Pith. sign in

REVIEW 5 major objections 6 minor 27 references

Analysing Health Misinformation with Advanced Centrality Metrics in Online Social Networks

T0 review · 5 major / 6 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read The paper claims that adding three purpose-built centrality scores to the usual four network measures expands the detected influencer set by 44.8% and lifts simulated misinformation reduction from 50% to 62.5%.

desk verdict The paper's quantitative claims about new centrality metrics rest on a random vulnerability init and an undescribed intervention simulation; the qualitative comparison is worth a look. read the letter →

arxiv 2507.09055 v1 pith:QGYTFAHS submitted 2025-07-11 cs.SI cs.IRphysics.soc-ph

Mkululi Sikosana , Sean Maudsley-Barton , Oluwaseun Ajao This is my paper · ORCID
classification cs.SIcs.IRphysics.soc-ph
keywords centralitymetricshealthmisinformationonlinesocialnetworksdynamicinfluencevulnerabilitypropagationnetworkinterventionCOVID-19
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

This paper argues that standard centrality scores (degree, eigenvector, betweenness, closeness) see only the static shape of a network and miss two things that matter for health misinformation: who is susceptible to false claims, and whose influence builds up over time. To fill that gap it introduces propagation centrality (PC), misinformation vulnerability centrality (MVC), and dynamic influence centrality (DIC). On a COVID-19 misinformation network, the new metrics take the combined top-influencer pool from 29 to 42 nodes, a 44.8% increase, and in simulated node-removal interventions they lift the reduction in misinformation volume from 50% to 62.5%. The same metrics were run on a second, broader medical-misinformation dataset and again surfaced influential actors that structural measures alone missed. If the results hold, a layered mix of static and dynamic, susceptibility-aware scores gives platforms and public-health agencies a more complete map of who starts, amplifies, and sustains false health narratives.

What carries the argument

The machinery is a triad of purpose-built scores, each defined against a directed interaction graph. Propagation centrality (PC) is an iterative PageRank-style diffusion rule: a node inherits importance from the out-degree-weighted importance of its in-neighbours, with a damping factor of 0.85, converging to a steady-state propagation potential. Misinformation vulnerability centrality (MVC) starts each node with a vulnerability draw from a uniform distribution between 0 and 1 and iteratively multiplies it by in-degree, so a node ranks high only when high exposure and high initial susceptibility coincide. Dynamic influence centrality (DIC) initialises every node at 1 and at each timestep adds the current scores of its in-neighbours, letting influence compound over roughly ten steps before normalisation. On top of these, a comparison protocol (union of the top-10 lists from each metric versus the traditional four, plus simulated node-removal interventions) turns the individual scores into the claim about complementary coverage.

What would settle it

Re-run the MVC calculation on the FibVID graph with a different random seed, or with vulnerability scores replaced by observable user features such as retweet-without-fact-check counts; if nodes 101358, 72378, and 130371 no longer rank in the top ten, the MVC-specific coverage gains are seed artifacts rather than network structure. Alternatively, remove only those three nodes in the simulation and check whether the misinformation-volume reduction still exceeds 50%.

Watch

Extended reading notes

Core claim

On the paper's terms, the central discovery is that influence in health-misinformation networks has at least three separable components that traditional centrality cannot see: diffusion potential, susceptibility, and temporal persistence. Propagation centrality (PC), implemented as a PageRank-style diffusion kernel, mostly agrees with degree and eigenvector rankings (90% overlap) and therefore anchors the new metrics to structural reality. Misinformation vulnerability centrality (MVC) multiplies each user's exposure by a vulnerability score, surfacing three nodes (101358, 72378, 130371) that no traditional top-10 list contains and raising the count of highly vulnerable nodes from seven to ten. Dynamic influence centrality (DIC) accumulates influence over repeated time steps and returns a top ten entirely disjoint from every static list, identifying 'long-tail' spreaders who keep reigniting rumours. The paper further reports that removing the traditional top nodes in simulation cuts misinformation volume by 50%, while also neutralising the nodes flagged by the three new metrics cuts it by 62.5%, and that the same metric family replicates on the Monant Medical Misinformation dataset, where 247 top influencers were unique to the advanced metrics.

Load-bearing premise

The load-bearing premise is that the vulnerability scores fed into MVC (random numbers between 0 and 1 chosen with a fixed seed instead of measured from user behaviour) really represent how susceptible people are to health misinformation; if they do not, the three nodes MVC newly flags and its share of the reported intervention gain are artifacts of the random draw.

Editorial extensions

If this is right

  • Platforms can use PC-ranked accounts as priority targets for early fact-checking and debunking because high-PC users can seed long-range cascades before corrections take hold.
  • MVC-identified users, being both reachable and susceptible, are the most cost-effective targets for media-literacy prompts and corrective content inserted into their feeds.
  • DIC shows that one-off hub removal is insufficient; continuous monitoring and phased counter-messaging are needed to handle 'long-tail' spreaders who reignite rumours after debunks fade.
  • A layered defence (traditional metrics for rapid triage of hubs, PC for high-throughput spreaders, MVC for vulnerable amplifiers, DIC for long-term surveillance) is more effective than any single metric.
  • Because the advanced metrics replicate on the Monant Medical Misinformation dataset, the combined framework should transfer beyond COVID-19 to broader health-misinformation topics such as vaccine hesitancy and alternative-medicine claims.

Reading between the lines

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

  • A natural next test is to replace MVC's random vulnerability draws with observable proxies such as retweet-without-fact-check rates, prior engagement with flagged posts, and account credibility, and check whether nodes 101358, 72378, and 130371 still rank in the top ten; if they disappear, the MVC-specific coverage gain is an artifact of the random seed rather than a property of the network.
  • The headline coverage gain is dominated by DIC's ten entirely disjoint nodes; a direct check would be to re-run DIC with varying iteration counts and decay factors to see whether its unique top-ten is stable or just an artefact of the chosen number of timesteps.
  • The 62.5% figure comes from a node-removal simulation on a single network snapshot; extending the same protocol to a temporal cascade model, where removed nodes can re-enter or adaptation occurs, would give an upper bound on the real-world mitigation gain.
  • Because the traditional and advanced metric families share so few top nodes, the paper implicitly reframes misinformation mitigation as a portfolio problem: any single score, structural or dynamic, will systematically overlook one class of spreaders, so intervention design should weight all three components deliberately.
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

5 major / 6 minor

Summary. The paper proposes three centrality metrics—Propagation Centrality (PC), Misinformation Vulnerability Centrality (MVC), and Dynamic Influence Centrality (DIC)—and compares them with degree, eigenvector, betweenness, and closeness on the FibVID COVID-19 misinformation dataset, with a replication on the Monant Medical Misinformation dataset. It reports that the novel metrics increase influencer coverage from 29 to 42 nodes (44.8%) and that simulated node-removal interventions reduce misinformation volume by 50% when only traditional nodes are removed and 62.5% when novel-metric nodes are also neutralised. The authors conclude that a combination of traditional and novel centrality measures offers a more robust and generalisable framework for understanding and mitigating health misinformation.

Significance. If the empirical claims were supported, the paper would offer a practical tool for prioritising fact-checking and content moderation, with concrete quantitative improvements (44.8% more influencers, 25% better intervention) and a generalisation test on a second dataset. The open availability of the code on Zenodo is commendable and facilitates checking. However, the central empirical claims rest on two pillars that are not as reported: the MVC vulnerability scores are seeded random draws, and the intervention simulation protocol is unspecified. The proxy-ground-truth validation is only illustrative. Consequently, the headline contributions are not established by the presented evidence. The idea of combining static, temporal, and susceptibility-aware metrics is plausible, but the paper as written does not demonstrate it.

major comments (5)
  1. [Materials and methods, Misinformation Vulnerability Centrality] The MVC vulnerability scores are initialised with random draws from U(0,1), seeded for reproducibility, because FibVID lacks credibility scores. With the update vult+1(vi) = in-degree(vi) × vult(vi), after any fixed number of iterations the min-max normalised final score is a monotone transformation of in-degree(vi)^t × vul0(vi). The MVC ranking is therefore essentially a random perturbation of in-degree. The three "MVC-Exclusive" nodes (101 358, 72 378, 130 371) and the associated percentage gains (30% influencers, 42.9% vulnerable nodes) are not demonstrated to be anything other than artifacts of the particular seed. The authors should replace the random initialisation with an observable vulnerability proxy from FibVID (e.g., retweet volume or emotion-word counts) or provide a multi-seed sensitivity analysis showing that the MVC-exclusive set is stable across seeds.
  2. [Results, Headline findings] The simulated node-removal intervention (50% vs 62.5% reduction in misinformation volume) is a headline result, but no protocol is provided. There is no cascade model, no removal order, no time horizon, no re-evaluation rule, and no code reference for this specific simulation. Without this information the numbers are not reproducible and the claimed 25% improvement cannot be assessed. The full simulation protocol, ideally with code, must be specified.
  3. [Results, Proxy ground truth evaluation (Table 2)] The proxy validation is based on five anonymised nodes with no explanation of how they were selected, no statistical test, and no baseline comparison. The table as printed does not support the claim that "nodes ranked highly by PC and DIC exhibit substantially higher retweet counts" because no aggregate statistics or significance tests are reported. A proper validation would compare the top-ranked nodes against a random baseline or against the traditional top-10 nodes in terms of retweet and emotion-word counts, with effect sizes and confidence intervals.
  4. [Results, Generalisability of advanced metrics beyond FibVID] The Monant dataset is access-restricted, and the description of the replication is thin. The manuscript does not state whether new vulnerability values were drawn for MVC (and with what seed), what top-K threshold was used, or how the time snapshots for DIC were constructed. The claim that the metrics "generalised successfully" is therefore not independently checkable. Please provide these details, or at least a clear pointer to code and a data-sharing arrangement.
  5. [Materials and methods, Dynamic Influence Centrality] The DIC iteration count is "typically around ten" and scores are normalised, but no sensitivity analysis is presented. Since the DIC recurrence has no decay and accumulates along walks, the claim that all ten DIC top nodes are unique to DIC could be sensitive to the stopping criterion. A plot or table of the top-10 set as a function of iteration count should be reported to establish robustness.
minor comments (6)
  1. [State-of-the-art, Closeness centrality] The displayed formula CC(v) = Σ_{u≠v} d(v,u) omits the reciprocal described in the text; please correct or clarify.
  2. [Materials and methods, Experimental setup (PC equation)] The rendered equation for x(vi) is malformed: the sum, the (1−d)/n term, and the d·Σ term are not shown correctly. Please provide the standard PageRank formula.
  3. [Data availability statement and Abstract] The dataset name is spelled both "Monant" (abstract, main text) and "Monat" (data availability statement); please standardise throughout.
  4. [Table 2] The column layout is ambiguous; the row "Node_A 1 2 1 200 30" does not clearly indicate which number is which rank or score. Please reformat with separate, labelled columns.
  5. [Introduction] The claim "To our knowledge, this research represents the first attempt to introduce and validate novel centrality metrics specifically tailored for the propagation of health misinformation" is too strong given the existing work cited (e.g., Refs. [4,5]); please soften or qualify.
  6. [Throughout] There are several typos, including "eficient" for "efficient" in the Computational Note and "suficient" for "sufficient" in the Experimental Setup; a careful proofread is needed.

Circularity Check

3 steps flagged · score 6.0 of 10

MVC's vulnerability scores are seeded random draws, so its 'vulnerable-node' predictions reduce to the injected random input; the retweet-based proxy validation and the PageRank-as-PC rebrand add self-referential and renaming circularity.

  1. self definitional [Materials and methods, Misinformation Vulnerability Centrality (MVC); Results, Misinformation Vulnerability Centrality (MVC) and Table 3]
    "In this study, due to the unavailability of detailed user credibility scores in Fibvid, we simulate vulnerability by assigning random values from a uniform distribution U(0,1), seeded for reproducibility."

    The MVC update is vult+1(vi) = in-degree(vi) x vult(vi); with static in-degree and monotone min-max normalisation, the final ranking is a monotone function of in-degree times the initial random vul0 draw. The three MVC-exclusive nodes (101358, 72378, 130371) are therefore selected by the seeded random values and degree, yet the Results describe them as 'combine moderate connectivity with high susceptibility' and Table 3 credits MVC with '+42.9% vulnerable nodes'. The vulnerability prediction is the random input, transformed and relabelled: no independent susceptibility signal enters the derivation.

  2. other [Results, Proxy ground truth evaluation]
    "Although the FibVID dataset lacks explicit ground truth labels for influence or vulnerability, we operationalised proxy ground truths based on observable network features. Specifically, we treated nodes with the highest number of retweet counts as proxies for influence, and posts containing higher proportions of emotionally charged language (fear, outrage, conspiracy) as proxies for vulnerability."

    PC and DIC are computed on the FibVID interaction graph whose edge weights are retweet frequencies; PageRank-style propagation (PC) and cumulative-neighbour updates (DIC) are functions of that same retweet signal. Validating these metrics against 'highest number of retweet counts' is therefore checking the metric against a component of its own input, not against an external ground truth. The claimed alignment is largely self-referential and cannot by itself establish that the new metrics capture influence or vulnerability.

1 more flagged steps
  1. renaming known result [Abstract; Materials and methods, Propagation Centrality (PC); Results, Propagation Centrality (PC)]
    "This study introduces and compares three novel centrality metrics: dynamic influence centrality (DIC), health misinformation vulnerability centrality (MVC), and propagation centrality (PC). ... PC, implemented as personalised PageRank, pinpoints nodes that can drive information cascades across the network."

    The paper's contribution list says 'Introduce DIC, MVC, and PC', but the PC formula is the standard PageRank recurrence with damping d=0.85 (x(vi) = (1-d)/n + d sum over in-neighbours x/dout). The text itself states 'PC, implemented as personalised PageRank'. The 'novel propagation centrality' is PageRank under a new name; its output is the known algorithm's output, so any PC-based 'discovery' is not a new metric derivation.

full rationale

The central claim that combining traditional and new centrality metrics yields a 'more robust and generalisable framework' rests partly on MVC's ability to find 'vulnerable' nodes and on the simulated intervention gain. That MVC result reduces by construction: vulnerability is initialised as a seeded uniform random draw and then multiplied by static in-degree, so the three MVC-exclusive nodes are an artefact of the random input, not an empirical discovery about susceptibility. The proxy-ground-truth validation is self-referential because the retweet counts used as the validation target are the same interaction signal from which the graph and the metrics are built. Separately, PC is explicitly identified as personalised PageRank, so presenting it as a novel metric is a renaming of a known algorithm. The paper does not rely on a load-bearing self-citation chain, and the Monant dataset is an external check, but the MVC random-input problem carries over to that generalisation claim. The simulated node-removal intervention (50% vs 62.5%) is also reported without a reproducible protocol, which further weakens, though is not itself circular, the headline mitigation claim. Overall, the score reflects that one key 'prediction'—the vulnerability-based discovery—reduces to its injected input, with additional self-referential validation and renaming issues, while other claims retain at least some independent content.

Assumptions & free parameters 6 free parameters · 5 assumptions · 1 invented entities

The central results depend on several hand-chosen parameters (damping factor, iteration counts, top-K cutoff, proxy definitions) and on an ad hoc random initialization of MVC vulnerability. The only standard axiom is Perron-Frobenius for PageRank convergence. No new physical or structural entity is proposed; the MVC vulnerability score is an invented latent quantity with no independent evidence, making it the most significant item in this ledger.

free parameters (6)
  • damping factor d = 0.85
    Used in Propagation Centrality, which is PageRank; this is a standard but hand-set parameter.
  • MVC vulnerability random values = U(0,1), seed not disclosed
    Initial vulnerability scores are drawn from a uniform distribution; the ordering of MVC results depends directly on these random draws.
  • MVC iteration count = 5 to 10
    The number of multiplications by in-degree is chosen by the authors and affects the final normalized scores.
  • DIC iteration count = about 10
    The cumulative influence sum is truncated at a small number of steps; the ranking depends on this truncation.
  • Top-K threshold = 10
    All comparisons of influential nodes use top-10 lists; the overlap statistics (29, 24, 42 nodes) depend on this cutoff.
  • Proxy ground truth definitions = retweet count as influence, emotion word count as vulnerability
    These proxies are chosen post hoc to validate the metrics and are not independently grounded or statistically tested.
assumptions (5)
  • standard math Perron-Frobenius theorem ensures convergence of PageRank-style iteration.
    Invoked in the Experimental Setup to justify PC convergence; standard spectral graph theory.
  • domain assumption The FibVID network representation with edges weighted by retweet frequency is a valid model of misinformation spread.
    All centralities are computed on this static directed graph; the validity of the findings depends on this representation matching real-world influence.
  • ad hoc to paper Random uniform values can stand in for missing user credibility scores in MVC.
    The paper states that due to unavailability of credibility scores, vulnerability is simulated with U(0,1). This is an arbitrary stand-in, not a measured quantity.
  • domain assumption Simulated node removal corresponds to real reduction in misinformation exposure.
    The 50% vs 62.5% headline result rests on an undescribed simulation whose mapping to real-world misinformation volume is assumed.
  • domain assumption The Monant Medical Misinformation dataset is a valid generalization benchmark.
    The dataset is access-restricted and its use is described only briefly; the generalization claim assumes the dataset represents diverse health misinformation discussions comparable to FibVID.
invented entities (1)
  • MVC vulnerability score
    purpose: To quantify a node's susceptibility to health misinformation based on connectivity and an initial vulnerability value.
    The vulnerability score is initialized from a seeded uniform random distribution, not from any measurement. The only purported evidence, Table 2, is unverifiable and lacks methodology. No external behavioral or psychological data supports the score.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Analysing Health Misinformation with Advanced Centrality Metrics in Online Social Networks." pith.science (2026). https://pith.science/paper/QGYTFAHS

@misc{pith2026250709055,
  author       = {Pith},
  title        = {Pith review of: Analysing Health Misinformation with Advanced Centrality Metrics in Online Social Networks},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/QGYTFAHS}},
  note         = {Machine review of arXiv:2507.09055}
}
read the original abstract

The rapid spread of health misinformation on online social networks (OSNs) during global crises such as the COVID-19 pandemic poses challenges to public health, social stability, and institutional trust. Centrality metrics have long been pivotal in understanding the dynamics of information flow, particularly in the context of health misinformation. However, the increasing complexity and dynamism of online networks, especially during crises, highlight the limitations of these traditional approaches. This study introduces and compares three novel centrality metrics: dynamic influence centrality (DIC), health misinformation vulnerability centrality (MVC), and propagation centrality (PC). These metrics incorporate temporal dynamics, susceptibility, and multilayered network interactions. Using the FibVID dataset, we compared traditional and novel metrics to identify influential nodes, propagation pathways, and misinformation influencers. Traditional metrics identified 29 influential nodes, while the new metrics uncovered 24 unique nodes, resulting in 42 combined nodes, an increase of 44.83%. Baseline interventions reduced health misinformation by 50%, while incorporating the new metrics increased this to 62.5%, an improvement of 25%. To evaluate the broader applicability of the proposed metrics, we validated our framework on a second dataset, Monant Medical Misinformation, which covers a diverse range of health misinformation discussions beyond COVID-19. The results confirmed that the advanced metrics generalised successfully, identifying distinct influential actors not captured by traditional methods. In general, the findings suggest that a combination of traditional and novel centrality measures offers a more robust and generalisable framework for understanding and mitigating the spread of health misinformation in different online network contexts.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

27 extracted references · 24 canonical work pages

  1. [1]

    Towards a methodology for validation of centrality measures in complex networks

    Batool K, Niazi MA. Towards a methodology for validation of centrality measures in complex networks. PLoS One. 2014;9(4):e90283. https://doi.org/10.1371/journal.pone.0090283 PMID: 24709999

    work page doi:10.1371/journal.pone.0090283 2014

  2. [2]

    How correlated are network centrality measures?

    Valente TW, Coronges K, Lakon C, Costenbader E. How correlated are network centrality measures?. Connect (Tor). 2008;28(1):16–26. PMID: 20505784

    work page 2008

  3. [3]

    Deciphering the laws of social network-transcendent COVID-19 misinformation dynamics and implications for combating misinformation phenomena

    Cheng M, Yin C, Nazarian S, Bogdan P. Deciphering the laws of social network-transcendent COVID-19 misinformation dynamics and implications for combating misinformation phenomena. Sci Rep. 2021;11(1):10424. https://doi.org/10.1038/s41598-021-89202-7 PMID: 34001937

    work page doi:10.1038/s41598-021-89202-7 2021

  4. [4]

    Temporal degree-degree and closeness-closeness: a new centrality metrics for social network analysis

    Elmezain M, Othman EA, Ibrahim HM. Temporal degree-degree and closeness-closeness: a new centrality metrics for social network analysis. Mathematics. 2021;9(22):2850. https://doi.org/10.3390/math9222850

    work page doi:10.3390/math9222850 2021

  5. [5]

    Centrality in complex networks with overlapping community structure

    Ghalmane Z, Cherifi C, Cherifi H, Hassouni ME. Centrality in complex networks with overlapping community structure. Sci Rep. 2019;9(1):10133. https://doi.org/10.1038/s41598-019-46507-y PMID: 31300702

    work page doi:10.1038/s41598-019-46507-y 2019

  6. [6]

    Health misinformation detection in the social web: an overview and a data science approach

    Di Sotto S, Viviani M. Health misinformation detection in the social web: an overview and a data science approach. Int J Environ Res Public Health. 2022;19(4):2173. https://doi.org/10.3390/ijerph19042173 PMID: 35206359

    work page doi:10.3390/ijerph19042173 2022

  7. [7]

    Analyzing the spread of misinformation on social networks: a process and software architecture for detection and analysis

    Duzen Z, Riveni M, Aktas MS. Analyzing the spread of misinformation on social networks: a process and software architecture for detection and analysis. Computers. 2023;12(11):232. https://doi.org/10.3390/computers12110232

    work page doi:10.3390/computers12110232 2023

  8. [8]

    Social network analysis for social neuroscientists

    Baek EC, Porter MA, Parkinson C. Social network analysis for social neuroscientists. Soc Cogn Affect Neurosci. 2021;16(8):883–901. https://doi.org/10.1093/scan/nsaa069 PMID: 32415969

    work page doi:10.1093/scan/nsaa069 2021

Show all 27 references
  1. [9]

    A systematic survey of centrality measures for protein-protein interaction networks

    Ashtiani M, Salehzadeh-Yazdi A, Razaghi-Moghadam Z, Hennig H, Wolkenhauer O, Mirzaie M, et al. A systematic survey of centrality measures for protein-protein interaction networks. BMC Syst Biol. 2018;12(1):80. https://doi.org/10.1186/s12918-018-0598-2 PMID: 30064421

    2018 doi

  2. [10]

    Heatmap centrality: a new measure to identify super-spreader nodes in scale-free networks

    Durón C. Heatmap centrality: a new measure to identify super-spreader nodes in scale-free networks. PLoS One. 2020;15(7):e0235690. PLOS Digital Health https://doi.org/10.1371/journal.pdig.0000888 June 16, 2025 17/ 18 PLOS DIGITAL HEALTH Advanced centrality metrics for analysin...

    2020 doi

  3. [11]

    Prevalence of health misinformation on social media-challenges and mitigation before, during, and beyond the COVID-19 pandemic: scoping literature review

    Kbaier D, Kane A, McJury M, Kenny I. Prevalence of health misinformation on social media-challenges and mitigation before, during, and beyond the COVID-19 pandemic: scoping literature review. J Med Internet Res. 2024;26:e38786. https://doi.org/10.2196/38786 PMID: 39159456

    2024 doi

  4. [12]

    Networks

    Newman MEJ. Networks. 2nd ed. Oxford University Press. 2018

    2018

  5. [13]

    Exploring network structure, dynamics, and function using NetworkX

    Hagberg AA, Schult DA, Swart PJ. Exploring network structure, dynamics, and function using NetworkX. In: Proceedings of the7th Python in Science Conference. 2008. p. 11–5

    2008

  6. [14]

    A set of measures of centrality based on betweenness

    Freeman LC. A set of measures of centrality based on betweenness. Sociometry. 1977;40(1):35. https://doi.org/10.2307/3033543

    1977 doi

  7. [15]

    Characterizing information leaders in Twitter during COVID-19 pandemic

    Pastor-Escuredo D, Tarazona C. Characterizing information leaders in Twitter during COVID-19 pandemic. 2022

    2022

  8. [16]

    Challenging social media threats using collective well-being-aware recommendation algorithms and an educational virtual companion

    Ognibene D, Wilkens R, Taibi D, Hernández-Leo D, Kruschwitz U, Donabauer G, et al. Challenging social media threats using collective well-being-aware recommendation algorithms and an educational virtual companion. Front Artif Intell. 2023;5:654930. https://doi.org/10.3389/frai...

    2023

  9. [17]

    Temporal networks

    Holme P, Saramäki J. Temporal networks. Phys Rep. 2012;519(3):97–125

    2012

  10. [18]

    The elaboration likelihood model of persuasion

    Petty RE, Cacioppo JT. The elaboration likelihood model of persuasion. Springer. 1986

    1986

  11. [19]

    FibVID: comprehensive fake news diffusion dataset during the COVID-19 period

    Kim J, Aum J, Lee S, Jang Y, Park E, Choi D. FibVID: comprehensive fake news diffusion dataset during the COVID-19 period. Telemat Inform. 2021;64:101688. https://doi.org/10.1016/j.tele.2021.101688 PMID: 36567815

    2021

  12. [20]

    Monant medical misinformation dataset: mapping articles to fact-checked claims

    Srba I, Pecher B, Tomlein M. Monant medical misinformation dataset: mapping articles to fact-checked claims. arXiv preprint 2022. https://arxiv.org/abs/2204.12294

    2022 arXiv

  13. [21]

    Unveiling influence in networks: a novel centrality metric and comparative analysis through graph-based models

    Bendahman N, Lotfi D. Unveiling influence in networks: a novel centrality metric and comparative analysis through graph-based models. Entropy (Basel). 2024;26(6):486. https://doi.org/10.3390/e26060486 PMID: 38920495

    2024 doi

  14. [22]

    A spatial interaction incorporated betweenness centrality measure

    Wu X, Cao W, Wang J, Zhang Y, Yang W, Liu Y. A spatial interaction incorporated betweenness centrality measure. PLoS One. 2022;17(5):e0268203. https://doi.org/10.1371/journal.pone.0268203 PMID: 35594259

    2022 doi

  15. [23]

    Virality prediction and community structure in social networks

    Weng L, Menczer F, Ahn YY. Virality prediction and community structure in social networks. Sci Rep. 2013;3(1):1–6

    2013

  16. [24]

    Fairness-aware fake news mitigation using counter information propagation

    Saxena A, Gutiérrez Bierbooms C, Pechenizkiy M. Fairness-aware fake news mitigation using counter information propagation. Appl Intell. 2023;53(22):27483–504

    2023

  17. [25]

    Maximizing the spread of influence through a social network

    Kempe D, Kleinberg J, Tardos É. Maximizing the spread of influence through a social network. In: Proceedings of the 9th ACM SIGKDD International Conference on Knowledge Discovery Data Minning. 2003. p. 137–46

    2003

  18. [26]

    Fighting misinformation on social media using crowdsourced judgments of news source quality

    Pennycook G, Rand DG. Fighting misinformation on social media using crowdsourced judgments of news source quality. Proc Natl Acad Sci U S A. 2019;116(7):2521–6. https://doi.org/10.1073/pnas.1806781116 PMID: 30692252

    2019 doi

  19. [27]

    CoAID: COVID-19 healthcare misinformation dataset

    Cui L, Lee D. CoAID: COVID-19 healthcare misinformation dataset. arXiv preprint 2020. https://arxiv.org/abs/2006.00885 PLOS Digital Health https://doi.org/10.1371/journal.pdig.0000888 June 16, 2025 18/ 18

    2020 arXiv

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.