations from degree heterogeneity alone. revision: yes
Referee: Discussion (interpretation as universal organizing principle): the claim that self-similarity at peak mobilization constitutes a universal principle enabling cross-scale amplification rests on the assumption that the renormalized networks faithfully encode social coordination. The manuscript would need temporal or shuffled baselines demonstrating that the modular-to-nested transitions and participation maxima are specific to the identified peak phases rather than generic features of the data.

Authors: Peak phases were identified a priori from independent indicators of user activity volume and documented event timelines, prior to network construction. In the revision we will add direct temporal comparisons showing that the modular-to-nested transitions and clustering collapse are most pronounced precisely at these peaks and weaken at other phases. While exhaustive shuffled baselines across the full multi-event dataset would require substantial additional computation, the combination of temporal specificity within each event and the replication across three independent movements provides convergent evidence for phase-specificity; we will clarify this distinction in the discussion. revision: partial

Circularity Check

0 steps flagged

No significant circularity in the reported derivation chain.

full rationale

The paper constructs co-occurrence networks from Twitter hashtag data, applies degree-thresholding renormalization, and reports observed self-similar signatures (modular-to-nested transitions, clustering collapse) at peak mobilization times across three events. No equations, definitions, or steps are presented that reduce the self-similarity claim to a fitted parameter renamed as prediction, a self-citation chain, or an ansatz smuggled in by definition. The central findings are framed as empirical patterns extracted from external data, rendering the derivation self-contained.

Axiom & Free-Parameter Ledger

1 free parameters · 2 axioms · 1 invented entities

The central claim rests on the assumption that renormalised co-occurrence networks reveal intrinsic self-similarity rather than methodological artifacts, plus the choice of degree threshold and the definition of peak times.

free parameters (1)

degree threshold for renormalisation
Used to coarse-grain the network; its specific value is not stated in the abstract and likely chosen to produce the reported scale invariance.

axioms (2)

domain assumption Hashtag co-occurrence faithfully proxies social interaction and coordination
Invoked when constructing the networks from Twitter data.
domain assumption Self-similarity is expected at critical points of collective action
Underpins the interpretation of the observed patterns as universal.

invented entities (1)

latent metric structure supporting hyperbolic embedding no independent evidence
purpose: To provide an estimate of effective social distance
Introduced as a hint from the embedding results; no independent falsifiable prediction is given in the abstract.

pith-pipeline@v0.9.0 · 5523 in / 1483 out tokens · 48832 ms · 2026-05-12T01:46:14.211623+00:00 · methodology

Review history (2 revisions) →

discussion (0)

Reference graph

Works this paper leans on

31 extracted references · 31 canonical work pages

[1]

M. D. Conover, E. Ferrara, F. Menczer, and A. Flammini, The digital evolution of occupy wall street, PloS one8, e64679 (2013)

work page 2013
[2]

Alvarez, D

R. Alvarez, D. Garcia, Y. Moreno, and F. Schweitzer, Sentiment cascades in the 15m movement, EPJ Data Sci- ence4, 6 (2015)

work page 2015
[3]

Z. C. Steinert-Threlkeld, D. Mocanu, A. Vespignani, and J. Fowler, Online social networks and offline protest, EPJ Data Science4, 1 (2015)

work page 2015
[4]

Borge-Holthoefer, A

J. Borge-Holthoefer, A. Rivero, I. Garc´ ıa, E. Cauh´ e, A. Ferrer, D. Ferrer, D. Francos, D. Iniguez, M. P. P´ erez, G. Ruiz,et al., Structural and dynamical patterns on on- line social networks: the spanish may 15th movement as a case study, PloS one6, e23883 (2011)

work page 2011
[5]

Esquirol, L

B. Esquirol, L. Prignano, A. D´ ıaz-Guilera, and E. Cozzo, Analyzing user activity on twitter during long-lasting cri- sis events: a case study of the covid-19 crisis in spain, Social Network Analysis and Mining14, 10.1007/s13278- 024-01215-y (2024)

work page doi:10.1007/s13278- 2024
[6]

M. G. Beir´ o, N. N. Chung, L. Y. Chew, and Y. Gandica, Signs of criticality in social explosions, Scientific Reports 14, 10.1038/s41598-024-53657-1 (2024)

work page doi:10.1038/s41598-024-53657-1 2024
[7]

Q. Liu, X. Zhao, W. Willinger, X. Wang, B. Y. Zhao, and H. Zheng, Self-similarity in social network dynam- ics, ACM Trans. Model. Perform. Eval. Comput. Syst.2, 10.1145/2994142 (2016). [8]Quantifying Political Self-Organization in Social Media. Fractal patterns in the Spanish 15M move- ment on Twitter, ALIFE 2022: The 2022 Con- ference on Artificial Life,...

work page doi:10.1145/2994142 2016
[8]

Mandelbrot, How long is the coast of britain? statisti- cal self-similarity and fractional dimension, Science156, 636–638 (1967)

B. Mandelbrot, How long is the coast of britain? statisti- cal self-similarity and fractional dimension, Science156, 636–638 (1967)

work page 1967
[9]

B. B. Mandelbrot,The fractal geometry of nature, up- dated and augmented, 18. print. ed. (Freeman, New York [u.a.], 1999)

work page 1999
[10]

Boguna, I

M. Boguna, I. Bonamassa, M. De Domenico, S. Havlin, D. Krioukov, and M. ´A. Serrano, Network geometry, Na- ture Reviews Physics3, 114 (2021)

work page 2021
[11]

C. Song, S. Havlin, and H. A. Makse, Self-similarity of complex networks, Nature433, 392 (2005)

work page 2005
[12]

C. Song, S. Havlin, and H. A. Makse, Origins of fractality in the growth of complex networks, Nature physics2, 275 (2006)

work page 2006
[13]

Serrano, M

D. Serrano, M. Angelesand Krioukov and M. Bogu˜ na, Self-similarity of complex networks and hidden metric spaces, Physical Review Letters100, 078701 (2008)

work page 2008
[14]

Krioukov, F

D. Krioukov, F. Papadopoulos, M. Kitsak, A. Vahdat, and M. Bogun´ a, Hyperbolic geometry of complex net- works, Physical Review E—Statistical, Nonlinear, and Soft Matter Physics82, 036106 (2010)

work page 2010
[15]

Bogun´ a, F

M. Bogun´ a, F. Papadopoulos, and D. Krioukov, Sustain- ing the internet with hyperbolic mapping, Nature com- munications1, 62 (2010)

work page 2010
[16]

Bl¨ asius, T

T. Bl¨ asius, T. Friedrich, A. Krohmer, and S. Laue, Ef- ficient embedding of scale-free graphs in the hyperbolic plane, IEEE/ACM transactions on Networking26, 920 (2018)

work page 2018
[17]

Garc´ ıa-P´ erez, A

G. Garc´ ıa-P´ erez, A. Allard, M. ´A. Serrano, and M. Bogu˜ n´ a, Mercator: uncovering faithful hyperbolic em- beddings of complex networks, New Journal of Physics 21, 123033 (2019)

work page 2019
[18]

Jankowski, A

R. Jankowski, A. Allard, M. Bogu˜ na, and M. A. Ser- rano, The d-mercator method for the multidimensional hyperbolic embedding of real networks, Nature Commu- nications14, 10.1038/s41467-023-43337-5 (2023)

work page doi:10.1038/s41467-023-43337-5 2023
[19]

Muscoloni, J

A. Muscoloni, J. M. Thomas, S. Ciucci, G. Bianconi, and C. V. Cannistraci, Machine learning meets complex net- works via coalescent embedding in the hyperbolic space, Nature communications8, 1615 (2017)

work page 2017
[20]

Garc´ ıa-P´ erez, M

G. Garc´ ıa-P´ erez, M. Bogu˜ n´ a, and M.´A. Serrano, Multi- scale unfolding of real networks by geometric renormal- ization, Nature Physics14, 583 (2018)

work page 2018
[21]

Gabrielli, D

A. Gabrielli, D. Garlaschelli, S. P. Patil, and M. ´A. Ser- rano, Network renormalization, Nature Reviews Physics , 1 (2025)

work page 2025
[22]

Zheng, A

M. Zheng, A. Allard, P. Hagmann, Y. Alem´ an-G´ omez, and M. ´A. Serrano, Geometric renormalization unravels self-similarity of the multiscale human connectome, Pro- ceedings of the National Academy of Sciences117, 20244 (2020)

work page 2020
[23]

Barjuan, M

L. Barjuan, M. Zheng, and M. ´A. Serrano, The multiscale self-similarity of the weighted human brain connectome, PLOS Computational Biology21, e1012848 (2025)

work page 2025
[24]

Zheng, G

M. Zheng, G. Garc´ ıa-P´ erez, M. Bogu˜ n´ a, and M.´A. Ser- rano, Scaling up real networks by geometric branching growth, Proceedings of the National Academy of Sciences 118, e2018994118 (2021)

work page 2021
[25]

Aliakbarisani, M

R. Aliakbarisani, M. Bogu˜ n´ a, and M.´A. Serrano, Clus- tering does not always imply latent geometry, Physical Review Letters135, 197402 (2025)

work page 2025
[26]

M. S. Granovetter, The strength of weak ties, inSocial Networks(Elsevier, 1977) pp. 347–367

work page 1977
[27]

Rico and Y

D. Rico and Y. Gandica, An entropic analysis of social demonstrations, Entropy26, 363 (2024)

work page 2024
[28]

Hagberg, P

A. Hagberg, P. Swart, and D. S Chult,Exploring network structure, dynamics, and function using NetworkX, Tech. Rep. (Los Alamos National Lab.(LANL), Los Alamos, NM (United States), 2008)

work page 2008
[29]

Newman,Networks(Oxford University Press, 2010)

M. Newman,Networks(Oxford University Press, 2010)

work page 2010
[30]

Almeida-Neto, P

M. Almeida-Neto, P. Guimar˜ aes, P. R. Guimar˜ aes, R. D. Loyola, and W. Ulrich, A consistent metric for nestedness analysis in ecological systems: reconciling concept and measurement, Oikos117, 1227–1239 (2008)

work page 2008
[31]

M. J. Palazzi Nieves, Structural and dynamical interde- pendencies in complex networks at meso- and macroscale: nestedness, modularity, and in-block nestedness, (2020)

work page 2020