REVIEW 2 major objections 2 minor 74 references
Adverse Online Social Interactions: A Multi-Level Evolutionary Analysis of Local Patterns, Diffusion, and Community Disruption
T0 review · 2 major / 2 minor · reviewed 2026-06-26 · grok-4.3
Pith's one-line read Structural disconnection marks subgroup disruption more persistently than toxic communication in online communities.
desk verdict The paper runs matched triadic and reference-model comparisons on X and Bluesky data to argue that structural negativity marks subgroup disruption more persistently than toxic content, but the controls leave room for algorithmic and self-selection confounds. read the letter →
The pith
A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.
The reading
What carries the argument
The multi-level framework that examines local friend-foe patterns, peer influence through matched triadic designs at the meso level, and subgroup disruption against random and recommendation-based references at the macro level.
What would settle it
A new dataset or controlled experiment in which toxic communication alone predicts the timing and persistence of subgroup disruption as strongly as structural disconnection does would falsify the claim that the two provide distinct complementary signals.
Extended reading notes
Core claim
Using data from X and Bluesky, the analysis shows that adverse social interactions operate as multi-scale processes: structural disconnection and toxic communication provide complementary signals where structural negativity more persistently marks subgroup disruption while toxic communication captures broader conflict both within and across communities.
Load-bearing premise
The matched triadic designs and random or recommendation-based reference groups at the middle and large scales remove confounding from platform algorithms and user self-selection in the two datasets.
Editorial extensions
If this is right
- Adverse social interactions influence how online communities form, fracture, and evolve as multi-scale processes.
- Structural negativity serves as a more persistent marker for subgroup disruption than toxic communication.
- Toxic communication signals conflict that spans both inside and across communities.
- Single-scale studies of adverse interactions will miss key dynamics of community change.
Reading between the lines
- Platform tools could combine structural monitoring with content flags to catch different kinds of disruption earlier.
- The same complementarity might appear in other online platforms if the multi-level method is applied there.
- Recommendation systems may need separate adjustments for structural versus content signals to limit community fracture.
Signed reviews
Editorial analysis
A structured set of objections, weighed in public.
Referee Report
Summary. The paper proposes a multi-level framework for studying adverse social interactions (ASIs) in online communities, drawing on large-scale X and Bluesky datasets. It examines friend/foe patterns at the micro level, peer influence via matched triadic designs at the meso level, and subgroup disruption at the macro level by comparing observed patterns against random and recommendation-based reference models. The central claim is that structural disconnection and toxic communication act as complementary signals: structural negativity more persistently marks subgroup disruption, while toxic communication reflects broader conflict within and across communities.
Significance. If the controls successfully isolate ASI effects, the work would provide a rare multi-scale integration of structural and content-based signals for community evolution, with the public code release aiding reproducibility. The directional findings on persistence and scope differences could inform platform moderation if the meso- and macro-level comparisons hold after addressing potential confounders.
major comments (2)
- [meso-level analysis (matched triadic designs)] Meso-level matched triadic designs: the claim that these designs isolate peer influence on ASIs (and thereby support the complementarity result) is load-bearing for the central claim, yet the manuscript provides no details on how self-selection into hostile neighborhoods or algorithmic exposure is blocked; residual correlation between tie formation and unobserved user traits could produce the reported persistence differences without the claimed structural vs. toxic distinction.
- [macro-level analysis] Macro-level reference models: the comparison of subgroup disruption against random and recommendation-based references is presented as evidence that structural negativity marks disruption more persistently, but without reported robustness checks (e.g., alternative matching criteria or sensitivity to recommendation algorithm parameters), it is unclear whether the scope differences survive plausible alternative reference constructions.
minor comments (2)
- [abstract and methods overview] The abstract and methods overview omit any mention of statistical controls, error estimation, data cleaning steps, or robustness checks; these should be added to allow verification of the directional findings.
- [results interpretation] Notation for 'structural disconnection' and 'toxic communication' should be defined explicitly with examples from the datasets before the results are interpreted.
Simulated Author's Rebuttal
We thank the referee for the constructive feedback. We address each major comment below, indicating revisions where the manuscript requires clarification or additional checks.
read point-by-point responses
-
Referee: [meso-level analysis (matched triadic designs)] Meso-level matched triadic designs: the claim that these designs isolate peer influence on ASIs (and thereby support the complementarity result) is load-bearing for the central claim, yet the manuscript provides no details on how self-selection into hostile neighborhoods or algorithmic exposure is blocked; residual correlation between tie formation and unobserved user traits could produce the reported persistence differences without the claimed structural vs. toxic distinction.
Authors: We acknowledge that the manuscript lacks explicit details on how the matched triadic designs address self-selection into hostile neighborhoods or algorithmic exposure. The design matches triads on observable covariates including degree, activity level, and temporal proximity, but residual confounding from unobserved traits remains possible in observational data. In the revision we will expand the methods section with a full description of the matching procedure and covariates, plus a limitations paragraph discussing the inability to fully block all selection effects. This will clarify the scope of the peer-influence claim without overstating isolation. revision: yes
-
Referee: [macro-level analysis] Macro-level reference models: the comparison of subgroup disruption against random and recommendation-based references is presented as evidence that structural negativity marks disruption more persistently, but without reported robustness checks (e.g., alternative matching criteria or sensitivity to recommendation algorithm parameters), it is unclear whether the scope differences survive plausible alternative reference constructions.
Authors: We agree that the macro-level results would be strengthened by explicit robustness checks. The current random and recommendation-based references follow common null-model practices, yet the manuscript does not report sensitivity to alternative matching criteria or recommendation parameters. In the revised version we will add these checks, including variation in recommendation-model parameters and alternative subgroup definitions, to verify that the reported differences in persistence and scope are not artifacts of the specific reference constructions. revision: yes
Circularity Check
Empirical multi-level analysis on external platform data exhibits no circularity
full rationale
The paper performs observational analysis on X and Bluesky datasets using matched triadic designs at meso level and random/recommendation references at macro level. These are standard empirical controls, not derived quantities. No equations, fitted parameters, or self-citations appear as load-bearing steps in the derivation chain; results are presented as direct measurements from data rather than predictions that reduce to inputs by construction. The central claim of complementary signals follows from the empirical comparisons without self-referential reduction.
Assumptions & free parameters
Cite this review
Pith. "Pith review of Adverse Online Social Interactions: A Multi-Level Evolutionary Analysis of Local Patterns, Diffusion, and Community Disruption." pith.science (2026). https://pith.science/paper/WUEQRDVD
@misc{pith2026260620846,
author = {Pith},
title = {Pith review of: Adverse Online Social Interactions: A Multi-Level Evolutionary Analysis of Local Patterns, Diffusion, and Community Disruption},
year = {2026},
howpublished = {\url{https://pith.science/paper/WUEQRDVD}},
note = {Machine review of arXiv:2606.20846}
}
read the original abstract
Adverse social interactions (ASIs) can shape how online communities evolve over the time. However, structural-based ASIs and content-based ASIs are often studied separately and at a single analytical scale. In this study, we propose a multi-level framework to examine how adverse social interactions appear locally, spread through neighborhoods, and disrupt cohesive subgroups. Using large-scale datasets from X and Bluesky, we analyze friend and foe patterns at the micro level, peer influence through matched triadic designs at the meso level, and subgroup disruption against random and recommendation-based references at the macro level. Our results show that structural disconnection and toxic communication provide complementary signals: structural negativity more persistently marks subgroup disruption, while toxic communication captures broader conflict both within and across communities. These findings suggest that adverse social interactions are multi-scale processes that influence how online communities form, fracture, and evolve. Our source code is publicly available at https://github.com/XueqiC/Adverse-Social-Interactions.
Figures
Figures from the paper (4 more)
Reference graph
Works this paper leans on
-
[1]
Shiza Ali, Mohammad Hammas Saeed, Esraa Aldreabi, Jeremy Blackburn, Emil- iano De Cristofaro, Savvas Zannettou, and Gianluca Stringhini. 2021. Under- standing the effect of deplatforming on social networks. InProceedings of the 13th ACM Web Science Conference 2021. 187–195
2021
-
[2]
Roy F Baumeister, Ellen Bratslavsky, Catrin Finkenauer, and Kathleen D Vohs
-
[3]
Bad is stronger than good.Review of general psychology5, 4 (2001), 323–370
2001
-
[4]
Olga Baysha. 2020. Dividing social networks: Facebook unfriending, unfollowing, and blocking in turbulent political times.Russian Journal of communication12, 2 (2020), 104–120
2020
-
[5]
Andres M Belaza, Jan Ryckebusch, Aaron Bramson, Corneel Casert, Kevin Hoef- man, Koen Schoors, Milan van den Heuvel, and Benjamin Vandermarliere. 2019. Social stability and extended social balance—Quantifying the role of inactive links in social networks.Physica A: Statistical Mechanics and its Applications518 (2019), 270–284
2019
-
[6]
Mireia Bolíbar. 2016. Macro, meso, micro: Broadening the ‘social’of social network analysis with a mixed methods approach.Quality & Quantity50, 5 (2016), 2217– 2236
2016
-
[7]
Ceren Budak, Divyakant Agrawal, and Amr El Abbadi. 2011. Limiting the spread of misinformation in social networks. InProceedings of the 20th international conference on World wide web. 665–674
2011
- [8]
Show all 74 references
-
[9]
Dorwin Cartwright and Frank Harary. 1956. Structural balance: a generalization of Heider’s theory.Psychological review63, 5 (1956), 277
1956
-
[10]
Xueqi Cheng. 2025. Edge-Centric Network Analytics. InProceedings of the Eighteenth ACM International Conference on Web Search and Data Mining. 1071– 1073
2025
-
[11]
Xueqi Cheng, Yu Wang, Yunchao Liu, Yuying Zhao, Charu C Aggarwal, and Tyler Derr. 2025. Edge classification on graphs: new directions in topological imbalance. InProceedings of the Eighteenth ACM International Conference on Web Search and Data Mining. 392–400
2025
-
[12]
Xueqi Cheng, Catherine Yang, Yuying Zhao, Yu Wang, Hamid Karimi, and Tyler Derr. 2025. BTS: A Comprehensive Benchmark for Tie Strength Prediction. In Proceedings of the 31st ACM SIGKDD Conference on Knowledge Discovery and Data Mining V. 2. 5345–5354
2025
-
[13]
Kai-Yang Chiang, Nagarajan Natarajan, Ambuj Tewari, and Inderjit S Dhillon
-
[14]
InPro- ceedings of the 20th ACM international conference on Information and knowledge management
Exploiting longer cycles for link prediction in signed networks. InPro- ceedings of the 20th ACM international conference on Information and knowledge management. 1157–1162
-
[15]
Matteo Cinelli, Gianmarco De Francisci Morales, Alessandro Galeazzi, Walter Quattrociocchi, and Michele Starnini. 2021. The echo chamber effect on social media.Proceedings of the national academy of sciences118, 9 (2021), e2023301118
2021
-
[16]
Robert Cote. 2019. The Evolution of Social Network Theory: Perceived Impact on Developing Networking Relationships.American Journal of Management19, 3 (2019)
2019
-
[17]
Paramita Dey and Sarbani Roy. 2017. Centrality based information blocking and influence minimization in online social network. In2017 IEEE international conference on advanced networks and telecommunications systems (ANTS). IEEE, 1–6
2017
-
[18]
2025.Mathematical analysis of signed networks: structure and dynamics
Fernando Diaz-Diaz. 2025.Mathematical analysis of signed networks: structure and dynamics. Ph. D. Dissertation. Institute of Cross-Disciplinary Physics and Complex Systems, IFISC
2025
-
[19]
Patrick Doreian and Andrej Mrvar. 2009. Partitioning signed social networks. Social Networks31, 1 (2009), 1–11
2009
-
[20]
Pouya Esmailian, Seyed Ebrahim Abtahi, and Mahdi Jalili. 2014. Mesoscopic analysis of online social networks-the role of negative ties.arXiv preprint arXiv:1411.6057(2014)
2014 arXiv
-
[21]
Diane Felmlee and Robert Faris. 2016. Toxic ties: Networks of friendship, dating, and cyber victimization.Social psychology quarterly79, 3 (2016), 243–262
2016
-
[22]
Dale Ganley and Cliff Lampe. 2009. The ties that bind: Social network principles in online communities.Decision support systems47, 3 (2009), 266–274
2009
-
[23]
Vasu Goel, Dhruv Sahnan, Subhabrata Dutta, Anil Bandhakavi, and Tanmoy Chakraborty. 2023. Hatemongers ride on echo chambers to escalate hate speech diffusion.PNAS nexus2, 3 (2023), pgad041
2023
-
[24]
Nicholas M Harrigan, Giuseppe Joe Labianca, and Filip Agneessens. 2020. Nega- tive ties and signed graphs research: Stimulating research on dissociative forces in social networks.Social Networks60 (2020), 1–10
2020
-
[25]
Fritz Heider. 1946. Attitudes and cognitive organization.The Journal of psychology 21, 1 (1946), 107–112
1946
-
[26]
Hong Huang, Yuxiao Dong, Jie Tang, Hongxia Yang, Nitesh V Chawla, and Xiaoming Fu. 2018. Will triadic closure strengthen ties in social networks?ACM Transactions on Knowledge Discovery from Data (TKDD)12, 3 (2018), 1–25
2018
-
[27]
Michael G Hudgens and M Elizabeth Halloran. 2008. Toward causal inference with interference.Journal of the american statistical association103, 482 (2008), 832–842
2008
-
[28]
Edda Humprecht, Lea Hellmueller, and Juliane A Lischka. 2020. Hostile emotions in news comments: A cross-national analysis of Facebook discussions.Social Media+ Society6, 1 (2020), 2056305120912481
2020
-
[29]
Abraham Israeli and Oren Tsur. 2022. Going extreme: Comparative analysis of hate speech in parler and gab.arXiv preprint arXiv:2201.11770(2022)
2022
-
[30]
Ana Jorge. 2019. Social media, interrupted: Users recounting temporary discon- nection on Instagram.Social Media+ Society5, 4 (2019), 2056305119881691
2019
-
[31]
Johannes Kaiser, Cristian Vaccari, and Andrew Chadwick. 2022. Partisan blocking: Biased responses to shared misinformation contribute to network polarization on social media.Journal of Communication72, 2 (2022), 214–240
2022
-
[32]
Funda Kivran-Swaine, Priya Govindan, and Mor Naaman. 2011. The impact of network structure on breaking ties in online social networks: unfollowing on twitter. InProceedings of the SIGCHI conference on human factors in computing systems. 1101–1104
2011
-
[33]
Deepak Kumar, Jeff Hancock, Kurt Thomas, and Zakir Durumeric. 2023. Under- standing the behaviors of toxic accounts on reddit. InProceedings of the ACM web conference 2023. 2797–2807
2023
-
[34]
Jérôme Kunegis, Stephan Schmidt, Andreas Lommatzsch, Jürgen Lerner, Ernesto W De Luca, and Sahin Albayrak. 2010. Spectral analysis of signed graphs for clustering, prediction and visualization. InProceedings of the 2010 SIAM international conference on data mining. SIAM, 559–570
2010
-
[35]
Doyeon Kwak and Wonjoon Kim. 2017. Understanding the process of social network evolution: Online-offline integrated analysis of social tie formation.Plos one12, 5 (2017), e0177729
2017
-
[36]
Haewoon Kwak, Hyunwoo Chun, and Sue Moon. 2011. Fragile online relationship: a first look at unfollow dynamics in twitter. InProceedings of the SIGCHI conference on human factors in computing systems. 1091–1100
2011
-
[37]
Jure Leskovec, Daniel Huttenlocher, and Jon Kleinberg. 2010. Predicting pos- itive and negative links in online social networks. InProceedings of the 19th international conference on World wide web. 641–650
2010
-
[38]
Li Li, Xiaohua Zheng, Jing Han, and Fei Hao. 2023. Information cascades blocking through influential nodes identification on social networks.Journal of Ambient Intelligence and Humanized Computing14, 6 (2023), 7519–7530
2023
-
[39]
Yunfei Paul Li, Kathleen J Propert, and Paul R Rosenbaum. 2001. Balanced risk set matching.J. Amer. Statist. Assoc.96, 455 (2001), 870–882
2001
-
[40]
Piper Liping Liu and Tien Ee Dominic Yeo. 2022. Weak ties matter: Social network dynamics of mobile media multiplexity and their impact on the social support and psychological well-being experienced by migrant workers.Mobile Media & Communication10, 1 (2022), 76–96
2022
-
[41]
Jay M Marlowe, Allen Bartley, and Francis Collins. 2017. Digital belongings: The intersections of social cohesion, connectivity and digital media.Ethnicities17, 1 (2017), 85–102
2017
-
[42]
Quinn McNemar. 1947. Note on the sampling error of the difference between correlated proportions or percentages.Psychometrika12, 2 (1947), 153–157
1947
-
[43]
Seth A Myers and Jure Leskovec. 2014. The bursty dynamics of the twitter information network. InProceedings of the 23rd international conference on World wide web. 913–924
2014
-
[44]
Shira Offer. 2021. Negative social ties: Prevalence and consequences.Annual Review of Sociology47, 1 (2021), 177–196
2021
-
[45]
Mina Young Pedersen, Sonja Smets, and Thomas Ågotnes. 2019. Analyzing echo chambers: A logic of strong and weak ties. InInternational Workshop on Logic, Rationality and Interaction. Springer, 183–198
2019
-
[46]
Dung V Pham, Giang L Nguyen, Tu N Nguyen, Canh V Pham, and Anh V Nguyen
-
[47]
Multi-topic misinformation blocking with budget constraint on online social networks.IEEE Access8 (2020), 78879–78889
2020
-
[48]
Nils Reimers and Iryna Gurevych. 2019. Sentence-bert: Sentence embeddings using siamese bert-networks.arXiv preprint arXiv:1908.10084(2019)
2019 arXiv
-
[49]
2010.Design of observational studies
Paul R Rosenbaum, P Rosenbaum, and Briskman. 2010.Design of observational studies. Vol. 10. Springer
2010
-
[50]
Paul R Rosenbaum and Donald B Rubin. 1983. The central role of the propensity score in observational studies for causal effects.Biometrika70, 1 (1983), 41–55
1983
-
[51]
Paul Rozin and Edward B Royzman. 2001. Negativity bias, negativity dominance, and contagion.Personality and social psychology review5, 4 (2001), 296–320
2001
-
[52]
Donald B Rubin. 2005. Causal inference using potential outcomes: Design, mod- eling, decisions.Journal of the American statistical Association100, 469 (2005), 322–331
2005
-
[53]
Giuseppe Russo, Manoel Horta Ribeiro, and Robert West. 2024. Stranger danger! cross-community interactions with fringe users increase the growth of fringe communities on reddit. InProceedings of the International AAAI Conference on Web and Social Media, Vol. 18. 1342–1353
2024
-
[54]
Thomas Ryberg and Malene Charlotte Larsen. 2008. Networked identities: under- standing relationships between strong and weak ties in networked environments. Journal of Computer Assisted Learning24, 2 (2008), 103–115
2008
-
[55]
Kazutoshi Sasahara, Wen Chen, Hao Peng, Giovanni Luca Ciampaglia, Alessandro Flammini, and Filippo Menczer. 2021. Social influence and unfollowing accelerate the emergence of echo chambers.Journal of Computational Social Science4, 1 (2021), 381–402. Conference’17, July 2017, W...
2021
-
[56]
Martin Saveski, Brandon Roy, and Deb Roy. 2021. The structure of toxic conver- sations on Twitter. InProceedings of the web conference 2021. 1086–1097
2021
-
[57]
Morgan Klaus Scheuerman, Jialun Aaron Jiang, Casey Fiesler, and Jed R Brubaker
-
[58]
A framework of severity for harmful content online.Proceedings of the ACM on Human-Computer Interaction5, CSCW2 (2021), 1–33
2021
-
[59]
Ozgur Can Seckin, Bao Tran Truong, Alessandro Flammini, and Filippo Menczer
-
[60]
Identifying Constructive Conflict in Online Discussions through Contro- versial yet Toxicity Resilient Posts.arXiv preprint arXiv:2509.18303(2025)
2025
-
[61]
Cuihua Shen, Peter Monge, and Dmitri Williams. 2014. The evolution of social ties online: A longitudinal study in a massively multiplayer online game.Journal of the Association for Information Science and Technology65, 10 (2014), 2127–2137
2014
-
[62]
Amit Sheth, Valerie L Shalin, and Ugur Kursuncu. 2022. Defining and detecting toxicity on social media: context and knowledge are key.Neurocomputing490 (2022), 312–318
2022
-
[63]
Tom Snijders, Christian Steglich, and Michael Schweinberger. 2017. Modeling the coevolution of networks and behavior. InLongitudinal models in the behavioral and related sciences. Routledge, 41–71
2017
-
[64]
Tingting Song, Qian Tang, and Jinghua Huang. 2019. Triadic closure, homophily, and reciprocation: an empirical investigation of social ties between content providers.Information Systems Research30, 3 (2019), 912–926
2019
-
[65]
Joep Sonnemans, Frans Van Dijk, and Frans Van Winden. 2006. On the dynamics of social ties structures in groups.Journal of economic psychology27, 2 (2006), 187–204
2006
-
[66]
Elizabeth A Stuart. 2010. Matching methods for causal inference: A review and a look forward.Statistical science: a review journal of the Institute of Mathematical Statistics25, 1 (2010), 1
2010
-
[67]
Guangmo Tong, Weili Wu, Ling Guo, Deying Li, Cong Liu, Bin Liu, and Ding-Zhu Du. 2017. An efficient randomized algorithm for rumor blocking in online social networks.IEEE Transactions on Network Science and Engineering7, 2 (2017), 845–854
2017
-
[68]
Jean M Twenge. 2013. Does online social media lead to social connection or social disconnection?Journal of College and Character14, 1 (2013), 11–20
2013
-
[69]
Victor Veitch, Dhanya Sridhar, and David Blei. 2020. Adapting text embeddings for causal inference. InConference on uncertainty in artificial intelligence. PMLR, 919–928
2020
-
[70]
Haozhe Wu, Zhiyuan Hu, Jia Jia, Yaohua Bu, Xiangnan He, and Tat-Seng Chua
-
[71]
InProceedings of the AAAI Conference on Artificial Intelli- gence, Vol
Mining unfollow behavior in large-scale online social networks via spatial- temporal interaction. InProceedings of the AAAI Conference on Artificial Intelli- gence, Vol. 34. 254–261
-
[72]
Bo Xu, Yun Huang, Haewoon Kwak, and Noshir Contractor. 2013. Structures of broken ties: exploring unfollow behavior on twitter. InProceedings of the 2013 conference on Computer supported cooperative work. 871–876
2013
-
[73]
Qinfeng Zhu and Marko M Skoric. 2022. Political implications of disconnection on social media: A study of politically motivated unfriending.New Media & Society24, 12 (2022), 2659–2679
2022
-
[74]
Help is where you find it
Shimrit Zigron and Jenny Bronstein. 2019. “Help is where you find it”: The role of weak ties networks as sources of information and support in virtual health communities.Journal of the Association for Information Science and Technology 70, 2 (2019), 130–139
2019
Reviewed June 26, 2026 · model on record in the stance chip above.
Discussion (0). Continue with ORCID to comment.