Pith. sign in

REVIEW 2 major objections 2 minor 59 references

Recommender logic sets the visibility regime in social media, with popularity creating reinforcement loops that concentrate attention and collaborative filtering distributing it broadly.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review · grok-4.3

2026-07-02 16:32 UTC pith:GTZBWPMH

load-bearing objection Simulations show popularity recommenders concentrate visibility through reinforcement loops while collaborative filtering spreads it, but the YSocial model lacks any reported calibration to real data. the 2 major comments →

arxiv 2607.00258 v1 pith:GTZBWPMH submitted 2026-06-30 cs.SI

Joint Effects of Recommender Systems and Network Structure on the Visibility of Content and Creators

Virginia Morini , Valentina Pansanella , Luca Pappalardo , Dino Pedreschi , Giulio Rossetti This is my paper
classification cs.SI
keywords recommender systemssocial media visibilitynetwork structureagent-based simulationpopularity reinforcementcollaborative filteringcreator inequality
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

The paper uses agent-based simulations to examine how recommendation algorithms and network structures together determine which content and creators gain visibility on social media. It finds that the choice of recommender primarily determines whether visibility concentrates on a few items or spreads across many. A reader would care because these mechanisms affect who gets seen and can influence opportunities for creators in digital spaces. Network position further shapes outcomes under certain recommenders.

Core claim

In simulations using YSocial with seven recommendation strategies and two network topologies, recommender logic sets the visibility regime: popularity creates a reinforcement loop in which early reactions increase later exposure, concentrating visibility on a small subset of content and limiting creator visibility to those whose content enters this loop, while collaborative filtering distributes visibility broadly across the active catalogue and user base. When the follower graph shapes candidate selection, network structure changes the direction of inequality under popularity ranking, redirecting visibility toward already socially popular creators. Network topology modulates the magnitude w

What carries the argument

YSocial agent-based virtual twin that simulates user interactions under seven recommendation strategies and two network topologies to measure visibility allocation.

Load-bearing premise

The chosen agent-based model with its seven strategies and two topologies sufficiently represents the joint effects in actual social media platforms.

What would settle it

A live experiment on a social media platform showing that a popularity-based recommender does not produce greater concentration of visibility on popular content than collaborative filtering would falsify the main claim.

Watch this falsifier. Get emailed when new claim-graph text bears on it.

If this is right

  • Visibility allocation must be assessed across content, creators, network position, and temporal reinforcement.
  • Under popularity ranking, visibility redirects toward creators who are already socially popular when network structure influences candidate selection.
  • Network topology affects the size of visibility effects but not their overall pattern.
  • Controlled simulations allow testing of feed designs for visibility distribution prior to real-world deployment.

Where Pith is reading between the lines

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

  • Platforms might benefit from using similar simulations to anticipate and mitigate visibility inequalities before implementing new algorithms.
  • Real-world networks that evolve over time could amplify or dampen the observed effects compared to static topologies.
  • Extending the model to include dynamic user behaviors or content creation could reveal additional feedback loops.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

2 major / 2 minor

Summary. The manuscript uses agent-based simulations in the YSocial virtual twin to study how seven recommendation strategies interact with two network topologies to shape visibility of content and creators. It claims that popularity-based recommenders induce reinforcement loops that concentrate visibility on a small subset of content while limiting creator visibility to those entering the loop, whereas collaborative filtering distributes visibility broadly; network structure modulates effect magnitudes and, under popularity ranking, redirects visibility toward already socially popular creators without altering the qualitative ordering.

Significance. If the simulation outcomes prove robust, the work would usefully demonstrate that visibility must be assessed across content, creators, network position, and temporal dynamics, and that controlled simulations can inform feed-design choices prior to deployment. The explicit comparison across multiple strategies and topologies is a constructive contribution to the literature on algorithmic visibility.

major comments (2)
  1. [Methods] Methods section (model description): the YSocial agent rules for reaction probabilities, candidate selection, and network evolution are not calibrated or validated against empirical distributions from real platforms (e.g., observed Gini coefficients, temporal reinforcement patterns, or visibility trajectories). Because the central claim that 'recommender logic sets the visibility regime' rests entirely on outcomes from this ungrounded virtual twin, the reported qualitative ordering may be an artifact of the chosen agent parameters rather than a general joint effect.
  2. [Results] Results (visibility-regime comparisons): the claim that network topology 'changes the direction of inequality' under popularity ranking is presented without quantitative tests of whether the redirection effect survives variation in the follower-graph density or the precise definition of 'socially popular' creators; this is load-bearing for the joint-effects conclusion.
minor comments (2)
  1. [Abstract] Abstract: the two network topologies are not named, making it difficult for readers to assess the scope of the reported modulation effect.
  2. [Figures] Figure captions (throughout): several panels lack explicit axis labels for the visibility or Gini metrics, complicating direct comparison across the seven strategies.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive comments on our manuscript. Below we provide point-by-point responses to the major comments. We have revised the manuscript to incorporate additional analyses and clarifications where feasible.

read point-by-point responses

  1. Referee: [Methods] Methods section (model description): the YSocial agent rules for reaction probabilities, candidate selection, and network evolution are not calibrated or validated against empirical distributions from real platforms (e.g., observed Gini coefficients, temporal reinforcement patterns, or visibility trajectories). Because the central claim that 'recommender logic sets the visibility regime' rests entirely on outcomes from this ungrounded virtual twin, the reported qualitative ordering may be an artifact of the chosen agent parameters rather than a general joint effect.

    Authors: We agree that direct calibration and validation against real-world empirical distributions would strengthen the generalizability of our findings. The YSocial virtual twin is constructed based on established social media interaction patterns from the literature, but we acknowledge the absence of platform-specific calibration in the current version. In the revised manuscript, we have expanded the Methods section to include a detailed discussion of parameter selection rationale, drawing from prior empirical studies on social media dynamics. Additionally, we have conducted and reported sensitivity analyses varying reaction probabilities and network evolution rules to demonstrate that the qualitative ordering of visibility regimes remains consistent across reasonable parameter ranges. We believe this addresses the concern that the results are mere artifacts, while noting that full validation would require proprietary data not available for this study. revision: partial

  2. Referee: [Results] Results (visibility-regime comparisons): the claim that network topology 'changes the direction of inequality' under popularity ranking is presented without quantitative tests of whether the redirection effect survives variation in the follower-graph density or the precise definition of 'socially popular' creators; this is load-bearing for the joint-effects conclusion.

    Authors: We appreciate this point and have performed additional robustness checks as suggested. Specifically, we varied the follower-graph density by simulating networks with different average degrees and tested alternative definitions of 'socially popular' creators (e.g., based on degree centrality versus betweenness). The redirection effect under popularity ranking persists in these variations, with network structure continuing to modulate inequality direction without altering the qualitative patterns. These new analyses have been added to the Results section and supplementary materials, reinforcing the joint-effects conclusion. revision: yes

Circularity Check

0 steps flagged

No circularity: results are direct simulation outputs, not self-referential derivations

full rationale

The paper reports outcomes from agent-based simulations in the YSocial virtual twin under explicitly enumerated recommendation strategies and network topologies. No equations, fitted parameters, or predictions are described that reduce the reported visibility regimes or reinforcement loops to the model inputs by construction. The central claims follow from running the defined agent rules rather than from any self-definition, renaming, or load-bearing self-citation chain. The model itself may draw on prior work, but the visibility findings are falsifiable experimental outputs within the stated setup and do not collapse into their own assumptions.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 1 invented entities

Results depend on the domain assumption that the YSocial simulation environment accurately models real social media dynamics; no free parameters or invented entities beyond the simulation framework itself are detailed in the abstract.

axioms (1)
  • domain assumption The YSocial agent-based model with its 7 recommendation strategies and 2 network topologies faithfully represents the joint effects of recommender systems and network structure on visibility in real social media.
    This assumption underpins all reported findings on visibility regimes and inequality patterns.
invented entities (1)
  • YSocial virtual twin no independent evidence
    purpose: Simulated environment for testing recommendation strategies and network topologies
    Introduced as the core experimental platform; no independent evidence provided in abstract.

pith-pipeline@v0.9.1-grok · 5741 in / 1283 out tokens · 43827 ms · 2026-07-02T16:32:12.382046+00:00 · methodology

0 comments
read the original abstract

Social media algorithms allocate users' visibility by ranking content within their social networks. Yet, how recommendation logic and network structure jointly shape visibility across content and creators remains largely understudied. In this work, we tackle this question through agent-based simulations using YSocial, a social media virtual twin, in which agents interact under 7 recommendation strategies and 2 network topologies. We find that recommender logic sets the visibility regime: popularity creates a reinforcement loop in which early reactions increase later exposure, concentrating visibility on a small subset of content and limiting creator visibility to those whose content enters this loop, while collaborative filtering distributes visibility broadly across the active catalogue and user base. When the follower graph shapes candidate selection, network structure changes the direction of inequality: under popularity ranking, creator-level concentration becomes comparable to global popularity, but visibility is systematically redirected toward creators who are already socially popular. Network topology modulates the magnitude of these effects without changing their qualitative ordering. These results show that visibility allocation should be evaluated across content, creators, network position, and temporal reinforcement, and that controlled simulations can help test how feed design distributes visibility before deployment.

Figures

Figures reproduced from arXiv: 2607.00258 by Dino Pedreschi, Giulio Rossetti, Luca Pappalardo, Valentina Pansanella, Virginia Morini.

Figure 1
Figure 1. Figure 1: Recommendation volume distributions (scale-free network). Distribution conditional on contents or creators receiving at least one recommendation. Panels (a) and (b) show global recommender systems per content, p(rco), and per creator, p(rcr), respectively. Panels (c) and (d) report the same quantities for network-aware recommender systems. 4 Results We organize the results around two classes of strategies … view at source ↗
Figure 2
Figure 2. Figure 2: Discrimination and coverage under global recommenders (scale-free network). Panel (a) reports changes in concentration (∆G) and panel (b) reports changes in coverage (∆C), both relative to RC. Hatched bars denote contents; plain bars denote creators. Error bars indicate cross-run standard deviation. sparse but persistent region of contents receives between roughly 850 and 1,050 recommendations, indicating … view at source ↗
Figure 3
Figure 3. Figure 3: Creator visibility for each recommender system (scale-free network). Each panel shows creator visibility in the BA social network under a different feed: (a) RC, (b) P, (c) UCF, (d) F, (e) FP, (f) LR. Node size is proportional to degree; node saturation reflects normalized visibility. Only high-degree or high-visibility nodes are shown. 4.2 Network-aware recommender systems We next examine network-aware re… view at source ↗
Figure 4
Figure 4. Figure 4: Discrimination and coverage under network-aware recommenders (scale-free network). Panel (a) reports changes in concentration (∆G) and panel (b) reports changes in coverage (∆C), both relative to F. Hatched bars denote contents; plain bars denote creators. Error bars indicate cross-run standard deviation. 10 [PITH_FULL_IMAGE:figures/full_fig_p010_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Degree-resolved creator visibility under network-aware recommenders (scale-free network). Panel (a) reports changes in mean recommendation volume (∆¯ra) and panel (b) reports changes in mean unique reach (∆¯ua), both relative to F. Creators are grouped into topology-specific degree bins; bins observed in fewer than three runs are omitted. Shaded areas indicate cross-run variability. Takeaway: The strength … view at source ↗
Figure 6
Figure 6. Figure 6: Popularity reinforcement (scale-free network). Spearman correlation ρtp between early popularity (reactions accumulated up to content age tp) and future recommendation exposure after tp. Panel (a) shows global recommenders (P and UCF); RC is omitted because its correlation is defined only at tp ≈ 2 and provides no meaningful trajectory. Panel (b) shows network-aware recommenders (F, FP, and LR). Values are… view at source ↗
Figure 7
Figure 7. Figure 7: Distribution of recommendation volume on a log–log scale in the ER topology, conditional on contents or [PITH_FULL_IMAGE:figures/full_fig_p018_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Changes in creator visibility by degree for network-aware recommenders in the ER topology, relative to the [PITH_FULL_IMAGE:figures/full_fig_p019_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Creator visibility for each recommender system (random network). Each panel shows creator visibility in the ER social network under a different feed: (a) RC, (b) P, (c) UCF, (d) F, (e) FP, (f) LR. Node size is proportional to degree; node saturation reflects normalized visibility. Only high-degree or high-visibility nodes are shown. 19 [PITH_FULL_IMAGE:figures/full_fig_p019_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: Spearman correlation ρtp between early popularity (reactions accumulated up to content age tp) and future recommendation exposure after tp, for the ER topology. Panel (a) compares global recommenders (RC, P, and UCF); panel (b) compares network-aware recommenders (F, FP, and LR). Values are reported in absolute terms; shaded areas indicate cross-run variability. 20 [PITH_FULL_IMAGE:figures/full_fig_p020_… view at source ↗
Figure 11
Figure 11. Figure 11: Lorenz curves of cumulative visibility for contents and creators under both network topologies. Rows [PITH_FULL_IMAGE:figures/full_fig_p021_11.png] view at source ↗
Figure 12
Figure 12. Figure 12: Changes in creator visibility by degree for global recommenders relative to the reverse-chronological [PITH_FULL_IMAGE:figures/full_fig_p022_12.png] view at source ↗
Figure 13
Figure 13. Figure 13: Visibility distributions for the reverse-chronological baseline RC and the follower-only baseline F. Panel (a): [PITH_FULL_IMAGE:figures/full_fig_p023_13.png] view at source ↗
Figure 14
Figure 14. Figure 14: Visibility distributions for the user–user collaborative filtering recommender (UCF) and the item–item [PITH_FULL_IMAGE:figures/full_fig_p024_14.png] view at source ↗

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

59 extracted references · 59 canonical work pages · 1 internal anchor

  1. [1]

    Multi-sided exposure bias in recommendation.arXiv preprint arXiv:2006.15772, 2020

    Himan Abdollahpouri and Masoud Mansoury. Multi-sided exposure bias in recommendation.arXiv preprint arXiv:2006.15772, 2020

    work page arXiv 2006

  2. [2]

    The unfairness of popularity bias in recommendation.arXiv preprint arXiv:1907.13286, 2019

    Himan Abdollahpouri, Masoud Mansoury, Robin Burke, and Bamshad Mobasher. The unfairness of popularity bias in recommendation.arXiv preprint arXiv:1907.13286, 2019

    work page arXiv 1907

  3. [3]

    Understanding longitudinal dynamics of recommender systems with agent-based modeling and simulation.arXiv preprint arXiv:2108.11068, 2021

    Gediminas Adomavicius, Dietmar Jannach, Stephan Leitner, and Jingjing Zhang. Understanding longitudinal dynamics of recommender systems with agent-based modeling and simulation.arXiv preprint arXiv:2108.11068, 2021

    work page arXiv 2021

  4. [4]

    The attention economy and the impact of artificial intelligence

    Ricardo Baeza-Yates and Usama M Fayyad. The attention economy and the impact of artificial intelligence. In Perspectives on digital humanism, pages 123–134. Springer, 2021

    work page 2021

  5. [5]

    Exposure to ideologically diverse news and opinion on facebook.Science, 348(6239):1130–1132, 2015

    Eytan Bakshy, Solomon Messing, and Lada A Adamic. Exposure to ideologically diverse news and opinion on facebook.Science, 348(6239):1130–1132, 2015

    work page 2015

  6. [6]

    Emergence of scaling in random networks.Science, 286(5439):509–512, 1999

    Albert-Lazlo Barabási and Réka Albert. Emergence of scaling in random networks.Science, 286(5439):509–512, 1999

    work page 1999

  7. [7]

    Auditing algorithmic bias on twitter

    Nathan Bartley, Andres Abeliuk, Emilio Ferrara, and Kristina Lerman. Auditing algorithmic bias on twitter. In Proceedings of the 13th ACM web science conference 2021, pages 65–73, 2021

    work page 2021

  8. [8]

    Impacts of personalization on social network exposure

    Nathan Bartley, Keith Burghardt, and Kristina Lerman. Impacts of personalization on social network exposure. In International Conference on Advances in Social Networks Analysis and Mining, pages 38–53. Springer, 2024

    work page 2024

  9. [9]

    Algorithmic amplification of politics and engagement maximization on social media

    Paul Bouchaud. Algorithmic amplification of politics and engagement maximization on social media. In International conference on complex networks and their applications, pages 131–142. Springer, 2023

    work page 2023

  10. [10]

    Skewed perspectives: examining the influence of engagement maximization on content diversity in social media feeds.Journal of Computational Social Science, 7(1):721–739, 2024

    Paul Bouchaud. Skewed perspectives: examining the influence of engagement maximization on content diversity in social media feeds.Journal of Computational Social Science, 7(1):721–739, 2024

    work page 2024

  11. [11]

    Exploring social network effects on popularity biases in recommender systems

    Rocío Cañamares and Pablo Castells. Exploring social network effects on popularity biases in recommender systems. InRSWeb@ RecSys, 2014

    work page 2014

  12. [12]

    Recommendation system simulations: A discussion of two key challenges.arXiv preprint arXiv:2109.02475, 2021

    Allison JB Chaney. Recommendation system simulations: A discussion of two key challenges.arXiv preprint arXiv:2109.02475, 2021

    work page arXiv 2021

  13. [13]

    Bias and debias in recommender system: A survey and future directions.ACM Transactions on Information Systems, 41(3):1– 39, 2023

    Jiawei Chen, Hande Dong, Xiang Wang, Fuli Feng, Meng Wang, and Xiangnan He. Bias and debias in recommender system: A survey and future directions.ACM Transactions on Information Systems, 41(3):1– 39, 2023

    work page 2023

  14. [14]

    The effect of people recommenders on echo chambers and polarization

    Federico Cinus, Marco Minici, Corrado Monti, and Francesco Bonchi. The effect of people recommenders on echo chambers and polarization. InProceedings of the International AAAI Conference on Web and Social Media, volume 16, pages 90–101, 2022. 13

    work page 2022

  15. [15]

    Princeton University Press, 2012

    Joshua M Epstein.Generative social science: Studies in agent-based computational modeling. Princeton University Press, 2012

    work page 2012

  16. [16]

    On random graphs

    Paul Erd ˝os and Alfred Rényi. On random graphs. i.Publicationes Mathematicae, 6:290–297, 1959

    work page 1959

  17. [17]

    The effect of homophily on disparate visibility of minorities in people recommender systems

    Francesco Fabbri, Francesco Bonchi, Ludovico Boratto, and Carlos Castillo. The effect of homophily on disparate visibility of minorities in people recommender systems. InProceedings of the International AAAI Conference on Web and Social Media, volume 14, pages 165–175, 2020

    work page 2020

  18. [18]

    Exposure inequality in people recommender systems: the long-term effects

    Francesco Fabbri, Maria Luisa Croci, Francesco Bonchi, and Carlos Castillo. Exposure inequality in people recommender systems: the long-term effects. InProceedings of the International AAAI Conference on Web and Social Media, volume 16, pages 194–204, 2022

    work page 2022

  19. [19]

    i’m in the bluesky tonight

    Andrea Failla and Giulio Rossetti. “i’m in the bluesky tonight”: insights from a year worth of social data.PloS one, 19(11):e0310330, 2024

    work page 2024

  20. [20]

    Link recommendations: Their impact on network structure and minorities

    Antonio Ferrara, Lisette Espín-Noboa, Fariba Karimi, and Claudia Wagner. Link recommendations: Their impact on network structure and minorities. InProceedings of the 14th ACM Web Science Conference 2022, pages 228–238. ACM, 2022

    work page 2022

  21. [21]

    Using agent-based modelling to evaluate the impact of algorithmic curation on social media.ACM Journal of Data and Information Quality, 15(1):1–24, 2022

    Anna Gausen, Wayne Luk, and Ce Guo. Using agent-based modelling to evaluate the impact of algorithmic curation on social media.ACM Journal of Data and Information Quality, 15(1):1–24, 2022

    work page 2022

  22. [22]

    Beyond accuracy: evaluating recommender systems by coverage and serendipity

    Mouzhi Ge, Carla Delgado-Battenfeld, and Dietmar Jannach. Beyond accuracy: evaluating recommender systems by coverage and serendipity. InProceedings of the fourth ACM conference on Recommender systems, pages 257–260, 2010

    work page 2010

  23. [23]

    The few-get-richer: a surprising consequence of popularity- based rankings? InThe world wide web conference, pages 2764–2770, 2019

    Fabrizio Germano, Vicenç Gómez, and Gaël Le Mens. The few-get-richer: a surprising consequence of popularity- based rankings? InThe world wide web conference, pages 2764–2770, 2019

    work page 2019

  24. [24]

    Reshares on social media amplify political news but do not detectably affect beliefs or opinions.Science, 381(6656):404–408, 2023

    Andrew M Guess, Neil Malhotra, Jennifer Pan, Pablo Barberá, Hunt Allcott, Taylor Brown, Adriana Crespo- Tenorio, Drew Dimmery, Deen Freelon, Matthew Gentzkow, et al. Reshares on social media amplify political news but do not detectably affect beliefs or opinions.Science, 381(6656):404–408, 2023

    work page 2023

  25. [25]

    How do social media feed algorithms affect attitudes and behavior in an election campaign?Science, 381(6656):398–404, 2023

    Andrew M Guess, Neil Malhotra, Jennifer Pan, Pablo Barberá, Hunt Allcott, Taylor Brown, Adriana Crespo- Tenorio, Drew Dimmery, Deen Freelon, Matthew Gentzkow, et al. How do social media feed algorithms affect attitudes and behavior in an election campaign?Science, 381(6656):398–404, 2023

    work page 2023

  26. [26]

    Muhammad Haroon, Magdalena Wojcieszak, Anshuman Chhabra, Xin Liu, Prasant Mohapatra, and Zubair Shafiq. Auditing youtube’s recommendation system for ideologically congenial, extreme, and problematic recommendations.Proceedings of the National Academy of Sciences, 120(50):e2213020120, 2023

    work page 2023

  27. [27]

    The second wave of attention economics

    Maxi Heitmayer. The second wave of attention economics. attention as a universal symbolic currency on social media and beyond.Interacting with Computers, 37(1):18–29, 2025

    work page 2025

  28. [28]

    Evaluating collaborative filtering recommender systems.ACM Transactions on Information Systems (TOIS), 22(1):5–53, 2004

    Jonathan L Herlocker, Joseph A Konstan, Loren G Terveen, and John T Riedl. Evaluating collaborative filtering recommender systems.ACM Transactions on Information Systems (TOIS), 22(1):5–53, 2004

    work page 2004

  29. [29]

    Causally estimating the effect of youtube’s recommender system using counterfactual bots.Proceedings of the National Academy of Sciences, 121(8), 2024

    Homa Hosseinmardi, Amir Ghasemian, Miguel Rivera-Lanas, Manoel Horta Ribeiro, Robert West, and Duncan J Watts. Causally estimating the effect of youtube’s recommender system using counterfactual bots.Proceedings of the National Academy of Sciences, 121(8), 2024

    work page 2024

  30. [30]

    Algorithmic amplification of politics on twitter.Proceedings of the National Academy of Sciences, 119(1):e2025334119, 2022

    Ferenc Huszár, Sofia Ira Ktena, Conor O’Brien, Luca Belli, Andrew Schlaikjer, and Moritz Hardt. Algorithmic amplification of politics on twitter.Proceedings of the National Academy of Sciences, 119(1):e2025334119, 2022

    work page 2022

  31. [31]

    Visibility allocation systems: How algorithmic design shapes online visibility and societal outcomes

    Stefania Ionescu, Robin Forsberg, Elsa Lichtenegger, Salima Jaoua, Kshitijaa Jaglan, Florian Dorfler, and Aniko Hannak. Visibility allocation systems: How algorithmic design shapes online visibility and societal outcomes. arXiv preprint arXiv:2510.17241, 2025

    work page arXiv 2025

  32. [32]

    A survey on popularity bias in recommender systems.User Modeling and User-Adapted Interaction, 34(5):1777–1834, 2024

    Anastasiia Klimashevskaia, Dietmar Jannach, Mehdi Elahi, and Christoph Trattner. A survey on popularity bias in recommender systems.User Modeling and User-Adapted Interaction, 34(5):1777–1834, 2024

    work page 2024

  33. [33]

    Impact of recommender systems on sales volume and diversity

    Dokyun Lee and Kartik Hosanagar. Impact of recommender systems on sales volume and diversity. InProceedings of the 2014 International Conference on Information Systems (ICIS 2014). Association for Information Systems, 2014

    work page 2014

  34. [34]

    Technology and democracy: A paradox wrapped in a contradiction inside an irony.Memory, mind & media, 1:e5, 2022

    Stephan Lewandowsky and Peter Pomerantsev. Technology and democracy: A paradox wrapped in a contradiction inside an irony.Memory, mind & media, 1:e5, 2022

    work page 2022

  35. [35]

    T-recs: A simulation tool to study the societal impact of recommender systems.arXiv preprint arXiv:2107.08959, 2021

    Eli Lucherini, Matthew Sun, Amy Winecoff, and Arvind Narayanan. T-recs: A simulation tool to study the societal impact of recommender systems.arXiv preprint arXiv:2107.08959, 2021. 14

    work page arXiv 2021

  36. [36]

    The urban impact of ai: modelling feedback loops in location-based recommender systems.Machine Learning, 115(1):19, 2026

    Giovanni Mauro, Marco Minici, and Luca Pappalardo. The urban impact of ai: modelling feedback loops in location-based recommender systems.Machine Learning, 115(1):19, 2026

    work page 2026

  37. [37]

    Social drivers and algorithmic mechanisms on digital media.Perspectives on Psychological Science, 19(5):735–748, 2024

    Hannah Metzler and David Garcia. Social drivers and algorithmic mechanisms on digital media.Perspectives on Psychological Science, 19(5):735–748, 2024

    work page 2024

  38. [38]

    Exploring the filter bubble: the effect of using recommender systems on content diversity

    Tien T Nguyen, Pik-Mai Hui, F Maxwell Harper, Loren Terveen, and Joseph A Konstan. Exploring the filter bubble: the effect of using recommender systems on content diversity. InProceedings of the 23rd international conference on World wide web, pages 677–686. ACM, 2014

    work page 2014

  39. [39]

    Like-minded sources on facebook are prevalent but not polarizing.Nature, 620(7972):137–144, 2023

    Brendan Nyhan, Jaime Settle, Emily Thorson, Magdalena Wojcieszak, Pablo Barberá, Annie Y Chen, Hunt Allcott, Taylor Brown, Adriana Crespo-Tenorio, Drew Dimmery, et al. Like-minded sources on facebook are prevalent but not polarizing.Nature, 620(7972):137–144, 2023

    work page 2023

  40. [40]

    Modeling individual attention dynamics on online social media.arXiv preprint arXiv:2507.01511, 2025

    Jaume Ojer, Filippo Radicchi, Santo Fortunato, Michele Starnini, and Romualdo Pastor-Satorras. Modeling individual attention dynamics on online social media.arXiv preprint arXiv:2507.01511, 2025

    work page arXiv 2025

  41. [41]

    From mean-field to complex topologies: Network effects on the algorithmic bias model

    Valentina Pansanella, Giulio Rossetti, and Letizia Milli. From mean-field to complex topologies: Network effects on the algorithmic bias model. InComplex Networks, volume 1016 ofStudies in Computational Intelligence. Springer, 2021

    work page 2021

  42. [42]

    Modeling algorithmic bias: simplicial complexes and evolving network topologies.Applied Network Science, 7(1):57, 2022

    Valentina Pansanella, Giulio Rossetti, and Letizia Milli. Modeling algorithmic bias: simplicial complexes and evolving network topologies.Applied Network Science, 7(1):57, 2022

    work page 2022

  43. [43]

    A survey on the impacts of recommender systems on users, items, and human-ai ecosystems.arXiv preprint arXiv:2407.01630, 2024

    Luca Pappalardo, Salvatore Citraro, Giuliano Cornacchia, Mirco Nanni, Valentina Pansanella, Giulio Rossetti, Gizem Gezici, Fosca Giannotti, Margherita Lalli, Giovanni Mauro, et al. A survey on the impacts of recommender systems on users, items, and human-ai ecosystems.arXiv preprint arXiv:2407.01630, 2024

    work page arXiv 2024

  44. [44]

    Human-ai coevolution

    Dino Pedreschi, Luca Pappalardo, Emanuele Ferragina, Ricardo Baeza-Yates, Albert-László Barabási, Frank Dignum, Virginia Dignum, Tina Eliassi-Rad, Fosca Giannotti, János Kertész, et al. Human-ai coevolution. Artificial Intelligence, 339:104244, 2025

    work page 2025

  45. [45]

    Opinion formation on social networks with algorithmic bias: dynamics and bias imbalance.Journal of Physics: Complexity, 2(4), 2021

    Antonio F Peralta, János Kertész, and Gerardo Iñiguez. Opinion formation on social networks with algorithmic bias: dynamics and bias imbalance.Journal of Physics: Complexity, 2(4), 2021

    work page 2021

  46. [46]

    Effect of algorithmic bias and network structure on coexistence, consensus, and polarization of opinions.Physical Review E, 104(4):044312, 2021

    Antonio F Peralta, Matteo Neri, János Kertész, and Gerardo Iñiguez. Effect of algorithmic bias and network structure on coexistence, consensus, and polarization of opinions.Physical Review E, 104(4):044312, 2021

    work page 2021

  47. [47]

    Modelling opinion dynamics in the age of algorithmic personalisation.Scientific reports, 9(1):7261, 2019

    Nicola Perra and Luis EC Rocha. Modelling opinion dynamics in the age of algorithmic personalisation.Scientific reports, 9(1):7261, 2019

    work page 2019

  48. [48]

    Cambridge University Press Cambridge, UK, 2016

    Márton Pósfai and Albert-László Barabási.Network science, volume 3. Cambridge University Press Cambridge, UK, 2016

    work page 2016

  49. [49]

    Auditing radicalization pathways on youtube

    Manoel Horta Ribeiro, Raphael Ottoni, Robert West, Virgílio AF Almeida, and Wagner Meira Jr. Auditing radicalization pathways on youtube. InProceedings of the 2020 conference on fairness, accountability, and transparency, pages 131–141. ACM, 2020

    work page 2020

  50. [50]

    The amplification paradox in recommender systems

    Manoel Horta Ribeiro, Veniamin Veselovsky, and Robert West. The amplification paradox in recommender systems. InProceedings of the International AAAI Conference on Web and Social Media, volume 17. AAAI Press, 2023

    work page 2023

  51. [51]

    Y social: an llm-powered social media digital twin

    Giulio Rossetti, Massimo Stella, Rémy Cazabet, Katherine Abramski, Erica Cau, Salvatore Citraro, Andrea Failla, Riccardo Improta, Virginia Morini, and Valentina Pansanella. Y social: an llm-powered social media digital twin. arXiv preprint arXiv:2408.00818, 2024

    work page arXiv 2024

  52. [52]

    The closed loop between opinion formation and personalized recommendations.IEEE Transactions on Control of Network Systems, 9(3):1092–1103, 2021

    Wilbert Samuel Rossi, Jan Willem Polderman, and Paolo Frasca. The closed loop between opinion formation and personalized recommendations.IEEE Transactions on Control of Network Systems, 9(3):1092–1103, 2021

    work page 2021

  53. [53]

    The Prosocial Ranking Challenge: Reducing Polarization on Social Media without Sacrificing Engagement

    Jonathan Stray, Ian Baker, George Beknazar-Yuzbashev, Ceren Budak, Julia Kamin, Kylan Rutherford, Mateusz Stalinski, Tin Acosta, Chris Bail, Michael Bernstein, et al. The prosocial ranking challenge: Reducing polarization on social media without sacrificing engagement.arXiv preprint arXiv:2603.19626, 2026

    work page arXiv 2026

  54. [54]

    Algorithmic bias amplifies opinion fragmentation and polarization: A bounded confidence model.PLOS ONE, 14(3):1–20, 03 2019

    Alina Sîrbu, Dino Pedreschi, Fosca Giannotti, and János Kertész. Algorithmic bias amplifies opinion fragmentation and polarization: A bounded confidence model.PLOS ONE, 14(3):1–20, 03 2019

    work page 2019

  55. [55]

    The drivers of online polarization: Fitting models to data.Information Sciences, 642:119152, 2023

    Carlo M Valensise, Matteo Cinelli, and Walter Quattrociocchi. The drivers of online polarization: Fitting models to data.Information Sciences, 642:119152, 2023

    work page 2023

  56. [56]

    Improving sales diversity by recommending users to items

    Saúl Vargas and Pablo Castells. Improving sales diversity by recommending users to items. InProceedings of the 8th ACM Conference on Recommender systems, pages 145–152, 2014. 15

    work page 2014

  57. [57]

    Auditing political exposure bias: Algorithmic amplification on twitter/x during the 2024 us presidential election

    Jinyi Ye, Luca Luceri, and Emilio Ferrara. Auditing political exposure bias: Algorithmic amplification on twitter/x during the 2024 us presidential election. InProceedings of the 2025 ACM Conference on Fairness, Accountability, and Transparency, pages 2349–2362. ACM, 2025

    work page 2024

  58. [58]

    The impact of youtube recommendation system on video views

    Renjie Zhou, Samamon Khemmarat, and Lixin Gao. The impact of youtube recommendation system on video views. InProceedings of the 10th ACM SIGCOMM conference on Internet measurement, pages 404–410, 2010

    work page 2010

  59. [59]

    Attention Inequality in Social Media

    Linhong Zhu and Kristina Lerman. Attention inequality in social media.arXiv preprint arXiv:1601.07200, 2016. A Appendix A - Linear Ranker Details Linear Ranker implementation details.The Linear Ranker (LR) follows a two-stage recommendation procedure. First, it generates a candidate set by taking the union of three pools: a reverse-chronological follower ...

    work page internal anchor Pith review Pith/arXiv arXiv 2016