pith. sign in

arxiv: 2605.30551 · v1 · pith:RRI2DGPInew · submitted 2026-05-28 · ⚛️ physics.soc-ph

Beyond Direct Retweets: Multi-Step Pathways in Italian COVID-19 Twitter

Edoardo Maggioni , Ren Manfredi , Fabio Saracco , Rossana Mastrandrea This is my paper

Pith reviewed 2026-06-28 23:42 UTC · model grok-4.3

classification ⚛️ physics.soc-ph
keywords retweet networkscommunity detectionhigher-order random walksattention redistributionCOVID-19 TwitterItalysocial mediarandom walks
0
0 comments X

The pith

Retweet attention in Italian COVID-19 Twitter starts concentrated inside communities but spreads unevenly to other groups over longer paths.

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

The paper combines community detection with higher-order random walks on a large Italian Twitter retweet network from the first COVID-19 wave. It treats short multi-step paths as a way to track where attention lands beyond single direct retweets. Attention stays mostly inside the same community at short distances but mixes more as path length grows. The mixing is uneven: certain communities gain share as endpoints while others lose it, patterns not explained by community size or direct links alone. The network also shows clear directional differences when the links are reversed.

Core claim

Motif-based random-walk paths show that attention concentration inside communities declines with path length while the resulting cross-community endpoint distribution is non-uniform, produces community prominence shifts not captured by size or first-order connectivity, and exhibits directional asymmetries under network reversal.

What carries the argument

Motif-based random-walk paths on the retweet network, used as a structural device to compare direct community-to-community connectivity against the distribution of multi-step endpoints.

If this is right

  • Moving from one-step to multi-step analysis changes which communities appear most prominent in the debate.
  • Community size and direct retweet volume do not fully predict longer-path attention sinks.
  • Reversing the network direction produces different endpoint prominence rankings.
  • First-order retweet graphs understate the higher-order mixing present in actual attention flows.

Where Pith is reading between the lines

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

  • The same path-length approach could be tested on other national or topic-specific Twitter datasets to check whether uneven redistribution is general.
  • Models of information diffusion might need to incorporate path-dependent community effects rather than assuming uniform mixing.
  • Monitoring tools for public discourse could add multi-step flow measures to detect emerging community prominence shifts earlier.

Load-bearing premise

Motif-based random-walk paths serve as a valid structural device to compare direct community-to-community connectivity with the distribution of multi-step endpoints and thereby measure attention redistribution.

What would settle it

If the observed distribution of endpoints after k-step paths exactly matches the direct (one-step) connectivity matrix for every k, or if community endpoint shares remain constant across increasing path lengths.

Figures

Figures reproduced from arXiv: 2605.30551 by Edoardo Maggioni, Fabio Saracco, Ren Manfredi, Rossana Mastrandrea.

Figure 1
Figure 1. Figure 1: The eight weighted random-walk motifs defined by Picciolo et al. (2022); the square represents the starting node for the random walker. Each motif corresponds to a distinct pattern of node visits and transitions before absorption. The motifs are mutually exclusive and collectively exhaustive for paths of length up to three steps. Figure adapted from Picciolo et al. (2022). with ℓ ∈ {1,2,3} before absorptio… view at source ↗
Figure 2
Figure 2. Figure 2: Connectivity of the validated projection on verified users. Nodes are colored by community. 973 nodes and 77 communities in total. Communities found with Leiden algorithm and resolution parameter γ = 2. To characterize the detected communities, we qualitatively inspected their most central users, using both degree and strength as indicators of centrality, together with the recurring narratives associated w… view at source ↗
Figure 3
Figure 3. Figure 3: Motif composition by starting community as stacked bars with broken y-axis to expand low-share motifs while retaining the dominant Motif 1 mass. ”G” indicates the global motif distribution (all paths). The top 10 communities are ordered by size rank. ”0” aggregates all remaining communities. To further characterize how motif compositions differ across communities, we consider the motif profile of each star… view at source ↗
Figure 4
Figure 4. Figure 4: Community similarity in the attention network (original orientation) based on Jensen–Shannon distances between motif-profile distributions (computed by starting community). 3.4 Between and Within-Community Direct Connectivity We begin by describing the first-order structure of direct retweet interactions at the community level, which serves as a reference point for the higher-order analyses that follow [P… view at source ↗
Figure 5
Figure 5. Figure 5: summarizes the first-order inter-community structure. The heatmap shows the raw matrix of direct retweet counts, while the barplots report the distribution of incoming and outgoing attention shares across communities. (a) Raw inter-community connectivity (counts). 1 2 3 4 5 6 7 8 9 10 0 community 0.00 0.05 0.10 0.15 0.20 0.25 Incoming attention Size-based attention share (b) Incoming attention share σ link… view at source ↗
Figure 6
Figure 6. Figure 6: a shows that persistence is highly concentrated, with community 2 capturing the largest share of within-community attention. 1 2 3 4 5 6 7 8 9 10 0 community 0.0 0.1 0.2 0.3 Within-community attention Size-based attention share (a) Within-community persistence, σ links within,c . 1 2 3 4 5 6 7 8 9 10 0 community 0.00 0.05 0.10 0.15 0.20 0.25 Size-based attention share (b) Size-based reference, σ size withi… view at source ↗
Figure 7
Figure 7. Figure 7: Community-level final-jump matrices (raw counts) for Motifs 1, 3, and 8. Each cell (c,c ′ ) represents the number of paths starting in community c and ending in community c ′ . Diagonal entries capture within-community persistence, while off-diagonal entries represent cross-community pathways. Color scales are logarithmic and defined separately for each motif. Across motifs, the matrices show a consistent … view at source ↗
Figure 8
Figure 8. Figure 8: reports the motif-specific within-community persistence share, that is, the share of realized paths starting in community c that also end in the same community (c → c), for Motifs 1, 3, and 8. 1 2 3 4 5 6 7 8 9 10 0 Community 0.0 0.2 0.4 0.6 0.8 paths,(m) within, c Within-community persistence across motif length Motif 1 Motif 3 Motif 8 [PITH_FULL_IMAGE:figures/full_fig_p021_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Observed versus first-order reference within-community persistence under Wlinks. Each point represents a community. The x-axis reports the first-order persistence share σ links within,c , while the y-axis reports the corresponding persistence share derived from paths σ paths,(m) within,c . Both axes are shown on logarithmic scales. The dashed line indicates equality between observed and reference values. T… view at source ↗
Figure 10
Figure 10. Figure 10: Reference-adjusted within-community persistence across motif lengths. Lines report log ratiolinks,(m) within,c for Motifs 1, 3, and 8. Negative values indicate lower persistence relative to the first-order reference. The downward trend highlights the progressive weakening of within-community retention as path length increases. Across communities, log ratiolinks,(m) within,c systematically decreases as mot… view at source ↗
Figure 11
Figure 11. Figure 11: Observed versus first-order reference share of received cross-community attention under Wlinks (final-jump, attention orientation). Each point is a community; panels correspond to Motif 1, Motif 3, and Motif 8. The dashed line indicates equality between observed and reference shares. Motif 1 Motif 3 Motif 8 Motif 0.8 0.6 0.4 0.2 0.0 0.2 0.4 0.6 Log-ratio of incoming attention (links baseline) (8 1) > 0 Co… view at source ↗
Figure 12
Figure 12. Figure 12: Reference deviations across motif length under Wlinks (final-jump, attention orientation). Lines show log ratiolinks,(m) c from Motif 1 to Motif 8. Left panel: communities with ∆ (8−1) c > 0 (increasing deviation). Right panel: communities with ∆ (8−1) c < 0 (decreasing deviation). The dashed horizontal line marks 0 (no deviation from the reference). The left panel shows communities with positive ∆ (8−1) … view at source ↗
Figure 13
Figure 13. Figure 13: reports the evolution of log ratiolinks,(m) c across Motifs 1, 3, and 8 under the reversed orientation, splitting communities according to the sign of ∆ (8−1) c . As in the attention case, longer pathways induce systematic divergence, with some communities becoming increasingly over-represented as endpoints of multi-step pathways while others are progressively under-represented. Motif 1 Motif 3 Motif 8 Mo… view at source ↗
Figure 14
Figure 14. Figure 14: Comparison of motif-length sensitivity across orientations. Each point is a community. The x-axis reports ∆ (8−1) c under the attention orientation and the y-axis reports ∆ (8−1) c under the reversed orientation. Dashed lines mark 0. Most communities lie in the same-sign quadrants, indicating that the qualitative direction of their higher-order behavior is broadly preserved across orientations. However, t… view at source ↗
Figure 15
Figure 15. Figure 15: Motif probabilities by starting community, with motifs ordered by global abundance and y-axis on log scale. Each line corresponds to one starting community (G, 1–10, 0). G 1 2 3 4 5 6 7 8 9 10 0 Starting Community 0.0 0.2 0.4 0.6 0.8 1.0 Motif Share in Starting Community Motif Motif 1 Motif 3 Motif 8 Motif 7 Motif 4 Motif 6 Motif 2 Motif 5 [PITH_FULL_IMAGE:figures/full_fig_p030_15.png] view at source ↗
Figure 16
Figure 16. Figure 16: Motif composition by starting community as stacked bars (linear scale), with motifs ordered by global abundance. 30/41 [PITH_FULL_IMAGE:figures/full_fig_p030_16.png] view at source ↗
Figure 17
Figure 17. Figure 17: Observed versus size-based baseline within-community persistence. Each point represents a community. The x-axis reports the size-based reference σ size within,c , while the y-axis reports the corresponding persistence share derived from paths σ paths,(m) within,c . Both axes are shown on logarithmic scales. The dashed line indicates equality between observed and baseline values. Motif 1 Motif 3 Motif 8 Mo… view at source ↗
Figure 18
Figure 18. Figure 18: Benchmark-adjusted within-community persistence across motif length under the size-only baseline. Lines report log ratio(m) within,c,size for Motifs 1, 3, and 8. Negative values indicate lower persistence relative to the size-based reference. 31/41 [PITH_FULL_IMAGE:figures/full_fig_p031_18.png] view at source ↗
Figure 19
Figure 19. Figure 19: Community-level log-ratio of within-community persistence under the Wlinks baseline. Bars report log σ paths,(m) within,c /σ links within,c  for Motifs 1, 3, and 8. 1 2 3 4 5 6 7 8 9 10 0 Community 4.0 3.5 3.0 2.5 2.0 1.5 1.0 0.5 0.0 lo g( paths,(m) within, c / size within, c) Within-community persistence (size baseline) Motif 1 Motif 3 Motif 8 [PITH_FULL_IMAGE:figures/full_fig_p032_19.png] view at source ↗
Figure 20
Figure 20. Figure 20: Community-level log-ratio of within-community persistence under the size-only baseline. Bars report log σ paths,(m) within,c /σ size within,c  for Motifs 1, 3, and 8. 1 2 3 4 5 6 7 8 9 10 0 Community 0.00 0.05 0.10 0.15 0.20 0.25 paths,(m) in, c Incoming attention share across motif length Motif 1 Motif 3 Motif 8 [PITH_FULL_IMAGE:figures/full_fig_p032_20.png] view at source ↗
Figure 21
Figure 21. Figure 21: Distribution of incoming attention across communities. Bars report the share of paths σ paths,(m) in,c for Motifs 1, 3, and 8. 32/41 [PITH_FULL_IMAGE:figures/full_fig_p032_21.png] view at source ↗
Figure 22
Figure 22. Figure 22: Community-level log-ratio of incoming attention under the Wlinks baseline. Bars report log σ paths,(m) in,c /σ links in,c  for Motifs 1, 3, and 8. 1 2 3 4 5 6 7 8 9 10 0 Community 1.0 0.5 0.0 0.5 1.0 lo g( paths,(m) in, c / size in, c) Incoming attention (size baseline) Motif 1 Motif 3 Motif 8 [PITH_FULL_IMAGE:figures/full_fig_p033_22.png] view at source ↗
Figure 23
Figure 23. Figure 23: Community-level log-ratio of incoming attention under the size-only baseline. Bars report log σ paths,(m) in,c /σ size in,c  for Motifs 1, 3, and 8. 10 2 10 1 Baseline incoming attention (size) 10 2 10 1 Observed incoming attention (paths) 1 3 2 4 5 6 7 8 109 0 Motif 1 10 2 10 1 Baseline incoming attention (size) 1 2 3 4 5 7 6 8 9 10 0 Motif 3 10 2 10 1 Baseline incoming attention (size) 1 2 3 4 5 6 7 8… view at source ↗
Figure 24
Figure 24. Figure 24: Observed versus baseline share of received cross-community attention under the Wsize benchmark (final-jump, attention orientation). Each point is a community; panels correspond to Motif 1, Motif 3, and Motif 8. The dashed line indicates equality between observed and baseline shares. 33/41 [PITH_FULL_IMAGE:figures/full_fig_p033_24.png] view at source ↗
Figure 25
Figure 25. Figure 25: Baseline deviations across motif length under the Wsize benchmark (final-jump, attention orientation). Lines show log ratiosize,(m) c from Motif 1 to Motif 8. Left panel: communities with ∆ size,(8−1) c > 0 (increasing deviation). Right panel: communities with ∆ size,(8−1) c < 0 (decreasing deviation). The dashed horizontal line marks 0 (no deviation from the baseline). 34/41 [PITH_FULL_IMAGE:figures/ful… view at source ↗
Figure 26
Figure 26. Figure 26: Community-level final-jump matrices (raw counts) for Motifs 1, 3, and 8 under the reversed orientation. 1 2 3 4 5 6 7 8 9 10 0 Community 0.0 0.2 0.4 0.6 0.8 paths,(m) within, c Within-community persistence across motif length Motif 1 Motif 3 Motif 8 [PITH_FULL_IMAGE:figures/full_fig_p036_26.png] view at source ↗
Figure 27
Figure 27. Figure 27: Within-community persistence across motif length under the reversed orientation. 10 3 10 2 10 1 Baseline within-community persistence (links) 10 3 10 2 10 1 Observed within-community persistence (paths) 1 2 3 4 5 6 78 9 10 0 Motif 1 10 3 10 2 10 1 Baseline within-community persistence (links) 1 2 3 4 5 6 7 8 9 10 0 Motif 3 10 3 10 2 10 1 Baseline within-community persistence (links) 1 2 3 4 5 6 7 8 9 10 0… view at source ↗
Figure 28
Figure 28. Figure 28: Observed versus baseline within-community persistence under the Wlinks benchmark (reversed orientation). E DISTRIBUTIONS 36/41 [PITH_FULL_IMAGE:figures/full_fig_p036_28.png] view at source ↗
Figure 29
Figure 29. Figure 29: Benchmark-adjusted within-community persistence across motif length under the reversed orientation. 10 2 10 1 Baseline incoming attention (links) 10 2 10 1 Observed incoming attention (paths) 1 2 3 4 5 6 7 8 9 10 0 Motif 1 10 2 10 1 Baseline incoming attention (links) 1 2 3 4 5 6 7 8 9 10 0 Motif 3 10 2 10 1 Baseline incoming attention (links) 1 2 3 4 5 6 7 8 9 10 0 Motif 8 [PITH_FULL_IMAGE:figures/full_… view at source ↗
Figure 30
Figure 30. Figure 30: Observed versus baseline share of received cross-community attention under the Wlinks benchmark (reversed orientation). (a) In-degree distribution. (b) Out-degree distribution [PITH_FULL_IMAGE:figures/full_fig_p037_30.png] view at source ↗
Figure 31
Figure 31. Figure 31: In- and out-degree distributions in the largest weakly connected component (LWCC) of the global retweet network. Both distributions are heavy-tailed, with a small fraction of users having very high degree. 37/41 [PITH_FULL_IMAGE:figures/full_fig_p037_31.png] view at source ↗
Figure 32
Figure 32. Figure 32: In- and out-strength distributions in the largest weakly connected component (LWCC) of the global retweet network. Both distributions are heavy-tailed, with a small fraction of users having very high degree. 38/41 [PITH_FULL_IMAGE:figures/full_fig_p038_32.png] view at source ↗
read the original abstract

We study how retweet interactions in large-scale Twitter debates are organized beyond direct links alone. Focusing on Twitter debate in Italy during the first phase of the COVID-19 pandemic, we combine a validated community-reconstruction pipeline with a higher-order random-walk framework to examine how short multi-step pathways redistribute attention across discursive communities. Rather than reconstructing observed cascades of individual tweets, we use motif-based random-walk paths as a structural device to compare direct community-to-community connectivity with the distribution of multi-step endpoints. We find that attention is initially concentrated within communities, but that this concentration weakens as path length increases. At the same time, the resulting cross-community redistribution is not uniform: some communities become increasingly prominent as endpoints of longer pathways, while others lose relative prominence. These differences are not fully captured by community size or by first-order retweet connectivity alone, and they also display important directional asymmetries when the network is analyzed under the reversed orientation. Taken together, the results show that moving beyond direct retweets changes the community-level representation of online debate and reveals higher-order structural patterns that remain invisible in first-order analyses.

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

2 major / 1 minor

Summary. The paper analyzes retweet networks from Italian COVID-19 Twitter discussions using community detection followed by motif-based random walks on the community graph. It claims that intra-community attention concentration weakens with increasing path length, while cross-community redistribution is non-uniform (some communities gain endpoint prominence, others lose it), independent of community size or first-order connectivity, and exhibits directional asymmetries under network reversal. The analysis relies on random-walk paths as a structural proxy rather than reconstructed cascades.

Significance. If the random-walk proxy is shown to faithfully capture multi-step attention flow, the work would demonstrate that higher-order network structure produces community-level representation shifts invisible in direct-retweet analyses, with potential implications for modeling information diffusion and polarization in crisis-related online debates.

major comments (2)
  1. [Abstract / Methods] The central claim that multi-step pathways produce non-uniform, size-independent redistribution with directional asymmetries rests on motif-based random walks serving as a faithful proxy for attention flow. The abstract states this choice is made instead of reconstructing observed cascades, yet no comparison (e.g., community-transition matrices, endpoint distributions, or Kolmogorov-Smirnov statistics) between walk-generated paths and any measured multi-step retweet chains is reported; without such grounding the reported weakening of intra-community concentration and prominence shifts risk being method artifacts.
  2. [Abstract / Results] No quantitative results, error bars, dataset sizes, or validation metrics appear in the abstract or are referenced in the provided description, making it impossible to assess whether the data support the stated patterns of non-uniform redistribution; the soundness assessment is therefore limited to 3.0 pending explicit reporting of these quantities in the full manuscript.
minor comments (1)
  1. [Methods] Clarify the precise definition of 'motif-based random-walk paths' (e.g., which motifs, how walks are sampled, and the stopping criterion for path length) to allow reproducibility.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive feedback on our manuscript. We respond to each major comment below, focusing on the methodological rationale and reporting.

read point-by-point responses

  1. Referee: [Abstract / Methods] The central claim that multi-step pathways produce non-uniform, size-independent redistribution with directional asymmetries rests on motif-based random walks serving as a faithful proxy for attention flow. The abstract states this choice is made instead of reconstructing observed cascades, yet no comparison (e.g., community-transition matrices, endpoint distributions, or Kolmogorov-Smirnov statistics) between walk-generated paths and any measured multi-step retweet chains is reported; without such grounding the reported weakening of intra-community concentration and prominence shifts risk being method artifacts.

    Authors: The abstract explicitly frames the motif-based random walks as a 'structural device' for comparing direct connectivity with multi-step endpoint distributions on the community graph, rather than as a model intended to reconstruct or faithfully replicate observed attention flows or cascades. This is a deliberate choice to isolate higher-order topological effects without depending on incomplete cascade data. No empirical comparison to measured multi-step chains is reported because the analysis targets structural patterns, not validated diffusion paths; we therefore do not view the results as method artifacts within the stated scope. revision: no

  2. Referee: [Abstract / Results] No quantitative results, error bars, dataset sizes, or validation metrics appear in the abstract or are referenced in the provided description, making it impossible to assess whether the data support the stated patterns of non-uniform redistribution; the soundness assessment is therefore limited to 3.0 pending explicit reporting of these quantities in the full manuscript.

    Authors: The full manuscript reports dataset sizes, community counts, transition statistics, and related metrics in the Results and Methods sections. To address the concern, we will revise the abstract to include key quantitative indicators such as the number of tweets analyzed and main effect sizes. revision: yes

Circularity Check

0 steps flagged

No circularity: results derived from direct application to observed data

full rationale

The paper applies a community-reconstruction pipeline followed by motif-based random-walk paths to retweet interaction data from the Italian COVID-19 Twitter debate. No equations or steps reduce by construction to fitted parameters that are then renamed as predictions, nor do any load-bearing claims rest on self-citations whose content is unverified within the paper. The central findings on weakening intra-community concentration and non-uniform redistribution with path length are obtained by computing endpoint distributions on the empirical network; the method is a structural proxy applied once to the data rather than a self-referential loop. This matches the default expectation of a self-contained empirical analysis.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

Abstract-only review; ledger reflects only elements explicitly named in the abstract.

axioms (2)
  • domain assumption The community-reconstruction pipeline is validated
    Abstract states 'validated community-reconstruction pipeline' without further detail.
  • domain assumption Motif-based random-walk paths accurately represent multi-step retweet interactions for attention redistribution
    Abstract uses them 'as a structural device to compare direct community-to-community connectivity with the distribution of multi-step endpoints'.

pith-pipeline@v0.9.1-grok · 5732 in / 1269 out tokens · 29656 ms · 2026-06-28T23:42:15.803807+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

60 extracted references · 2 canonical work pages

  1. [1]

    Badawy, A., Ferrara, E., and Lerman, K. (2018). Analyzing the digital traces of political manipulation: The 2016 russian interference twitter campaign. In 2018 IEEE/ACM international conference on advances in social networks analysis and mining (ASONAM) , pages 258--265. IEEE

    2018

  2. [2]

    A., Borge-Holthoefer, J., and Moreno, Y

    Baños, R. A., Borge-Holthoefer, J., and Moreno, Y. (2013). The role of hidden influentials in the diffusion of online information cascades. EPJ Data Science , 2(1)

    2013

  3. [3]

    Becatti, C., Caldarelli, G., Lambiotte, R., and Saracco, F. (2019). Extracting significant signal of news consumption from social networks: the case of twitter in italian political elections. Palgrave Communications , 5(1)

    2019

  4. [4]

    Bennett, W. L. and Manheim, J. B. (2006). The one-step flow of communication. The ANNALS of the American Academy of Political and Social Science , 608(1):213--232

    2006

  5. [5]

    R., Gleich, D

    Benson, A. R., Gleich, D. F., and Leskovec, J. (2016). Higher-order organization of complex networks. Science , 353(6295):163--166

    2016

  6. [6]

    D., Guillaume, J.-L., Lambiotte, R., and Lefebvre, E

    Blondel, V. D., Guillaume, J.-L., Lambiotte, R., and Lefebvre, E. (2008). Fast unfolding of communities in large networks. Journal of statistical mechanics: theory and experiment , 2008(10):P10008

    2008

  7. [7]

    Bracciale, R., Martella, A., and Visentin, C. (2018). From super-participants to super-echoed. participation in the 2018 italian electoral twittersphere. PaCO Vol. 11, No. 2 (2018). Special Issue: From Big Data in Politics to the Politics of Big Data , 11:361--393

    2018

  8. [8]

    Caldarelli, G., Artime, O., Fischetti, G., Guarino, S., Nowak, A., Saracco, F., Holme, P., and de Domenico, M. (2025). The physics of news, rumors, and opinions. arXiv preprint arXiv:2510.15053

    work page arXiv 2025

  9. [9]

    Caldarelli, G., De Nicola, R., Petrocchi, M., Pratelli, M., and Saracco, F. (2021). Flow of online misinformation during the peak of the covid-19 pandemic in italy. EPJ data science , 10(1):34

    2021

  10. [10]

    Chang, X., Cai, C.-R., Zhang, J.-Q., and Wang, C.-Y. (2021). Analytical solution of epidemic threshold for coupled information-epidemic dynamics on multiplex networks with alterable heterogeneity. Physical Review E , 104(4):044303

    2021

  11. [11]

    Cimini, G., Squartini, T., Saracco, F., Garlaschelli, D., Gabrielli, A., and Caldarelli, G. (2019). The statistical physics of real-world networks. Nature Reviews Physics , 1(1):58--71

    2019

  12. [12]

    Cinelli, M., De Francisci Morales, G., Galeazzi, A., Quattrociocchi, W., and Starnini, M. (2021). The echo chamber effect on social media. Proceedings of the national academy of sciences , 118(9):e2023301118

    2021

  13. [13]

    Conover, M., Ratkiewicz, J., Francisco, M., Gon c alves, B., Menczer, F., and Flammini, A. (2011). Political polarization on twitter. In Proceedings of the international aaai conference on web and social media , volume 5, pages 89--96

    2011

  14. [14]

    D., Gon c alves, B., Flammini, A., and Menczer, F

    Conover, M. D., Gon c alves, B., Flammini, A., and Menczer, F. (2012). Partisan asymmetries in online political activity. EPJ Data science , 1(1):6

    2012

  15. [15]

    E., and Quattrociocchi, W

    Del Vicario, M., Bessi, A., Zollo, F., Petroni, F., Scala, A., Caldarelli, G., Stanley, H. E., and Quattrociocchi, W. (2016). The spreading of misinformation online. Proceedings of the national academy of Sciences , 113(3):554--559

    2016

  16. [16]

    Dubois, E., Minaeian, S., Paquet-Labelle, A., and Beaudry, S. (2020). Who to trust on social media: How opinion leaders and seekers avoid disinformation and echo chambers. Social Media + Society , 6(2):2056305120913993

    2020

  17. [17]

    C., Peixoto, T

    Fischer, R., Leit\ ao, J. C., Peixoto, T. P., and Altmann, E. G. (2015). Sampling motif-constrained ensembles of networks. Phys. Rev. Lett. , 115:188701

    2015

  18. [18]

    W., Cross, B., Zhou, Z., Serafino, M., Bovet, A., Makse, H

    Flamino, J., Galeazzi, A., Feldman, S., Macy, M. W., Cross, B., Zhou, Z., Serafino, M., Bovet, A., Makse, H. A., and Szymanski, B. K. (2023). Political polarization of news media and influencers on Twitter in the 2016 and 2020 US presidential elections . Nature Human Behaviour , 7(6):904--916

    2023

  19. [19]

    Fortunato, S. (2010). Community detection in graphs. Physics reports , 486(3-5):75--174

    2010

  20. [20]

    and Loffredo, M

    Garlaschelli, D. and Loffredo, M. I. (2008). Maximum likelihood: Extracting unbiased information from complex networks. Phys. Rev. E , 78:015101

    2008

  21. [21]

    Garrett, R. K. (2009). Echo chambers online?: Politically motivated selective exposure among internet news users. Journal of computer-mediated communication , 14(2):265--285

    2009

  22. [22]

    Goel, S., Anderson, A., Hofman, J., and Watts, D. J. (2016). The structural virality of online diffusion. Management science , 62(1):180--196

    2016

  23. [23]

    Gomez Rodriguez, M., Leskovec, J., and Krause, A. (2010). Inferring networks of diffusion and influence. In Proceedings of the 16th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining , KDD '10, page 1019–1028, New York, NY, USA. Association for Computing Machinery

    2010

  24. [24]

    Gong, X., Huskey, R., Xue, H., Shen, C., and Frey, S. (2023). Broadcast information diffusion processes on social media networks: Exogenous events lead to more integrated public discourse. Journal of Communication , 73(3):247--259

    2023

  25. [25]

    González-Bailón, S., Borge-Holthoefer, J., and Moreno, Y. (2013). Broadcasters and hidden influentials in online protest diffusion. American Behavioral Scientist

    2013

  26. [26]

    Guarino, S., Mounim, A., Caldarelli, G., and Saracco, F. (2026). Leveraging content producer networks and user perception to detect online discursive communities. Scientific Reports

    2026

  27. [27]

    Guarino, S., Pierri, F., Di Giovanni, M., and Celestini, A. (2021). Information disorders during the covid-19 infodemic: The case of italian facebook. Online Social Networks and Media , 22:100124

    2021

  28. [28]

    Hanna, A., Wells, C., Maurer, P., Friedland, L., Shah, D., and Matthes, J. (2013). Partisan alignments and political polarization online: A computational approach to understanding the french and us presidential elections. In Proceedings of the 2nd Workshop on Politics, Elections and Data , pages 15--22

    2013

  29. [29]

    Hilbert, M., Vásquez, J., Halpern, D., Valenzuela, S., and Arriagada, E. (2017). One step, two step, network step? complementary perspectives on communication flows in twittered citizen protests. Social Science Computer Review , 35(4):444--461

    2017

  30. [30]

    and Lazarsfeld, P

    Katz, E. and Lazarsfeld, P. F. (1955). Personal influence: the part played by people in the flow of mass communications. Free Press

    1955

  31. [31]

    Lambiotte, R., Rosvall, M., and Scholtes, I. (2019). From networks to optimal higher-order models of complex systems. Nature physics , 15(4):313--320

    2019

  32. [32]

    and Fortunato, S

    Lancichinetti, A. and Fortunato, S. (2009). Community detection algorithms: a comparative analysis. Physical review E , 80(5):056117

    2009

  33. [33]

    LaRock, T., Scholtes, I., and Eliassi-Rad, T. (2022). Sequential motifs in observed walks. Journal of Complex Networks , 10(5):cnac036

    2022

  34. [34]

    Mastrandrea, R., Squartini, T., Fagiolo, G., and Garlaschelli, D. (2014). Enhanced reconstruction of weighted networks from strengths and degrees. New Journal of Physics , 16(4):043022

    2014

  35. [35]

    Mattei, M., Pratelli, M., Caldarelli, G., Petrocchi, M., and Saracco, F. (2022). Bow-tie structures of twitter discursive communities. sci rep 12 (1): 12944

    2022

  36. [36]

    Newman, M. (2018). Networks . Oxford university press

    2018

  37. [37]

    O'Sullivan, D. J. P., O'Keeffe, G. J., Fennell, P. G., and Gleeson, J. P. (2015). Mathematical modeling of complex contagion on clustered networks. Frontiers in Physics , Volume 3 - 2015

    2015

  38. [38]

    Parisi, F., Squartini, T., and Garlaschelli, D. (2020). A faster horse on a safer trail: generalized inference for the efficient reconstruction of weighted networks. New Journal of Physics , 22(5):053053

    2020

  39. [39]

    Paul, I., Khattar, A., Kumaraguru, P., Gupta, M., and Chopra, S. (2019). Elites tweet? characterizing the twitter verified user network. In 2019 IEEE 35th International Conference on Data Engineering Workshops (ICDEW) , pages 278--285

    2019

  40. [40]

    P., Peel , L., Gross , T., and De Domenico , M

    Peixoto , T. P., Peel , L., Gross , T., and De Domenico , M. (2026). Graphs are maximally expressive for higher-order interactions . arXiv e-prints , page arXiv:2602.16937

    work page arXiv 2026

  41. [41]

    Peixoto, T. P. and Rosvall, M. (2017). Modelling sequences and temporal networks with dynamic community structures. Nature communications , 8(1):582

    2017

  42. [42]

    Picciolo, F., Ruzzenenti, F., Holme, P., and Mastrandrea, R. (2022). Weighted network motifs as random walk patterns. New journal of physics , 24(5):053056

    2022

  43. [43]

    Pratelli, M., Saracco, F., and Petrocchi, M. (2024). Entropy-based detection of twitter echo chambers. PNAS nexus , 3(5):pgae177

    2024

  44. [44]

    Radicioni, T., Saracco, F., Pavan, E., and Squartini, T. (2021). Analysing twitter semantic networks: the case of 2018 italian elections. Scientific Reports , 11(1):13207

    2021

  45. [45]

    N., Albert, R., and Kumara, S

    Raghavan, U. N., Albert, R., and Kumara, S. (2007). Near linear time algorithm to detect community structures in large-scale networks. Physical Review E—Statistical, Nonlinear, and Soft Matter Physics , 76(3):036106

    2007

  46. [46]

    Riascos, A. P. and Mateos, J. L. (2021). Random walks on weighted networks: a survey of local and non-local dynamics. Journal of Complex Networks , 9(5):cnab032

    2021

  47. [47]

    T., and Lambiotte, R

    Salnikov, V., Schaub, M. T., and Lambiotte, R. (2016). Using higher-order markov models to reveal flow-based communities in networks. Scientific reports , 6(1):23194

    2016

  48. [48]

    Saracco, F., Di Clemente, R., Gabrielli, A., and Squartini, T. (2015). Randomizing bipartite networks: the case of the world trade web. Scientific reports , 5(1):10595

    2015

  49. [49]

    J., Clemente, R

    Saracco, F., Straka, M. J., Clemente, R. D., Gabrielli, A., Caldarelli, G., and Squartini, T. (2017). Inferring monopartite projections of bipartite networks: an entropy-based approach. New Journal of Physics , 19(5):053022

    2017

  50. [50]

    Scholtes, I. (2017). When is a network a network? multi-order graphical model selection in pathways and temporal networks. In Proceedings of the 23rd ACM SIGKDD international conference on knowledge discovery and data mining , pages 1037--1046

    2017

  51. [51]

    Scholtes, I., Wider, N., and Garas, A. (2016). Higher-order aggregate networks in the analysis of temporal networks: path structures and centralities. The European Physical Journal B , 89(3):61

    2016

  52. [52]

    K., Lizardo, O., and Makse, H

    Serafino, M., Virginio Clemente, G., Flamino, J., Szymanski, B. K., Lizardo, O., and Makse, H. A. (2024). Analysis of flows in social media uncovers a new multi-step model of information spread. Journal of Statistical Mechanics: Theory and Experiment , 2024(11):113402

    2024

  53. [53]

    S., Srivastava, D., Verma, M., and Muhuri, S

    Singh, S. S., Srivastava, D., Verma, M., and Muhuri, S. (2025). Information diffusion analysis: process, model, deployment, and application. The Knowledge Engineering Review , 39:e11

    2025

  54. [54]

    Sobkowicz, P. (2018). Opinion dynamics model based on cognitive biases of complex agents. Journal of Artificial Societies and Social Simulation (JASSS) , 21(4):8

    2018

  55. [55]

    A., Waltman, L., and Van Eck, N

    Traag, V. A., Waltman, L., and Van Eck, N. J. (2019). From louvain to leiden: guaranteeing well-connected communities. Scientific reports , 9(1):5233

    2019

  56. [56]

    Vallarano, N., Bruno, M., Marchese, E., Trapani, G., Saracco, F., Cimini, G., Zanon, M., and Squartini, T. (2021). Fast and scalable likelihood maximization for exponential random graph models with local constraints. Scientific Reports 2021 11:1 , 11:1--33

    2021

  57. [57]

    Vosoughi, S., Roy, D., and Aral, S. (2018). The spread of true and false news online. science , 359(6380):1146--1151

    2018

  58. [58]

    Watts, D. J. and Dodds, P. S. (2007). Influentials, networks, and public opinion formation . Journal of Consumer Research , 34(4):441--458

    2007

  59. [59]

    M., Mason, W

    Wu, S., Hofman, J. M., Mason, W. A., and Watts, D. J. (2011). Who says what to whom on twitter. In Proceedings of the 20th International Conference on World Wide Web , WWW '11, page 705–714, New York, NY, USA. Association for Computing Machinery

    2011

  60. [60]

    Zhao, Y., Bai, W., Qiao, T., and Wang, W. (2025). Modularity of online social networks acts as a reliable predictor of both whole-network and ego-network characteristics over time. Humanities and Social Sciences Communications , 12(1):1--10

    2025