pith. machine review for the scientific record. sign in

arxiv: 2605.09197 · v1 · submitted 2026-05-09 · 💻 cs.SI

Recognition: no theorem link

An Experimental Method to Study Opinion Diffusion in Human-AI Hybrid Societies

Antoine Jardin, Diana Mangalagiu, Elif \c{C}elen, L\'ena Gaubert, Nori Jacoby, Raja Marjieh, R\'emi Devaux

Pith reviewed 2026-05-12 03:41 UTC · model grok-4.3

classification 💻 cs.SI
keywords opinion dynamicshuman-AI interactionpolarizationsocial networksexperimental methodLLM agentsgrid lattice
0
0 comments X

The pith

Hybrid human-AI networks reach lower polarization than human-only networks after repeated opinion updates on a grid.

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

The paper introduces an experimental method that places human participants and AI agents together in small 5 by 5 grid networks and tracks how their statements on a polarizing topic change over eight rounds. Three conditions are run in parallel: groups made only of humans, only of AI agents, and equal mixtures of both. The mixed groups finish with the smallest spread of opposing views, while human-only groups retain more division and show weaker alignment between neighbors. The setup also tests different ways of instructing the AI agents through prompts to see whether those instructions alter how fast or how far the group converges. Such controlled networks give researchers a direct way to observe whether introducing AI changes the trajectory of opinion formation.

Core claim

In 5x5 grid lattice networks, hybrid groups containing equal shares of humans and AI agents reach the lowest final polarization after eight iterative rounds of statement selection and revision, while human-only networks display higher polarization together with lower neighbor agreement.

What carries the argument

The 5x5 grid lattice network in which participants iteratively select and revise statements on a given polarizing topic, with AI agents implemented through large language model prompts.

If this is right

  • Hybrid human-AI networks can produce lower final polarization than human-only networks under repeated opinion updates.
  • Changes in how AI agents are prompted can shift the observed convergence patterns.
  • Grid-based experimental networks isolate the effect of AI presence on human opinion dynamics.

Where Pith is reading between the lines

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

  • The method could be extended to vary network size or run more rounds to test whether the polarization reduction persists.
  • If the pattern holds, platforms might explore limited AI participation to reduce echo-chamber effects on specific topics.
  • Results may still depend on the choice of polarizing topic and the exact prompt wording given to the AI.

Load-bearing premise

The small grid size, fixed number of rounds, and use of prompted AI agents sufficiently represent how opinions actually spread in larger human-AI societies.

What would settle it

A follow-up experiment on larger networks or with real deployed AI tools that finds hybrid groups do not end with lower polarization than human-only groups.

Figures

Figures reproduced from arXiv: 2605.09197 by Antoine Jardin, Diana Mangalagiu, Elif \c{C}elen, L\'ena Gaubert, Nori Jacoby, Raja Marjieh, R\'emi Devaux.

Figure 1
Figure 1. Figure 1: Experimental design. A. Humans or AI agents were embedded in a 5 × 5 social network that evolved over eight iterations. The network was initialized with positive opinions in the upper half and negative opinions in the lower half (supporting vs. not supporting the central claim). In each trial, participants viewed three to five opinions from neighboring nodes in the previous iteration (up, down, left, right… view at source ↗
Figure 2
Figure 2. Figure 2: Example statements. Example statements from the three conditions at the third and eighth iterations, shown. Method Experiment Design We conducted large-scale online experiments in which partici￾pants (human, hybrid, or fully AI) were embedded in a directed unweighted 5×5 grid lattice social network (25 nodes; figure 1A). All networks were initialized with the question: Does red meat cause cancer and cardio… view at source ↗
Figure 3
Figure 3. Figure 3: Example trajectories. Statements are annotated by an LLM as positive (blue), negative (red), or neutral (green) in regards to the given question. Experimental Conditions, Participants, and AI calls We compared three experimental conditions (Figure 1B): (1) Human-only, in which all network nodes were occupied by human participants (𝑁 = 200); (2) AI-only, in which all nodes were populated by Grok-4-Fast agen… view at source ↗
Figure 4
Figure 4. Figure 4: presents the evolution of polarization (𝑃𝑧 ) and NCI over iterations for the three experimental conditions. The AI only condition was run 21 times where the human and hy￾brid experiment were run only once. All conditions exhibited an initial decrease in polarization, indicating opinion conver￾gence over the course of the experiment. Changes in NCI were more variable, but followed a similar overall downward… view at source ↗
Figure 5
Figure 5. Figure 5: Experimental results. Evolution of Mean Polariza￾tion (top) and Mean Neighbor Correlation Index (bottom) over iterations in the Grok-only condition across the two instruc￾tion framings: opinion (green) and consensus (dark orange). Shaded area represent +/−1 standard error of the mean across 21 runs per condition. Explaining Opinion Convergence in AI Networks To examine how task framing and prompt variation… view at source ↗
read the original abstract

As artificial intelligence increasingly mediates public discourse, it becomes important to understand how human-AI collectives shape opinion formation, deliberation, and democratic outcomes. We present a novel experimental method for studying opinion dynamics in hybrid human-AI social networks. Participants, human or AI, were embedded in $5\times5$ grid lattice networks and iteratively asked to select and revise statements on a given polarizing topic over eight rounds. We compared three conditions: human-only, AI-only, and hybrid networks with equal proportions of human and AI participants. Hybrid human-AI networks achieved the lowest final polarization while, in contrast, human-only networks exhibited higher polarization with lower neighbor agreement. We also ran additional experiments varying Large Language Model (LLM) prompt framing to explore whether instruction design might influence convergence patterns. Although these early findings are preliminary and cannot yet support broad generalizations, they highlight the potential value of experimental social networks for understanding opinion dynamics in human-AI hybrid societies.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

3 major / 1 minor

Summary. The manuscript presents a novel experimental method to study opinion diffusion in human-AI hybrid societies. Participants, either human or AI agents powered by LLMs, are placed in 5x5 grid lattice networks and iteratively select and revise statements on polarizing topics over eight rounds. The study compares three conditions: human-only, AI-only, and hybrid networks with equal proportions of humans and AIs. Key findings indicate that hybrid networks result in the lowest final polarization, whereas human-only networks show higher polarization and lower neighbor agreement. Additional experiments explore variations in LLM prompt framing.

Significance. If the reported differences in polarization hold under more rigorous conditions, this work could provide an important experimental framework for investigating how AI integration affects opinion formation and deliberation in social networks. It offers preliminary evidence that hybrid human-AI setups might mitigate polarization compared to purely human groups, which has implications for designing AI-mediated public discourse platforms. The method itself is a strength as an initial step toward controlled studies in this emerging area.

major comments (3)
  1. [Abstract] The abstract states that hybrid human-AI networks achieved the lowest final polarization while human-only networks exhibited higher polarization with lower neighbor agreement, but provides no sample sizes, exact definition of the polarization metric, statistical tests, controls for individual differences, or summaries of raw data, rendering the comparative claims unverifiable.
  2. [Experimental Design] The choice of a 5×5 grid lattice with only eight iterative rounds limits the emergence of global mixing or cascade effects characteristic of real opinion diffusion on larger platforms, since the network diameter is small (4–8 hops) and local neighbor influence dominates.
  3. [AI Agent Model] Since LLM agents are deterministic given fixed prompts, the observed convergence in hybrid and AI-only conditions may reflect prompt-induced averaging rather than the stochastic, identity-driven processes in human groups; the paper reports exploratory prompt variants but without per-run variance or sensitivity analyses.
minor comments (1)
  1. The abstract mentions 'additional experiments varying Large Language Model (LLM) prompt framing' but does not specify the exact framings or their quantitative impacts.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for their thoughtful and constructive comments, which have helped us identify areas to strengthen the manuscript. We address each major comment below and describe the planned revisions.

read point-by-point responses

  1. Referee: [Abstract] The abstract states that hybrid human-AI networks achieved the lowest final polarization while human-only networks exhibited higher polarization with lower neighbor agreement, but provides no sample sizes, exact definition of the polarization metric, statistical tests, controls for individual differences, or summaries of raw data, rendering the comparative claims unverifiable.

    Authors: We agree that the abstract requires additional detail to support verifiability. In the revised version we will incorporate the sample sizes (number of networks and participants per condition), a concise definition of the polarization metric (network-level opinion variance), reference to the statistical tests performed (e.g., ANOVA with reported p-values), mention of baseline controls for individual opinion differences, and a brief note on raw-data consistency across runs. These additions will be made while remaining within abstract length constraints. revision: yes

  2. Referee: [Experimental Design] The choice of a 5×5 grid lattice with only eight iterative rounds limits the emergence of global mixing or cascade effects characteristic of real opinion diffusion on larger platforms, since the network diameter is small (4–8 hops) and local neighbor influence dominates.

    Authors: We acknowledge the limitation of the chosen scale. The 5×5 grid and eight-round design were selected to enable precise observation and coding of every local interaction in a controlled setting. We will add an explicit limitations paragraph in the Discussion section that notes the restricted network diameter and the consequent focus on local rather than global dynamics, and we will outline planned extensions to larger or alternative topologies for future work. The core experimental design itself will not be altered, as it fulfills the paper’s stated goal of providing an initial controlled method. revision: partial

  3. Referee: [AI Agent Model] Since LLM agents are deterministic given fixed prompts, the observed convergence in hybrid and AI-only conditions may reflect prompt-induced averaging rather than the stochastic, identity-driven processes in human groups; the paper reports exploratory prompt variants but without per-run variance or sensitivity analyses.

    Authors: We appreciate the distinction drawn between prompt-driven and stochastic processes. Although we used a non-zero temperature (0.8) during generation to introduce sampling variability, we agree that variance reporting and sensitivity checks are needed. The revision will add (i) per-run standard deviations for polarization and agreement metrics, (ii) a sensitivity analysis across temperature values and prompt-framing variants, and (iii) a short discussion clarifying how these controls help separate prompt effects from human-like stochasticity. These elements will appear in the Results and Methods sections. revision: yes

Circularity Check

0 steps flagged

No circularity: purely empirical experimental report

full rationale

The paper is an experimental report describing participant trials in 5x5 grids over eight rounds across human-only, AI-only, and hybrid conditions. No equations, derivations, model predictions, fitted parameters, or self-citations appear in the provided text. Results are presented as direct observations from the setup rather than reductions from prior claims or definitions. The analysis is therefore self-contained with no load-bearing steps that reduce to inputs by construction.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The central claim depends on the assumption that LLM-prompted agents can stand in for AI participants in opinion networks and that the small-grid iterative design captures relevant diffusion dynamics.

axioms (1)
  • domain assumption LLM agents given appropriate prompts can simulate human-like opinion revision and neighbor influence in a social network setting.
    The hybrid condition and prompt-framing experiments rest on this to produce comparable behavior to humans.

pith-pipeline@v0.9.0 · 5486 in / 1260 out tokens · 67145 ms · 2026-05-12T03:41:47.578066+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

73 extracted references · 73 canonical work pages · 3 internal anchors

  1. [1]

    Epidemic modeling with generative agents

    Epidemic modeling with generative agents , author=. arXiv preprint arXiv:2307.04986 , year=

    work page arXiv

  2. [2]

    OpenRouter.ai: Unified API for AI Models , author=

    work page

  3. [3]

    arXiv preprint arXiv:2502.00879 , year=

    Generating computational cognitive models using large language models , author=. arXiv preprint arXiv:2502.00879 , year=

    work page arXiv

  4. [4]

    Large Language Models for Scientific and Societal Advances , year=

    Collective Social Behaviors in LLMs: An Analysis of LLMs Social Networks , author=. Large Language Models for Scientific and Societal Advances , year=

    work page

  5. [5]

    Scientific Reports , volume=

    Humans create more novelty than ChatGPT when asked to retell a story , author=. Scientific Reports , volume=. 2024 , publisher=

    work page 2024

  6. [6]

    Proceedings-british academy , volume=

    Why culture is common, but cultural evolution is rare , author=. Proceedings-british academy , volume=. 1996 , organization=

    work page 1996

  7. [7]

    arXiv preprint arXiv:2409.11112 , year=

    Strategic Insights in Human and Large Language Model Tactics at Word Guessing Games , author=. arXiv preprint arXiv:2409.11112 , year=

    work page arXiv

  8. [8]

    arXiv preprint arXiv:2501.10476 , year=

    Revisiting Rogers' Paradox in the Context of Human-AI Interaction , author=. arXiv preprint arXiv:2501.10476 , year=

    work page arXiv

  9. [9]

    Nature human behaviour , volume=

    Building machines that learn and think with people , author=. Nature human behaviour , volume=. 2024 , publisher=

    work page 2024

  10. [10]

    Gibbs Sampling with People , url =

    Harrison, Peter and Marjieh, Raja and Adolfi, Federico and van Rijn, Pol and Anglada-Tort, Manuel and Tchernichovski, Ofer and Larrouy-Maestri, Pauline and Jacoby, Nori , booktitle =. Gibbs Sampling with People , url =

    work page

  11. [11]

    Nature Human Behaviour , volume=

    A new sociology of humans and machines , author=. Nature Human Behaviour , volume=. 2024 , publisher=

    work page 2024

  12. [12]

    Science , volume=

    AI can help humans find common ground in democratic deliberation , author =. Science , volume=. 2024 , publisher=

    work page 2024

  13. [13]

    UMAP: Uniform Manifold Approximation and Projection for Dimension Reduction

    Umap: Uniform manifold approximation and projection for dimension reduction , author=. arXiv preprint arXiv:1802.03426 , year=

    work page internal anchor Pith review Pith/arXiv arXiv

  14. [14]

    Efficient Estimation of Word Representations in Vector Space

    Efficient estimation of word representations in vector space , author=. arXiv preprint arXiv:1301.3781 , year=

    work page internal anchor Pith review Pith/arXiv arXiv

  15. [15]

    Science advances , volume=

    Generative AI enhances individual creativity but reduces the collective diversity of novel content , author=. Science advances , volume=. 2024 , publisher=

    work page 2024

  16. [16]

    Nature Human Behaviour , pages=

    A large-scale comparison of divergent creativity in humans and large language models , author=. Nature Human Behaviour , pages=. 2025 , publisher=

    work page 2025

  17. [17]

    Towards an AI co-scientist

    Towards an AI co-scientist , author=. arXiv preprint arXiv:2502.18864 , year=

    work page internal anchor Pith review Pith/arXiv arXiv

  18. [18]

    Nature , volume=

    AI can spark creativity—if we ask it how, not what, to think , author=. Nature , volume=. 2026 , publisher=

    work page 2026

  19. [19]

    Nature communications , volume=

    Simple autonomous agents can enhance creative semantic discovery by human groups , author=. Nature communications , volume=. 2024 , publisher=

    work page 2024

  20. [20]

    Trends in cognitive sciences , volume=

    Exploration versus exploitation in space, mind, and society , author=. Trends in cognitive sciences , volume=. 2015 , publisher=

    work page 2015

  21. [21]

    , author=

    Optimal foraging in semantic memory. , author=. Psychological review , volume=. 2012 , publisher=

    work page 2012

  22. [22]

    arXiv preprint arXiv:2310.13018 , year=

    Getting aligned on representational alignment , author=. arXiv preprint arXiv:2310.13018 , year=

    work page arXiv

  23. [23]

    Administrative science quarterly , volume=

    The network structure of exploration and exploitation , author=. Administrative science quarterly , volume=. 2007 , publisher=

    work page 2007

  24. [24]

    Journal of the Royal Society Interface , volume=

    Individual exploration and selective social learning: balancing exploration--exploitation trade-offs in collective foraging , author=. Journal of the Royal Society Interface , volume=. 2022 , publisher=

    work page 2022

  25. [25]

    Nature , volume=

    Locally noisy autonomous agents improve global human coordination in network experiments , author=. Nature , volume=. 2017 , publisher=

    work page 2017

  26. [26]

    Nature Computational Science , volume=

    Using LLMs to advance the cognitive science of collectives , author=. Nature Computational Science , volume=. 2025 , publisher=

    work page 2025

  27. [27]

    Nature Human Behaviour , volume=

    Machine culture , author=. Nature Human Behaviour , volume=. 2023 , publisher=

    work page 2023

  28. [28]

    Nature , pages=

    Artificial intelligence tools expand scientists’ impact but contract science’s focus , author=. Nature , pages=. 2026 , publisher=

    work page 2026

  29. [29]

    Scientific Reports , volume=

    AI-generated poetry is indistinguishable from human-written poetry and is rated more favorably , author=. Scientific Reports , volume=. 2024 , publisher=

    work page 2024

  30. [30]

    Nature , pages=

    A foundation model to predict and capture human cognition , author=. Nature , pages=. 2025 , publisher=

    work page 2025

  31. [31]

    2020 , howpublished =

    PsyNet: An Online Framework for Behavioral Experiments , author =. 2020 , howpublished =

    work page 2020

  32. [32]

    Behavior Research Methods , volume=

    Empirica: a virtual lab for high-throughput macro-level experiments , author=. Behavior Research Methods , volume=. 2021 , publisher=

    work page 2021

  33. [33]

    arXiv preprint arXiv:2406.04278 , year=

    Characterizing similarities and divergences in conversational tones in humans and llms by sampling with people , author=. arXiv preprint arXiv:2406.04278 , year=

    work page arXiv

  34. [34]

    Findings of the Association for Computational Linguistics: EMNLP 2025 , pages=

    Agent laboratory: Using llm agents as research assistants , author=. Findings of the Association for Computational Linguistics: EMNLP 2025 , pages=. 2025 , publisher=

    work page 2025

  35. [35]

    nature , volume=

    Highly accurate protein structure prediction with AlphaFold , author=. nature , volume=. 2021 , publisher=

    work page 2021

  36. [36]

    Nature , volume=

    Autonomous chemical research with large language models , author=. Nature , volume=. 2023 , publisher=

    work page 2023

  37. [37]

    Proceedings of the National Academy of Sciences , volume=

    Bayesian modeling of human--AI complementarity , author=. Proceedings of the National Academy of Sciences , volume=. 2022 , publisher=

    work page 2022

  38. [38]

    arXiv preprint arXiv:2502.12847 , year=

    Characterizing the Interaction of Cultural Evolution Mechanisms in Experimental Social Networks , author=. arXiv preprint arXiv:2502.12847 , year=

    work page arXiv

  39. [39]

    Scientific Reports , volume=

    Divergent creativity in humans and large language models , author=. Scientific Reports , volume=. 2026 , publisher=

    work page 2026

  40. [40]

    Advances in neural information processing systems , volume=

    Gibbs sampling with people , author=. Advances in neural information processing systems , volume=

    work page

  41. [41]

    Nat Hum Behav , author =

    Commonality and variation in mental representations of music revealed by a cross-cultural comparison of rhythm priors in 15 countries , volume =. Nat Hum Behav , author =. 2024 , pages =. doi:10.1038/s41562-023-01800-9 , abstract =

    work page doi:10.1038/s41562-023-01800-9 2024

  42. [42]

    Cognitive Psychology , author =

    Uncovering mental representations with. Cognitive Psychology , author =. 2010 , pages =. doi:10.1016/j.cogpsych.2009.07.001 , language =

    work page doi:10.1016/j.cogpsych.2009.07.001 2010

  43. [43]

    Shiiku, Shota and Marjieh, Raja and Anglada-Tort, Manuel and Jacoby, Nori , month = may, year =. The. doi:10.48550/arXiv.2502.17962 , abstract =

    work page doi:10.48550/arxiv.2502.17962

  44. [44]

    URL https://www.science.org/doi/abs/10.1126/ science.adq2852

    Science , author =. 2024 , pages =. doi:10.1126/science.adq2852 , abstract =

    work page doi:10.1126/science.adq2852 2024

  45. [45]

    doi:10.48550/arXiv.2506.06366 , abstract =

    Chen, Lin and Zhang, Yunke and Feng, Jie and Chai, Haoye and Zhang, Honglin and Fan, Bingbing and Ma, Yibo and Zhang, Shiyuan and Li, Nian and Liu, Tianhui and Sukiennik, Nicholas and Zhao, Keyu and Li, Yu and Liu, Ziyi and Xu, Fengli and Li, Yong , month = jun, year =. doi:10.48550/arXiv.2506.06366 , abstract =

    work page doi:10.48550/arxiv.2506.06366

  46. [46]

    A data-driven kinetic model for opinion dynamics with social network contacts , volume =. Eur. J. Appl. Math , author =. 2025 , pages =. doi:10.1017/S0956792524000068 , abstract =

    work page doi:10.1017/s0956792524000068 2025

  47. [47]

    Conflicting narratives and polarization on social media , url =

    Pournaki, Armin , month = jul, year =. Conflicting narratives and polarization on social media , url =. doi:10.48550/arXiv.2507.15600 , abstract =

    work page doi:10.48550/arxiv.2507.15600

  48. [48]

    Examining. Proc. ACM Hum.-Comput. Interact. , author =. 2025 , pages =. doi:10.1145/3757591 , abstract =

    work page doi:10.1145/3757591 2025

  49. [49]

    Potter, Yujin and Lai, Shiyang and Kim, Junsol and Evans, James and Song, Dawn , month = nov, year =. Hidden. doi:10.48550/arXiv.2410.24190 , abstract =

    work page doi:10.48550/arxiv.2410.24190

  50. [50]

    How will advanced

    Summerfield, Christopher and Argyle, Lisa and Bakker, Michiel and Collins, Teddy and Durmus, Esin and Eloundou, Tyna and Gabriel, Iason and Ganguli, Deep and Hackenburg, Kobi and Hadfield, Gillian and Hewitt, Luke and Huang, Saffron and Landemore, Helene and Marchal, Nahema and Ovadya, Aviv and Procaccia, Ariel and Risse, Mathias and Schneier, Bruce and S...

    work page doi:10.48550/arxiv.2409.06729

  51. [51]

    and Busby, Ethan C

    Argyle, Lisa P. and Busby, Ethan C. and Gubler, Joshua R. and Hepner, Bryce and Lyman, Alex and Wingate, David , month = apr, year =. Arti-"fickle". doi:10.48550/arXiv.2504.03822 , abstract =

    work page doi:10.48550/arxiv.2504.03822

  52. [52]

    AI & Soc , author =

    Hybrid collective intelligence in a human–. AI & Soc , author =. 2021 , pages =. doi:10.1007/s00146-020-01005-y , language =

    work page doi:10.1007/s00146-020-01005-y 2021

  53. [53]

    Mnemonic convergence in social networks:. Proc. Natl. Acad. Sci. U.S.A. , author =. 2016 , pages =. doi:10.1073/pnas.1525569113 , abstract =

    work page doi:10.1073/pnas.1525569113 2016

  54. [54]

    Donkers, Tim and Ziegler, Jürgen , month = jun, year =. Human-. doi:10.48550/arXiv.2502.01340 , abstract =

    work page doi:10.48550/arxiv.2502.01340

  55. [55]

    Knowl Inf Syst , author =

    Artificial intelligence in the battle against disinformation and misinformation: a systematic review of challenges and approaches , volume =. Knowl Inf Syst , author =. 2025 , pages =. doi:10.1007/s10115-024-02337-7 , language =

    work page doi:10.1007/s10115-024-02337-7 2025

  56. [56]

    Decoding

    Wang, Chenxi and Liu, Zongfang and Yang, Dequan and Chen, Xiuying , year =. Decoding. doi:10.48550/ARXIV.2409.19338 , abstract =

    work page doi:10.48550/arxiv.2409.19338

  57. [57]

    Human Behavior and Emerging Technologies , author =

    The. Human Behavior and Emerging Technologies , author =. 2023 , pages =. doi:10.1155/2023/9094933 , abstract =

    work page doi:10.1155/2023/9094933 2023

  58. [58]

    Opinion cascades and the unpredictability of partisan polarization , volume =. Sci. Adv. , author =. 2019 , pages =. doi:10.1126/sciadv.aax0754 , abstract =

    work page doi:10.1126/sciadv.aax0754 2019

  59. [59]

    Network structure shapes consensus dynamics through individual decisions , volume =. Proc. Natl. Acad. Sci. U.S.A. , author =. 2026 , pages =. doi:10.1073/pnas.2520483123 , abstract =

    work page doi:10.1073/pnas.2520483123 2026

  60. [60]

    Why more social interactions lead to more polarization in societies , volume =. Proc. Natl. Acad. Sci. U.S.A. , author =. 2025 , pages =. doi:10.1073/pnas.2517530122 , abstract =

    work page doi:10.1073/pnas.2517530122 2025

  61. [61]

    Analyzing the

    Chitra, Uthsav and Musco, Christopher , month = jan, year =. Analyzing the. Proceedings of the 13th. doi:10.1145/3336191.3371825 , language =

    work page doi:10.1145/3336191.3371825

  62. [62]

    Quantifying. Trans. Soc. Comput. , author =. 2018 , pages =. doi:10.1145/3140565 , abstract =

    work page doi:10.1145/3140565 2018

  63. [63]

    ICWSM , author =

    A. ICWSM , author =. 2021 , pages =. doi:10.1609/icwsm.v7i1.14421 , abstract =

    work page doi:10.1609/icwsm.v7i1.14421 2021

  64. [64]

    ICWSM , author =

    The. ICWSM , author =. 2022 , pages =. doi:10.1609/icwsm.v16i1.19275 , abstract =

    work page doi:10.1609/icwsm.v16i1.19275 2022

  65. [65]

    Data Min Knowl Disc , author =

    Measuring and moderating opinion polarization in social networks , volume =. Data Min Knowl Disc , author =. 2017 , pages =. doi:10.1007/s10618-017-0527-9 , language =

    work page doi:10.1007/s10618-017-0527-9 2017

  66. [66]

    Minimizing

    Cinus, Federico and Miyauchi, Atsushi and Kuroki, Yuko and Bonchi, Francesco , month = jan, year =. Minimizing. doi:10.48550/arXiv.2501.16076 , abstract =

    work page doi:10.48550/arxiv.2501.16076

  67. [67]

    Journal of mathematical sociology , volume=

    Social influence and opinions , author=. Journal of mathematical sociology , volume=. 1990 , publisher=

    work page 1990

  68. [68]

    Advances in Complex Systems , volume=

    Mixing beliefs among interacting agents , author=. Advances in Complex Systems , volume=. 2000 , publisher=

    work page 2000

  69. [69]

    Dynamic Games and Applications , volume=

    Opinion dynamics and learning in social networks , author=. Dynamic Games and Applications , volume=. 2011 , publisher=

    work page 2011

  70. [70]

    The Journal of Mathematical Sociology , volume=

    Disambiguation of social polarization concepts and measures , author=. The Journal of Mathematical Sociology , volume=. 2016 , publisher=

    work page 2016

  71. [71]

    arXiv preprint arxiv:2602.21236 , year =

    Migliarini, Matteo and Ercevik, Berat and Olowe, Oluwagbemike and Fatima, Saira and Zhao, Sarah and Le, Minh Anh and Sharma, Vasu and Panda, Ashwinee , title =. arXiv preprint arxiv:2602.21236 , year =

    work page arXiv

  72. [72]

    arXiv preprint arXiv:2602.11286 , year =

    Mei, Katelyn Xiaoying and Wolfe, Robert and Weber, Nicholas and Saveski, Martin , title =. arXiv preprint arXiv:2602.11286 , year =

    work page arXiv

  73. [73]

    , title =

    Conger, Kate and Freedman, Dylan and Thompson, Stuart A. , title =. The New York Times , year =

    work page