Pith. sign in

REVIEW 4 major objections 6 minor 38 references

The Impact of COVID-19 on Twitter Ego Networks: Structure, Sentiment, and Topics

T0 review · 4 major / 6 minor · reviewed 2026-08-07 · deepseek-v4-flash

Pith's one-line read Lockdowns temporarily expanded Twitter users' ego networks: more circles, more negative ties, and wider topic diversity, all reverting once restrictions lifted.

desk verdict A careful longitudinal Twitter study whose headline 'temporary lockdown adaptation' result is vulnerable to a tweet-volume confound the authors never fully address. read the letter →

arxiv 2506.03788 v1 pith:MAN3RY6K submitted 2025-06-04 cs.SI physics.soc-ph

classification cs.SIphysics.soc-ph
keywords egonetworksCOVID-19lockdownTwittersignedsentimentanalysistopicmodelingBERTopiclongitudinalsocialmedia
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper claims that the COVID-19 lockdown temporarily changed how Twitter users maintain their personal online networks. Using seven years of activity from over a thousand regular users, it finds that during the lockdown year ego networks grew, social circles became more structured, relationships were more likely to be negative, and users touched a wider range of topics. It then shows these metrics largely returned to pre-pandemic levels once restrictions were lifted. The point of the study is that online social life absorbed the cognitive effort displaced from offline contact, and that this adaptation was reversible rather than a permanent reorganization of relationships.

What carries the argument

The central object is the ego network itself: each user's active alters are those contacted at least once per year through replies, mentions, or retweets, and these alters are grouped into concentric intimacy circles by Mean Shift clustering on interaction frequency. Relationship polarity is assigned per alter using a threshold of 17% negative interactions, based on sentiment labels from the BertTweet model, and thematic diversity is measured as the number of distinct BERTopic clusters per user per year. The argument is carried by comparing differences of consecutive annual growth rates across the seven yearly intervals, which isolates the lockdown period as the only interval with a statistically significant expansion and the post-lockdown period as the only one with a significant contraction.

What would settle it

Recompute active ego network size and unique topic counts per unit of posting activity (e.g., per 1,000 tweets or per interaction) across periods I4, I5, and I6. If the per-activity values are flat or declining while raw values peak in I5, the claimed lockdown-driven expansion is a volume artifact.

Watch

Extended reading notes

Core claim

During the main lockdown interval (March 2020 to March 2021), the average active ego network size jumped from about 123 to 150 alters, the share of negative relationships rose significantly, and the average number of unique topics per user rose from about 84 to 102. In the following year all three measures fell, and t-tests on the difference of consecutive annual growth rates are statistically significant only for the lockdown and post-lockdown triplets. The authors interpret this as a reallocation of cognitive resources: with offline socialization curtailed, users invested more in online ties, explored more diverse content, and, under pandemic stress, interacted more negatively; when restrictions ended, attention returned offline and the metrics reverted.

Load-bearing premise

The load-bearing premise is that the growth in ego-network size and unique topic counts during lockdown reflects deeper social engagement, not merely the fact that users tweeted more; the metrics are not normalized by tweet volume.

Editorial extensions

If this is right

  • If the paper is right, future forced reductions in offline contact should reproduce the same temporary pattern: larger active ego networks, more structured outer circles, more negative ties, and higher topic diversity during the isolation period.
  • The layer structure is resilient: inner circles stay stable while outer circles absorb the expansion, so cognitive capacity limits still bind even when online engagement intensifies.
  • The return of all three dimensions to pre-pandemic levels means the lockdown did not permanently reorganize online relationships; the effect is a reversible adaptation.
  • The rise in negative relationships during lockdown supports a stress-driven account of online conflict during societal crises.

Reading between the lines

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

  • Because network size and topic count both scale with tweet volume, a tweet-normalized reanalysis (e.g., unique topics per 1,000 tweets, active alters per interaction) is the clearest way to test whether the lockdown peak is a true change in relationship structure or only an activity effect.
  • The snowball sample is seeded from a single well-known public figure, so the reversibility pattern is most directly evidence about that community; random-cohort or cross-cultural replications would show whether the temporary-adaptation story generalizes.
  • If the same rise-and-revert pattern appears in other mass social-isolation events (or in platform-specific lockdowns), the finding becomes a general behavioral law; if not, it is a one-off pandemic artifact.
  • The broader-topic, larger-network period also implies wider information exposure, so the results suggest that lockdowns may have amplified both the reach and the conflict potential of online information diffusion.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 6 minor

Summary. The manuscript analyzes 1,286 Twitter users' ego networks over a seven-year window (March 2015 to March 2022), divided into yearly periods with the early-2020 lockdown as the boundary between I4 and I5. For each period it measures active ego-network size (Eq. 1), social-circle structure via Mean Shift, relationship polarity via BERTweet sentiment with a 17% negativity threshold, and thematic diversity via the number of unique BERTopic clusters per user. The reported results show a peak during the lockdown year I5 in network size, circle count, negative-relationship share, and topic counts, followed by a decline in I6. The authors interpret this as a temporary reallocation of cognitive resources to online social life.

Significance. If the confounds discussed below were resolved, this would be a valuable longitudinal contribution: it is one of the few large-scale analyses of ego-network structure, sentiment, and topic diversity across a major societal disruption, and it explicitly tests for temporal reversion. The data span is substantial (67M tweets from 1,286 consistently active users), and the growth-rate-difference testing is transparent. However, the central claim is not yet established because the key structural and thematic metrics are raw counts that scale mechanically with tweet volume, and because the post-lockdown period still contains restrictions. With volume normalization and a better-defined post-lockdown phase, the paper could make a solid contribution.

major comments (4)
  1. [Section 5, Eq. (2), Figs. 5 and 8] The active ego-network size |A_i^u| (Eq. 1) and the unique-topic count |T_{I_i}^j| are both unscaled counts that increase mechanically with tweet production. Figure 2 shows a sharp rise in tweet volume beginning in March 2020 that persists through I5; this is exactly where Figures 5(a) and 8(a) peak. An alter is active when w >= 1 interaction occurs per year, so a user who tweets more will accumulate more active alters even if no underlying relationship changes; similarly, more tweets will naturally span more BERTopic clusters. The t-tests in Tables 1 and 4 compare growth-rate differences between periods but do not control for tweet volume. Please re-run the analysis normalizing by number of tweets or interactions (e.g., active alters per tweet, unique topics per tweet) or including tweet volume as a covariate; otherwise the central claim that lockdown expanded and intensified ego networks and broadened topics is confounded by a mechanical volume effect.
  2. [Section 3 (period definitions)] The post-lockdown interval I6 (March 2021 to March 2022) is not a clean 'restrictions lifted' period: many countries in Europe and North America still had substantial COVID-19 restrictions, including curfews and vaccine-pass mandates, during parts of that year. The manuscript's claim that metrics 'largely reverted to pre-pandemic norms' in I6 therefore conflates the easing of restrictions with ongoing pandemic conditions. Please use a quantitative restriction index (e.g., the Oxford COVID-19 Government Response Tracker) to define lockdown and post-lockdown phases, or model stringency continuously, or explicitly restrict the reversion claim to the period when restrictions were actually lifted for the majority of users in the sample.
  3. [Sections 6.3 and 4.3.2] The topic-diversity metric counts unique BERTopic clusters per user, but 72.66% of texts are classified as outliers by HDBSCAN and excluded. The per-period distribution of these outliers is not reported; if the outlier fraction varies with tweet volume or with the pandemic period, the unique-topic count may reflect changes in the outlier rate rather than genuine thematic diversity. Please report per-period outlier proportions and test the sensitivity of the |T| result to alternative outlier handling (e.g., including outliers as a separate topic, or varying min_cluster_size).
  4. [Section 3 (sampling)] The dataset is a snowball sample seeded from Roberto Burioni, an Italian virologist active in vaccine-related debates, with 87% of users tweeting in English. The manuscript acknowledges that the group is 'not fully representative' but then interprets findings as general properties of 'users' and 'online ego networks.' This sampling frame is likely enriched in users engaged in health-science and political arguments, which could bias both the sentiment (negative-interaction) and topic-diversity results. Please either temper the generalizing language in the abstract and conclusions, or add robustness checks on a more diverse sample or population-weighted partition of the data.
minor comments (6)
  1. [Table 4] The p-value for the (I3, I4, I5) triplet is 0.0364 in the table but the text states '0.0 × 10−4'; please correct this inconsistency.
  2. [Figure 3] The bar for the third language is labeled 'France'; it should read 'French'.
  3. [Abstract] The abstract says 'five years pre-pandemic and two years post,' but the period scheme I0-I4 is five years, I5 is the lockdown year, and I6 is a single post-lockdown year; please clarify the counting.
  4. [Sections 4.1 and 4.2] Eq. (1) counts mentions, replies, and retweets for active relationships, while Section 4.2 includes quotes in the interaction set for polarity classification; please clarify why quotes are treated differently in the two analyses.
  5. [Conclusions] There is a typo in 'Interstingly' that should read 'Interestingly.'
  6. [Tables] Several table headers are rendered as 'T able' in the PDF; please fix the formatting.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the temporal trends are measured directly, and the self-citations are to prior falsifiable work rather than reductions of the derivation to its inputs.

full rationale

The paper's main metrics are empirical measurements, not quantities defined in terms of the outcomes they are used to support. Active ego-network size (Eq. 1, with active alters defined by w>=1 per period) counts distinct alters with at least one interaction; the lockdown peak is a measured pattern, not a consequence of the definition. The 17% negativity threshold, Mean Shift circle extraction, and BERTopic pipeline are methodological imports from prior work (Tacchi et al. 2024a; Cekini et al. 2024) or external libraries, but none is fitted to the lockdown trend, and the polarity and topic-diversity comparisons are t-tests on measured growth rates. Section 6.1 recaps circle-level findings from the authors' own Cekini et al. (2024), but that is a published, externally falsifiable prior study, and the current paper independently presents the headline network-size result in Figure 5 and Table 1; the structural recap is a completeness limitation, not a load-bearing circular reduction. The reader's concern that network size and topic counts scale mechanically with tweet volume is a substantive correctness/identification risk (volume is not included as a covariate or normalizer), but it is not circularity: the metrics are not defined in terms of the cognitive-resource conclusion they are used to explain.

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

The central claims rest on several chosen thresholds and domain assumptions rather than on a fitted predictive model. The most consequential are the activity threshold defining ego-network membership, the 17% negativity cutoff defining signed relationships, and the BERTopic configuration producing topic counts. The paper also assumes that Twitter interaction volume proxies cognitive investment and that the snowball sample from Roberto Burioni represents general users. No new entities are introduced.

free parameters (5)
  • Active relationship threshold (w >= 1 interaction per year) = 1 interaction per year
    Defines which alters are in the active ego network (Eq. 1); changes in tweet volume mechanically affect this count and hence ego network size.
  • Negative relationship threshold = 17% negative interactions
    A relationship is classified negative if over 17% of interactions are negative (Section 4.2), adopted from Tacchi et al. (2024a); this drives all polarity percentages.
  • BERTopic hyperparameters = n_components=2, n_neighbors=10, min_cluster_size=20, leaf
    Selected by DBCV grid search on a 500k-tweet sample (Appendix A.1); determines the topics and the 72.66% outlier rate that shapes the unique-topic metric.
  • User retention criteria = active in >=50% of months; inactivity less than 6 months beyond typical
    Filters the 53,837 users down to 1,627 regular active users (Section 3); this selection affects all downstream metrics.
  • IQR outlier exclusion = IQR-based, applied per period
    Users with active ego network size flagged as outliers in any interval are removed (Section 3, Appendix A.2), reducing the sample to 1,286; this post-hoc exclusion could affect observed trends.
assumptions (5)
  • domain assumption Twitter interaction frequency reflects tie strength and intimacy (Eq. 1)
    Central to ego-network extraction; if replies, mentions, and retweets per year do not proxy closeness, the circle structure is not meaningful.
  • domain assumption The 17% negativity threshold, grounded in Gottman and Hart/Risley, transfers from face-to-face relationship studies to Twitter relationships
    Section 4.2 relies on this external threshold to label whole relationships positive or negative, without validation on Twitter relationship outcomes.
  • ad hoc to paper The snowball sample seeded from Roberto Burioni is sufficiently representative of general Twitter users
    Section 3 starts collection from a single prominent Italian virologist; the network is likely biased toward COVID and vaccine discourse and toward Italian and English science-engaged users.
  • domain assumption March 1, 2020 marks the start of lockdown, and March 1, 2021 to March 1, 2022 is a post-lockdown period
    The period definitions in Section 3 are coarse; I6 still includes pandemic waves and restrictions in many countries, so 'return to pre-pandemic norms' is only partially supported.
  • ad hoc to paper The number of unique BERTopic clusters per user per year measures thematic diversity independently of tweet volume
    Section 6.3 uses the unique-topic count as the diversity metric without normalizing by number of tweets; higher activity during lockdown mechanically allows more clusters to be detected.

how reviews work

0 comments
Cite this review

Pith. "Pith review of The Impact of COVID-19 on Twitter Ego Networks: Structure, Sentiment, and Topics." pith.science (2026). https://pith.science/paper/MAN3RY6K

@misc{pith2026250603788,
  author       = {Pith},
  title        = {Pith review of: The Impact of COVID-19 on Twitter Ego Networks: Structure, Sentiment, and Topics},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/MAN3RY6K}},
  note         = {Machine review of arXiv:2506.03788}
}
read the original abstract

Lockdown measures, implemented by governments during the initial phases of the COVID-19 pandemic to reduce physical contact and limit viral spread, imposed significant restrictions on in-person social interactions. Consequently, individuals turned to online social platforms to maintain connections. Ego networks, which model the organization of personal relationships according to human cognitive constraints on managing meaningful interactions, provide a framework for analyzing such dynamics. The disruption of physical contact and the predominant shift of social life online potentially altered the allocation of cognitive resources dedicated to managing these digital relationships. This research aims to investigate the impact of lockdown measures on the characteristics of online ego networks, presumably resulting from this reallocation of cognitive resources. To this end, a large dataset of Twitter users was examined, covering a seven-year period of activity. Analyzing a seven-year Twitter dataset -- including five years pre-pandemic and two years post -- we observe clear, though temporary, changes. During lockdown, ego networks expanded, social circles became more structured, and relationships intensified. Simultaneously, negative interactions increased, and users engaged with a broader range of topics, indicating greater thematic diversity. Once restrictions were lifted, these structural, emotional, and thematic shifts largely reverted to pre-pandemic norms -- suggesting a temporary adaptation to an extraordinary social context.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

38 extracted references · 37 canonical work pages

  1. [1]

    In: ACM SIGMOD Record, vol

    Ankerst, M., Breunig, M.M., Kriegel, H.-P., Sander, J.: Optics: ordering points to identify the clustering structure. In: ACM SIGMOD Record, vol. 28, pp. 49–60 (1999). ACM

  2. [2]

    In: 2013 Proceedings IEEE INFOCOM, pp

    Arnaboldi, V., Conti, M., Passarella, A., Pezzoni, F.: Ego networks in twitter: an exper- imental analysis. In: 2013 Proceedings IEEE INFOCOM, pp. 3459–3464 (2013). IEEE

  3. [3]

    In: ICISP’20, pp

    Allaoui, M., Kherfi, M.L., Cheriet, A.: Considerably improving clustering algo- rithms using UMAP dimensionality reduction technique: A comparative study. In: ICISP’20, pp. 317–325 (2020). Springer

  4. [4]

    Journal of Computer-Mediated Communication22(5), 231–247 (2017)

    Arnaboldi, V., Passarella, A., Conti, M., Dunbar, R.: Structure of ego-alter relation- ships of politicians in twitter. Journal of Computer-Mediated Communication22(5), 231–247 (2017)

  5. [5]

    American Journal of Sociology 117(1), 90–171 (2011)

    Aral, S., Van Alstyne, M.: The diversity-bandwidth trade-off. American Journal of Sociology 117(1), 90–171 (2011)

  6. [6]

    In: Proceedings of the Biennial GSCL Conference 2009, pp

    Bouma, G.: Normalized (Pointwise) Mutual Information in collocation extraction. In: Proceedings of the Biennial GSCL Conference 2009, pp. 31–40 (2009)

  7. [7]

    In: Companion Proceedings of the The Web Conference 2018, pp

    Boldrini, C., Toprak, M., Conti, M., Passarella, A.: Twitter and the press: an ego- centred analysis. In: Companion Proceedings of the The Web Conference 2018, pp. 1471–1478 (2018)

  8. [8]

    In: International Conference on Discovery Science, pp

    Cekini, K., Biondi, E., Boldrini, C., Passarella, A., Conti, M.: Social Isolation, Dig- ital Connection: COVID-19’s Impact on Twitter Ego Networks. In: International Conference on Discovery Science, pp. 260–274 (2024). Springer

Show all 38 references
  1. [9]

    In: Pei, J., Tseng, V.S., Cao, L., Motoda, H., Xu, G

    Campello, R.J.G.B., Moulavi, D., Sander, J.: Density-based clustering based on hier- archical density estimates. In: Pei, J., Tseng, V.S., Cao, L., Motoda, H., Xu, G. (eds.) PAKDD’13, pp. 160–172 (2013). Springer

  2. [10]

    Social Networks 43, 39–47 (2015)

    Dunbar, R.I.M., Arnaboldi, V., Conti, M., Passarella, A.: The structure of online social networks mirrors those in the offline world. Social Networks 43, 39–47 (2015)

  3. [11]

    Transac- tions of the Association for Computational Linguistics 8, 439–453 (2020)

    Dieng, A.B., Ruiz, F.J.R., Blei, D.M.: Topic modeling in embedding spaces. Transac- tions of the Association for Computational Linguistics 8, 439–453 (2020)

  4. [12]

    Human nature 6(3), 273–290 (1995)

    Dunbar, R.I., Spoors, M.: Social networks, support cliques, and kinship. Human nature 6(3), 273–290 (1995)

  5. [13]

    Evolutionary Anthropology: Issues, News, and Reviews 6(5), 178–190 (1998) 21

    Dunbar, R.I.: The social brain hypothesis. Evolutionary Anthropology: Issues, News, and Reviews 6(5), 178–190 (1998) 21

  6. [14]

    In: KDD’96, pp

    Ester, M., Kriegel, H.-P., Sander, J., Xu, X.: A density-based algorithm for discovering clusters in large spatial databases with noise. In: KDD’96, pp. 226–231 (1996). AAAI Press Ford Rojas, J.-P.: Coronavirus: Lockdowns drive record growth in Twitter usage. https://news.sky....

  7. [15]

    PeerJ Computer Science 1, 26 (2015)

    Ferrara, E., Yang, Z.: Quantifying the effect of sentiment on information diffusion in social media. PeerJ Computer Science 1, 26 (2015)

  8. [16]

    Simon and Schuster, New York, US (1995) Gon¸ calves, B., Perra, N., Vespignani, A.: Modeling users’ activity on twitter networks: Validation of dunbar’s number

    Gottman, J.M.: Why Marriages Succeed or Fail: And How You Can Make Yours Last. Simon and Schuster, New York, US (1995) Gon¸ calves, B., Perra, N., Vespignani, A.: Modeling users’ activity on twitter networks: Validation of dunbar’s number. PloS one 6(8), 22656 (2011)

  9. [17]

    https://arxiv.org/abs/2203.05794

    Grootendorst, M.: BERTopic: Neural topic modeling with a class-based TF-IDF procedure (2022). https://arxiv.org/abs/2203.05794

  10. [18]

    Human Nature 14(1), 53–72 (2003) https://doi.org/10.1007/s12110-003-1016-y

    Hill, R.A., Dunbar, R.I.: Social network size in humans. Human Nature 14(1), 53–72 (2003) https://doi.org/10.1007/s12110-003-1016-y

  11. [19]

    PloS one 15(11), 0241957 (2020)

    Huang, X., Li, Z., Jiang, Y., Li, X., Porter, D.: Twitter reveals human mobility dynamics during the COVID-19 pandemic. PloS one 15(11), 0241957 (2020)

  12. [20]

    Hart, B., Risley, T.R.: Meaningful Differences in the Everyday Experience of Young American Children. Paul H. Brookes Publishing Co., Baltimore, MD, US (1995)

  13. [21]

    Proceedings of the 20th International Conference Companion on World Wide Web, WWW 2011, 87–88 (2011)

    Maniu, S., Abdessalem, T., Cautis, B.: Casting a web of trust over wikipedia: An interaction-based approach. Proceedings of the 20th International Conference Companion on World Wide Web, WWW 2011, 87–88 (2011)

  14. [22]

    EPJ Data Science 10(1), 47 (2021)

    Mattei, M., Caldarelli, G., Squartini, T., Saracco, F.: Italian Twitter semantic network during the Covid-19 epidemic. EPJ Data Science 10(1), 47 (2021)

  15. [23]

    Journal of Open Source Software 3(29), 861 (2018)

    McInnes, L., Healy, J., Melville, J.: UMAP: Uniform Manifold Approximation and Projection for Dimension Reduction. Journal of Open Source Software 3(29), 861 (2018)

  16. [24]

    In: Proceedings of the 2014 SIAM International Conference on Data Mining, pp

    Moulavi, D., Jaskowiak, P.A., Campello, R.J.G.B., Zimek, A., Sander, J.: Density- based clustering validation. In: Proceedings of the 2014 SIAM International Conference on Data Mining, pp. 839–847 (2014). SIAM

  17. [25]

    Scientific reports 3(1), 1–7 (2013)

    Miritello, G., Lara, R., Cebrian, M., Moro, E.: Limited communication capacity unveils strategies for human interaction. Scientific reports 3(1), 1–7 (2013)

  18. [26]

    This is Fake 22 News

    Miyazaki, K., Uchiba, T., Tanaka, K., An, J., Kwak, H., Sasahara, K.: “This is Fake 22 News”: Characterizing the Spontaneous Debunking from Twitter Users to COVID- 19 False Information. In: ICWSM’23, vol. 17, pp. 650–661 (2023)

  19. [27]

    ego network

    Ollivier, K., Boldrini, C., Passarella, A., Conti, M.: Structural invariants and semantic fingerprints in the “ego network” of words. PLOS ONE 17(11), 0277182 (2022)

  20. [28]

    Journal of Individual Differences 32(3), 161–169 (2011)

    Pollet, T.V., Roberts, S.G., Dunbar, R.I.: Extraverts have larger social network layers. Journal of Individual Differences 32(3), 161–169 (2011)

  21. [29]

    ICWSM 12, 298–305 (2012)

    Quercia, D., Capra, L., Crowcroft, J.: The social world of twitter: Topics, geography, and emotions. ICWSM 12, 298–305 (2012)

  22. [30]

    Social Networks 31(2), 138–146 (2009)

    Roberts, S.G., Dunbar, R.I., Pollet, T.V., Kuppens, T.: Exploring variation in active network size: Constraints and ego characteristics. Social Networks 31(2), 138–146 (2009)

  23. [31]

    In: EMNLP-IJCNLP’19, pp

    Reimers, N., Gurevych, I.: Sentence-BERT: Sentence Embeddings Using Siamese BERT-Networks. In: EMNLP-IJCNLP’19, pp. 3982–3992 (2019)

  24. [32]

    British journal of psychology 103(2), 149–168 (2012)

    Sutcliffe, A., Dunbar, R., Binder, J., Arrow, H.: Relationships and the social brain: integrating psychological and evolutionary perspectives. British journal of psychology 103(2), 149–168 (2012)

  25. [33]

    https://about.fb.com/news/2020/03/ keeping-our-apps-stable-during-covid-19/ (2020)

    Schultz, A., Parikh, J.: Keeping Our Services Stable and Reliable Dur- ing the COVID-19 Outbreak. https://about.fb.com/news/2020/03/ keeping-our-apps-stable-during-covid-19/ (2020)

  26. [34]

    IEEE Transactions on Computational Social Systems 10(1), 48–60 (2022)

    Toprak, M., Boldrini, C., Passarella, A., Conti, M.: Harnessing the power of ego net- work layers for link prediction in online social networks. IEEE Transactions on Computational Social Systems 10(1), 48–60 (2022)

  27. [35]

    EPJ Data Science 13(1), 55 (2024)

    Tacchi, J., Boldrini, C., Passarella, A., Conti, M.: Keep your friends close, and your enemies closer: Structural properties of negative relationships on twitter. EPJ Data Science 13(1), 55 (2024)

  28. [36]

    https://arxiv.org/ abs/2402.18235 Wikipedia contributors: COVID-19 lockdowns

    Tacchi, J., Boldrini, C., Passarella, A., Conti, M.: On the Joint Effect of Culture and Discussion Topics on X (Twitter) Signed Ego Networks (2024). https://arxiv.org/ abs/2402.18235 Wikipedia contributors: COVID-19 lockdowns. https://en.wikipedia.org/wiki/ COVID-19 lockdowns (2024)

  29. [37]

    Journal of Affective Disorders 277, 55–64 (2020) 23

    Iacobucci, M., Ho, R., Majeed, A., McIntyre, R.S.: Impact of COVID-19 pandemic on mental health in the general population: A systematic review. Journal of Affective Disorders 277, 55–64 (2020) 23

  30. [38]

    Proceedings of the Royal Society B: Biological Sciences 272(1561), 439–444 (2005) https://doi.org/10.1098/rspb.2004.2970 24

    Zhou, W.X., Sornette, D., Hill, R.A., Dunbar, R.I.: Discrete hierarchical organiza- tion of social group sizes. Proceedings of the Royal Society B: Biological Sciences 272(1561), 439–444 (2005) https://doi.org/10.1098/rspb.2004.2970 24

Pith tools

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