pith. sign in

arxiv: 2605.16245 · v1 · pith:4OH7OC24new · submitted 2026-05-15 · 💻 cs.CY · cs.AI· cs.CL· cs.LG· cs.SI

AI-Mediated Communication Can Steer Collective Opinion

Stratis Tsirtsis , Kai Rawal , Chris Russell , Brent Mittelstadt , Sandra Wachter This is my paper

Pith reviewed 2026-05-19 21:34 UTC · model grok-4.3

classification 💻 cs.CY cs.AIcs.CLcs.LGcs.SI
keywords AI-mediated communicationopinion dynamicsLLM biascollective opinionsocial networksmessage editingplatform auditingcontent moderation
0
0 comments X

The pith

AI editing of messages can amplify biases through social networks and shift collective opinions.

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

The paper demonstrates that popular large language models introduce consistent directional biases when editing human-written texts on contested topics such as gun control or atheism. It builds a mathematical model placing an AI mediator between users on a network, where the AI transforms expressed opinions before they reach others. Analysis of the model's equilibrium and simulations on real network data show these biases can grow through repeated interactions and move overall group opinion in the biased direction. The authors further audit an existing platform feature and trace a pro-life tilt in its outputs to specific design choices. A reader would care because this mechanism operates on platforms people already use daily for discussion.

Core claim

Generative AI now polishes posts and explains content on platforms, and when instructed to edit human texts on contested topics the models from several families consistently nudge the output toward one side, such as favoring gun control or opposing atheism. Placing such an AI between users in an opinion-dynamics model on a network allows analytic characterization of the equilibrium; both the math and simulations on real social-network data establish that the introduced biases are amplified across connections and shift the collective opinion distribution in the direction of the bias. An audit of X's 'Explain this post' feature confirms pro-life bias traceable to concrete design decisions.

What carries the argument

A mathematical model of opinion dynamics in which an AI mediator sits between users on a social network and transforms the opinions they express and perceive before transmission.

If this is right

  • Biases in AI message editing can produce measurable long-term shifts in collective opinion across connected users.
  • Platform design choices that embed AI mediation can unintentionally steer public discourse on divisive issues.
  • Audits of specific features can identify and trace bias sources back to training or instruction decisions.
  • Regulatory efforts on AI in communication platforms may need to address mediation effects in addition to direct generation.

Where Pith is reading between the lines

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

  • If the amplification holds in live settings, small consistent nudges in everyday AI tools could accumulate into noticeable changes in public sentiment over months or years.
  • The same model framework could be used to test whether different network structures, such as echo chambers versus diverse graphs, change how fast or how far the bias spreads.
  • Platforms might counter the effect by randomizing editing instructions or adding explicit neutrality constraints, though this remains untested here.

Load-bearing premise

Biases observed when AI edits isolated texts on selected topics will keep appearing and steer opinions the same way in ongoing, real-time human conversations on actual platforms.

What would settle it

Measure whether average opinions on a contested topic shift measurably toward the AI's editing direction after a platform rolls out AI polishing or explanation features, compared with a matched control group or time period without those features.

Figures

Figures reproduced from arXiv: 2605.16245 by Brent Mittelstadt, Chris Russell, Kai Rawal, Sandra Wachter, Stratis Tsirtsis.

Figure 1
Figure 1. Figure 1: Analysis of bias introduced by LLMs when improving human-written posts. Panel (a) shows the original opinions of 400 posts on feminism from the SemEval dataset against those of their LLM-improved counterparts generated by gemma-3-12b-it, where the green and pink marker cor￾respond to average values for posts labeled “in favor” and “against”, respectively. Panel (b) shows the posterior means and 95% credibl… view at source ↗
Figure 2
Figure 2. Figure 2: Opinion dynamics when gemma-3-12b-it is used to edit posts. Panel (a) shows the average opinion on abortion over time, for different fractions ϕ of AI adopters. Panel (b) shows the long-run average opinion under AI transformations from different topics and datasets (see Appendix B.4 for abbreviations) against the AI’s bias, as measured by the posterior mean of the intercept in Eq. 1. “X” indicates no AI tr… view at source ↗
Figure 3
Figure 3. Figure 3: Bias introduced by Grok when contextualizing X posts on abortion. Panel (a) shows the stance distribution of contextual claims generated by Grok based on the implementation of X’s “Explain this post” feature, broken down by whether the post is pro-choice or pro-life. Panel (b) shows the four guidelines included by X in the model’s instructions and the change in the stance distribution of Grok’s contextual … view at source ↗
Figure 4
Figure 4. Figure 4: Bias introduced by LLMs when improving human-written posts. The panels show the posterior means and 95% credible intervals of the intercepts capturing the average bias β (see Section 2) by different LLMs across topics from the SemEval dataset, using prompts for the improvement task (see B.1.2). −0.05 0.00 0.05 0.10 0.15 Bias towards “in favor” Death penalty School uniforms Cloning Minimum wage Marijuana le… view at source ↗
Figure 5
Figure 5. Figure 5: Bias introduced by LLMs when drafting social-media posts. The panels show the posterior means and 95% credible intervals of the intercepts capturing the average bias β (see Section 2) by different LLMs across topics from the UKP dataset, using prompts for the drafting task (see B.1.1). 0.0 0.2 0.4 0.6 0.8 1.0 Average directly expressed opinion −0.05 0.00 0.05 0.10 Average bias towards “in favor” Gun contro… view at source ↗
Figure 6
Figure 6. Figure 6: Average LLM-induced bias vs. average directly expressed opinion (UKP). The figure shows the mean of the bias β against the average directly expressed opinion of each model on each topic. Each point represents one model-topic pair with different markers used for Llama-3.1-8B-Instruct ( ), Ministral-3-8B-Instruct-2512 ( ), gemma-3-12b-it ( ), and Qwen3-8B ( ). 27 [PITH_FULL_IMAGE:figures/full_fig_p027_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Opinion transformations resulting from gemma-3-12b-it across topics. In each panel, each gray point shows the original opinion x expressed in a human-written text from the respective dataset against the opinion y expressed in its LLM-generated counterpart, averaged across prompt variants and random seeds used for the generation. The respective pink line corresponds to the AI transformation f, fitted on the… view at source ↗
Figure 8
Figure 8. Figure 8: AI Bias vs long-run average opinion across AI transformations. The panels show the long-run average opinion under AI transformations based on different topics and datasets against the AI’s bias, as measured by the posterior mean of the intercept in Eq. 1. “X” indicates no AI transformation. All simulations were conducted with the gemma-3-12b-it model with κ = 0.4, λ = 0.3, and ϕ = 0.6. 0.1 0.2 0.3 0.4 0.5 … view at source ↗
Figure 9
Figure 9. Figure 9: Shift in long-run average opinion under different model parameters using the Twitter network. Heatmaps show the change in average long-run opinion between simulations with AI mediation (ϕ = 0.6) and without mediation (ϕ = 0), across values of κ and λ, for each topic in the SemEval dataset using gemma-3-12b-it. 29 [PITH_FULL_IMAGE:figures/full_fig_p029_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: Shift in long-run average opinion under different model parameters using the Google Plus network. Heatmaps show the change in average long-run opinion between simulations with AI me￾diation (ϕ = 0.6) and without mediation (ϕ = 0), across values of κ and λ, for each topic in the SemEval dataset using gemma-3-12b-it. 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 Average stubbornness 10% 20% 30% 40% 50% 60% 70% 80% 90… view at source ↗
Figure 11
Figure 11. Figure 11: Shift in long-run average opinion under different model parameters using the Face￾book network. Heatmaps show the change in average long-run opinion between simulations with AI mediation (ϕ = 0.6) and without mediation (ϕ = 0), across values of κ and λ, for each topic in the SemEval dataset using gemma-3-12b-it. 30 [PITH_FULL_IMAGE:figures/full_fig_p030_11.png] view at source ↗
Figure 12
Figure 12. Figure 12: Analysis of convergence of individual opinions under AI-mediated opinion dynamics across topics and networks. Each panel shows the maximum change individuals’ opinions per time step against the average stubbornness λ. 31 [PITH_FULL_IMAGE:figures/full_fig_p031_12.png] view at source ↗
Figure 13
Figure 13. Figure 13: Analysis of convergence of the average opinion under AI-mediated opinion dynamics across topics and networks. Each panel shows the change in average opinion per time step against the average stubbornness λ. 32 [PITH_FULL_IMAGE:figures/full_fig_p032_13.png] view at source ↗
Figure 14
Figure 14. Figure 14: Biases introduced by gemma-3-12b-it under different ideological viewpoint prefixes in its system prompt. Markers represent the posterior means of the intercept of the Bayesian linear mixed effects model given by Eq. 1, capturing the bias β that the LLM introduces on each of the 6 SemEval topics, with error bars representing 95% credible intervals. Within each topic, each marker corresponds to a system pro… view at source ↗
read the original abstract

Generative artificial intelligence (AI) is increasingly integrated into the online platforms where humans exchange opinions; large language models (LLMs) now polish users' posts on LinkedIn and provide context for content shared on X. While prior work has shown that AI can express biased opinions and shape individuals' opinions during human-AI interactions, less attention has been paid to its influence on collective opinion formation when mediating human-to-human communication. We address this gap via a combination of empirical and theoretical analyses. We show empirically that LLMs from multiple popular families introduce directional biases when instructed to edit human-written texts on contested topics, for example, nudging texts in favor of gun control and against atheism. Building on this observation, we introduce a mathematical model of opinion dynamics in which an AI system sits between users on a social network, transforming the opinions they express and perceive. By analytically characterizing the equilibrium of this model and performing simulations on real social network data, we show that biases introduced by AI in human-to-human communication can be amplified through the network and shift collective opinion in their direction. In light of these findings, we investigate whether such biases are controllable by online platforms. We audit the "Explain this post" feature on X and find evidence of pro-life bias in Grok's outputs on abortion-related content, which we trace back to specific design choices. We conclude with a discussion of the broader implications of our findings in relation to ongoing legislative efforts in the European Union.

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

2 major / 2 minor

Summary. The paper claims that LLMs from multiple families introduce directional biases when editing human texts on contested topics (e.g., favoring gun control or opposing atheism). It introduces a mathematical model of opinion dynamics on a social network in which an AI mediator applies a transformation to expressed and perceived opinions, analytically characterizes the equilibrium, and uses simulations on real network data to argue that these biases amplify through the network and shift collective opinion. The work also audits X's 'Explain this post' feature, finding pro-life bias in Grok outputs traceable to design choices, and discusses regulatory implications.

Significance. If the central modeling and empirical-to-theoretical bridge hold, the result identifies a plausible mechanism for AI to steer collective opinion via mediation of human-to-human exchanges, with direct relevance to platform design and EU regulatory efforts. The combination of multi-LLM empirical tests, closed-form equilibrium analysis, and network simulations on real data provides a structured framework that could be extended to other mediation scenarios.

major comments (2)
  1. [§3] §3 (Mathematical Model and Equilibrium Derivation): The model treats the AI transformation as a fixed directional bias applied uniformly to opinions in ongoing exchanges, but the empirical editing experiments are static, single-shot tasks with explicit instructions; the paper does not demonstrate or test whether this bias persists under variable user prompts, overrides, or real-time context, which is load-bearing for the amplification claim in the simulations.
  2. [§4] §4 (Simulations on Real Network Data): The reported opinion shifts rely on the assumption of consistent directional bias magnitude across interactions; without reported sensitivity analysis on bias variability (observed across LLM families and topics in the empirical section) or on network topology parameters, it is unclear whether the equilibrium shift is robust or an artifact of the chosen bias value.
minor comments (2)
  1. [Empirical section] Empirical section: sample sizes, exact statistical tests, data exclusion criteria, and prompt templates for the editing tasks are not fully specified, making it difficult to assess reproducibility of the directional bias findings.
  2. [Audit section] The audit of the 'Explain this post' feature would benefit from a clearer description of the sampling procedure for abortion-related posts and the exact criteria used to classify outputs as pro-life biased.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive comments, which help clarify the connection between our empirical findings and the modeling framework. We address each major comment below and describe the revisions we will make.

read point-by-point responses

  1. Referee: [§3] §3 (Mathematical Model and Equilibrium Derivation): The model treats the AI transformation as a fixed directional bias applied uniformly to opinions in ongoing exchanges, but the empirical editing experiments are static, single-shot tasks with explicit instructions; the paper does not demonstrate or test whether this bias persists under variable user prompts, overrides, or real-time context, which is load-bearing for the amplification claim in the simulations.

    Authors: We acknowledge that the empirical experiments are single-shot editing tasks under fixed instructions, while the model assumes repeated application of a consistent transformation. The empirical results establish the presence and direction of biases for standard editing prompts on contested topics, which directly inform the model's transformation parameters. We agree that explicit tests of persistence under prompt variation would strengthen the link. In revision we will add a dedicated limitations subsection in §3 discussing this assumption and include supplementary experiments testing bias consistency across varied prompts and contexts for representative LLMs and topics. revision: partial

  2. Referee: [§4] §4 (Simulations on Real Network Data): The reported opinion shifts rely on the assumption of consistent directional bias magnitude across interactions; without reported sensitivity analysis on bias variability (observed across LLM families and topics in the empirical section) or on network topology parameters, it is unclear whether the equilibrium shift is robust or an artifact of the chosen bias value.

    Authors: The simulations employ bias magnitudes calibrated to empirical averages to illustrate network-level amplification. We recognize that variability across LLMs and topics, as well as network parameters, warrants explicit robustness checks. In the revised manuscript we will add sensitivity analyses that sweep bias magnitudes over the range observed in the empirical section and vary key network topology parameters (e.g., density, degree distribution) to confirm that the directional equilibrium shift remains qualitatively stable. revision: yes

Circularity Check

0 steps flagged

No significant circularity: model derived and solved independently

full rationale

The paper separates empirical bias measurements (LLMs editing texts on contested topics) from the mathematical model of AI-mediated opinion dynamics. The model is introduced based on the observation but then formulated and its equilibrium characterized analytically without any reduction of predictions to fitted parameters or self-definitional loops. Simulations apply the independent model to network data. No self-citations, uniqueness theorems, or ansatzes smuggled via prior work are described as load-bearing for the central claims. The derivation chain remains self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 0 invented entities

The central claim rests on standard assumptions from opinion dynamics literature plus the empirical observation that LLM edits produce consistent directional shifts; no new entities are postulated and free parameters appear limited to bias magnitude.

free parameters (1)
  • AI bias magnitude
    The model uses a parameter controlling how strongly the AI transforms messages; its value is informed by the empirical editing experiments rather than derived from first principles.
axioms (1)
  • domain assumption Users update opinions based on the AI-transformed messages they receive according to a standard averaging or bounded-confidence rule
    This is the core update mechanism in the opinion dynamics model introduced in the theoretical section.

pith-pipeline@v0.9.0 · 5810 in / 1346 out tokens · 99649 ms · 2026-05-19T21:34:55.708665+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

100 extracted references · 100 canonical work pages · 1 internal anchor

  1. [1]

    Accessed: 2026-04-22

    https://www.linkedin.com/help/linkedin/answer/a1517763. Accessed: 2026-04-22

    work page 2026

  2. [2]

    Accessed: 2026-04-22

    https://blog.youtube/inside-youtube/2024-in-youtube-ai/. Accessed: 2026-04-22

    work page 2024

  3. [3]

    Accessed: 2026-04-22

    https://x.ai/news/grok-1212. Accessed: 2026-04-22

    work page 2026

  4. [4]

    Fine-tuning language models to find agreement among humans with diverse preferences.Advances in neural information processing systems, 35:38176–38189, 2022

    Michiel Bakker, Martin Chadwick, Hannah Sheahan, Michael Tessler, Lucy Campbell-Gillingham, Jan Balaguer, Nat McAleese, Amelia Glaese, John Aslanides, Matt Botvinick, et al. Fine-tuning language models to find agreement among humans with diverse preferences.Advances in neural information processing systems, 35:38176–38189, 2022

    work page 2022

  5. [5]

    Ai can help humans find common ground in democratic deliberation.Science, 386(6719):eadq2852, 2024

    Michael Henry Tessler, Michiel A Bakker, Daniel Jarrett, Hannah Sheahan, Martin J Chadwick, Raphael Koster, Georgina Evans, Lucy Campbell-Gillingham, Tantum Collins, David C Parkes, et al. Ai can help humans find common ground in democratic deliberation.Science, 386(6719):eadq2852, 2024

    work page 2024

  6. [6]

    Whose opinions do language models reflect? InInternational conference on machine learning, pages 29971–30004

    Shibani Santurkar, Esin Durmus, Faisal Ladhak, Cinoo Lee, Percy Liang, and Tatsunori Hashimoto. Whose opinions do language models reflect? InInternational conference on machine learning, pages 29971–30004. PMLR, 2023

    work page 2023

  7. [7]

    Bias in opinion summarisation from pre- training to adaptation: A case study in political bias

    Nannan Huang, Haytham Fayek, and Xiuzhen Jenny Zhang. Bias in opinion summarisation from pre- training to adaptation: A case study in political bias. InProceedings of the 18th Conference of the European Chapter of the Association for Computational Linguistics (Volume 1: Long Papers), pages 1041–1055, 2024

    work page 2024

  8. [8]

    Evaluating the persuasive influence of political microtargeting with large language models.Proceedings of the National Academy of Sciences, 121(24):e2403116121, 2024

    Kobi Hackenburg and Helen Margetts. Evaluating the persuasive influence of political microtargeting with large language models.Proceedings of the National Academy of Sciences, 121(24):e2403116121, 2024

    work page 2024

  9. [9]

    The levers of political persuasion with conversational artificial intelligence.Science, 390(6777):eaea3884, 2025

    Kobi Hackenburg, Ben M Tappin, Luke Hewitt, Ed Saunders, Sid Black, Hause Lin, Catherine Fist, Helen Margetts, David G Rand, and Christopher Summerfield. The levers of political persuasion with conversational artificial intelligence.Science, 390(6777):eaea3884, 2025

    work page 2025

  10. [10]

    On the conversational persuasiveness of gpt-4.Nature Human Behaviour, 9(8):1645–1653, 2025

    Francesco Salvi, Manoel Horta Ribeiro, Riccardo Gallotti, and Robert West. On the conversational persuasiveness of gpt-4.Nature Human Behaviour, 9(8):1645–1653, 2025

    work page 2025

  11. [11]

    Co-writing with opinionated language models affects users’ views

    Maurice Jakesch, Advait Bhat, Daniel Buschek, Lior Zalmanson, and Mor Naaman. Co-writing with opinionated language models affects users’ views. InProceedings of the 2023 CHI conference on human factors in computing systems, pages 1–15, 2023

    work page 2023

  12. [12]

    Reaching a consensus.Journal of the American Statistical association, 69(345): 118–121, 1974

    Morris H DeGroot. Reaching a consensus.Journal of the American Statistical association, 69(345): 118–121, 1974

    work page 1974

  13. [13]

    Social influence and opinions.Journal of mathematical sociology, 15(3-4):193–206, 1990

    Noah E Friedkin and Eugene C Johnsen. Social influence and opinions.Journal of mathematical sociology, 15(3-4):193–206, 1990

    work page 1990

  14. [14]

    Opinion Dynamics and Bounded Confidence: Models, Analysis and Simulation.Journal of Artificial Societies and Social Simulation, 5(3), 2002

    Hegselmann Rainer and Ulrich Krause. Opinion Dynamics and Bounded Confidence: Models, Analysis and Simulation.Journal of Artificial Societies and Social Simulation, 5(3), 2002

    work page 2002

  15. [15]

    Opinion dynamics: A comprehensive overview.arXiv preprint arXiv:2511.00401, 2025

    Mohammad Shirzadi, Emilio Cruciani, and Ahad N Zehmakan. Opinion dynamics: A comprehensive overview.arXiv preprint arXiv:2511.00401, 2025

    work page arXiv 2025

  16. [16]

    Opinion maximization in social networks

    Aristides Gionis, Evimaria Terzi, and Panayiotis Tsaparas. Opinion maximization in social networks. InProceedings of the 2013 SIAM international conference on data mining, pages 387–395. SIAM, 2013

    work page 2013

  17. [17]

    How bad is forming your own opinion?Games and Economic Behavior, 92:248–265, 2015

    David Bindel, Jon Kleinberg, and Sigal Oren. How bad is forming your own opinion?Games and Economic Behavior, 92:248–265, 2015. 13

    work page 2015

  18. [18]

    Minimizing polarization and disagreement in social networks

    Cameron Musco, Christopher Musco, and Charalampos E Tsourakakis. Minimizing polarization and disagreement in social networks. InProceedings of the 2018 world wide web conference, pages 369–378, 2018

    work page 2018

  19. [19]

    Adversarial perturbations of opinion dynamics in networks

    Jason Gaitonde, Jon Kleinberg, and Eva Tardos. Adversarial perturbations of opinion dynamics in networks. InProceedings of the 21st ACM Conference on Economics and Computation, pages 471–472, 2020

    work page 2020

  20. [20]

    Adversaries with limited information in the friedkin- johnsen model

    Sijing Tu, Stefan Neumann, and Aristides Gionis. Adversaries with limited information in the friedkin- johnsen model. InProceedings of the 29th ACM SIGKDD Conference on Knowledge Discovery and Data Mining, pages 2201–2210, 2023

    work page 2023

  21. [21]

    A survey on algorithmic interventions in opinion dynamics.arXiv preprint arXiv:2603.10756, 2026

    Atsushi Miyauchi, Yuko Kuroki, Federico Cinus, Stefan Neumann, and Francesco Bonchi. A survey on algorithmic interventions in opinion dynamics.arXiv preprint arXiv:2603.10756, 2026

    work page arXiv 2026

  22. [22]

    Ai and the problem of knowledge collapse.AI & SOCIETY, 40(5):3249–3269, 2025

    Andrew J Peterson. Ai and the problem of knowledge collapse.AI & SOCIETY, 40(5):3249–3269, 2025

    work page 2025

  23. [23]

    Do large language models have a legal duty to tell the truth?Royal Society Open Science, 11(8):240197, 2024

    Sandra Wachter, Brent Mittelstadt, and Chris Russell. Do large language models have a legal duty to tell the truth?Royal Society Open Science, 11(8):240197, 2024

    work page 2024

  24. [24]

    The impact of advanced ai systems on democracy.Nature Human Behaviour, 9(12):2420–2430, 2025

    Christopher Summerfield, Lisa P Argyle, Michiel Bakker, Teddy Collins, Esin Durmus, Tyna Eloundou, Iason Gabriel, Deep Ganguli, Kobi Hackenburg, Gillian K Hadfield, et al. The impact of advanced ai systems on democracy.Nature Human Behaviour, 9(12):2420–2430, 2025

    work page 2025

  25. [25]

    How ai threatens democracy.Journal of Democracy, 34(4):122–131, 2023

    Sarah Kreps and Doug Kriner. How ai threatens democracy.Journal of Democracy, 34(4):122–131, 2023

    work page 2023

  26. [26]

    Stance detection: A survey.ACM Computing Surveys (CSUR), 53(1):1–37, 2020

    Dilek K¨ u¸ c¨ uk and Fazli Can. Stance detection: A survey.ACM Computing Surveys (CSUR), 53(1):1–37, 2020

    work page 2020

  27. [27]

    Cross-topic argu- ment mining from heterogeneous sources

    Christian Stab, Tristan Miller, Benjamin Schiller, Pranav Rai, and Iryna Gurevych. Cross-topic argu- ment mining from heterogeneous sources. InProceedings of the 2018 conference on empirical methods in natural language processing, pages 3664–3674, 2018

    work page 2018

  28. [28]

    Semeval- 2016 task 6: Detecting stance in tweets

    Saif Mohammad, Svetlana Kiritchenko, Parinaz Sobhani, Xiaodan Zhu, and Colin Cherry. Semeval- 2016 task 6: Detecting stance in tweets. InProceedings of the 10th international workshop on semantic evaluation (SemEval-2016), pages 31–41, 2016

    work page 2016

  29. [29]

    Bayesian linear mixed models using stan: A tutorial for psy- chologists, linguists, and cognitive scientists.arXiv preprint arXiv:1506.06201, 2015

    Tanner Sorensen and Shravan Vasishth. Bayesian linear mixed models using stan: A tutorial for psy- chologists, linguists, and cognitive scientists.arXiv preprint arXiv:1506.06201, 2015

    work page arXiv 2015

  30. [30]

    Advanced Bayesian Multilevel Modeling with the R Package brms

    Paul-Christian B¨ urkner. Advanced bayesian multilevel modeling with the r package brms.arXiv preprint arXiv:1705.11123, 2017

    work page internal anchor Pith review Pith/arXiv arXiv 2017

  31. [31]

    Symbolic description of factorial models for analysis of variance.Journal of the Royal Statistical Society Series C: Applied Statistics, 22(3):392–399, 1973

    GN Wilkinson and CE Rogers. Symbolic description of factorial models for analysis of variance.Journal of the Royal Statistical Society Series C: Applied Statistics, 22(3):392–399, 1973

    work page 1973

  32. [32]

    Linear representations of political perspective emerge in large language models

    Junsol Kim, James Evans, and Aaron Schein. Linear representations of political perspective emerge in large language models. InThe Thirteenth International Conference on Learning Representations, 2025

    work page 2025

  33. [33]

    Cambridge University Press, 2011

    Noah E Friedkin and Eugene C Johnsen.Social influence network theory: A sociological examination of small group dynamics, volume 33. Cambridge University Press, 2011

    work page 2011

  34. [34]

    Cultural reception and production: The social construction of meaning in book clubs.American Sociological Review, 77(1):45–68, 2012

    C Clayton Childress and Noah E Friedkin. Cultural reception and production: The social construction of meaning in book clubs.American Sociological Review, 77(1):45–68, 2012. 14

    work page 2012

  35. [35]

    Learning a linear influence model from transient opinion dynamics

    Abir De, Sourangshu Bhattacharya, Parantapa Bhattacharya, Niloy Ganguly, and Soumen Chakrabarti. Learning a linear influence model from transient opinion dynamics. InProceedings of the 23rd ACM international conference on conference on information and knowledge management, pages 401–410, 2014

    work page 2014

  36. [36]

    Network science on belief system dynamics under logic constraints.Science, 354(6310):321–326, 2016

    Noah E Friedkin, Anton V Proskurnikov, Roberto Tempo, and Sergey E Parsegov. Network science on belief system dynamics under logic constraints.Science, 354(6310):321–326, 2016

    work page 2016

  37. [37]

    A theory of the evolution of social power: Natural trajectories of interpersonal influence systems along issue sequences.Sociological Science, 3:444–472, 2016

    Noah E Friedkin, Peng Jia, and Francesco Bullo. A theory of the evolution of social power: Natural trajectories of interpersonal influence systems along issue sequences.Sociological Science, 3:444–472, 2016

    work page 2016

  38. [38]

    How truth wins in opinion dynamics along issue sequences

    Noah E Friedkin and Francesco Bullo. How truth wins in opinion dynamics along issue sequences. Proceedings of the National Academy of Sciences, 114(43):11380–11385, 2017

    work page 2017

  39. [39]

    Achieving consensus in multilateral international negotiations: The case study of the 2015 paris agree- ment on climate change.Science Advances, 7(51):eabg8068, 2021

    Carmela Bernardo, Lingfei Wang, Francesco Vasca, Yiguang Hong, Guodong Shi, and Claudio Altafini. Achieving consensus in multilateral international negotiations: The case study of the 2015 paris agree- ment on climate change.Science Advances, 7(51):eabg8068, 2021

    work page 2015

  40. [40]

    Opinion dynamics in social networks with stubborn agents: Equilibrium and convergence rate.Automatica, 50(12):3209–3215, 2014

    Javad Ghaderi and Rayadurgam Srikant. Opinion dynamics in social networks with stubborn agents: Equilibrium and convergence rate.Automatica, 50(12):3209–3215, 2014

    work page 2014

  41. [41]

    Opinion dynamics with local inter- actions

    Dimitris Fotakis, Dimitris Palyvos-Giannas, and Stratis Skoulakis. Opinion dynamics with local inter- actions. InIJCAI, pages 279–285, 2016

    work page 2016

  42. [42]

    Opinion dynamics with varying susceptibility to persuasion

    Rediet Abebe, Jon Kleinberg, David Parkes, and Charalampos E Tsourakakis. Opinion dynamics with varying susceptibility to persuasion. InProceedings of the 24th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining, pages 1089–1098, 2018

    work page 2018

  43. [43]

    Quantifying and minimizing risk of conflict in social networks

    Xi Chen, Jefrey Lijffijt, and Tijl De Bie. Quantifying and minimizing risk of conflict in social networks. InProceedings of the 24th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining, pages 1197–1205, 2018

    work page 2018

  44. [44]

    Analyzing the impact of filter bubbles on social network po- larization

    Uthsav Chitra and Christopher Musco. Analyzing the impact of filter bubbles on social network po- larization. InProceedings of the 13th international conference on web search and data mining, pages 115–123, 2020

    work page 2020

  45. [45]

    Minimizing polarization and disagreement in social networks via link recommendation.Advances in Neural Information Processing Systems, 34:2072–2084, 2021

    Liwang Zhu, Qi Bao, and Zhongzhi Zhang. Minimizing polarization and disagreement in social networks via link recommendation.Advances in Neural Information Processing Systems, 34:2072–2084, 2021

    work page 2072

  46. [46]

    On the relationship between relevance and conflict in online social link recommendations.Advances in Neural Information Processing Systems, 36:36708–36725, 2023

    Yanbang Wang and Jon Kleinberg. On the relationship between relevance and conflict in online social link recommendations.Advances in Neural Information Processing Systems, 36:36708–36725, 2023

    work page 2023

  47. [47]

    A tutorial on modeling and analysis of dynamic social networks

    Anton V Proskurnikov and Roberto Tempo. A tutorial on modeling and analysis of dynamic social networks. part i.Annual Reviews in Control, 43:65–79, 2017

    work page 2017

  48. [48]

    Bullo.Contraction Theory for Dynamical Systems

    F. Bullo.Contraction Theory for Dynamical Systems. Kindle Direct Publishing, 1.3 edition, 2026. ISBN 979-8836646806

    work page 2026

  49. [49]

    Learning to discover social circles in ego networks.Advances in neural information processing systems, 25, 2012

    Jure Leskovec and Julian Mcauley. Learning to discover social circles in ego networks.Advances in neural information processing systems, 25, 2012

    work page 2012

  50. [50]

    On estimating regression.Theory of Probability & Its Applications, 9(1):141–142, 1964

    Elizbar A Nadaraya. On estimating regression.Theory of Probability & Its Applications, 9(1):141–142, 1964

    work page 1964

  51. [51]

    Smooth regression analysis.Sankhy¯ a: The Indian Journal of Statistics, Series A, pages 359–372, 1964

    Geoffrey S Watson. Smooth regression analysis.Sankhy¯ a: The Indian Journal of Statistics, Series A, pages 359–372, 1964. 15

    work page 1964

  52. [52]

    University of Chicago Press, 2023

    Alexander Coppock.Persuasion in parallel: How information changes minds about politics. University of Chicago Press, 2023

    work page 2023

  53. [53]

    Judging llm-as-a-judge with mt-bench and chatbot arena

    Lianmin Zheng, Wei-Lin Chiang, Ying Sheng, Siyuan Zhuang, Zhanghao Wu, Yonghao Zhuang, Zi Lin, Zhuohan Li, Dacheng Li, Eric Xing, et al. Judging llm-as-a-judge with mt-bench and chatbot arena. Advances in neural information processing systems, 36:46595–46623, 2023

    work page 2023

  54. [54]

    Language models are few-shot learn- ers.Advances in neural information processing systems, 33:1877–1901, 2020

    Tom Brown, Benjamin Mann, Nick Ryder, Melanie Subbiah, Jared D Kaplan, Prafulla Dhariwal, Arvind Neelakantan, Pranav Shyam, Girish Sastry, Amanda Askell, et al. Language models are few-shot learn- ers.Advances in neural information processing systems, 33:1877–1901, 2020

    work page 1901

  55. [55]

    Limitations and loopholes in the eu ai act and ai liability directives: what this means for the european union, the united states, and beyond.Yale JL & Tech., 26:702, 2023

    Sandra Wachter. Limitations and loopholes in the eu ai act and ai liability directives: what this means for the european union, the united states, and beyond.Yale JL & Tech., 26:702, 2023

    work page 2023

  56. [56]

    europa.eu/legal-content/EN/TXT/HTML/?uri=CELEX%3A52020DC0790,

    https://eur-lex. europa.eu/legal-content/EN/TXT/HTML/?uri=CELEX%3A52020DC0790, . Ac- cessed: 2026-05-13

    work page 2026

  57. [57]

    Accessed: 2026-05-13

    https://digital-strategy.ec.europa.eu/en/news/press-statement-european-board-digital-services- following-its-16th-meeting, . Accessed: 2026-05-13

    work page 2026

  58. [58]

    https://democracy-reporting.org/en/office/EU/publications/big-tech-is-backing- out-of-commitments-countering-disinformation-whats-next-for-the-eus-code-of- practice#WhatLiesAheadfortheCodeofPractice(Conduct)onDisinformation? Accessed: 2026-05-13

    work page 2026

  59. [59]

    Accessed: 2026-05-13

    https://www.euractiv.com/news/eu-slaps-elon-musks-x-e120-million-for-first-confirmed-dsa-breaches/, . Accessed: 2026-05-13

    work page 2026

  60. [60]

    Accessed: 2026-05-13

    https://www.techpolicy.press/the-us-governments-use-of-elon-musks-grok-ai-undermines-its-own- rules/, . Accessed: 2026-05-13

    work page 2026

  61. [61]

    To protect science, we must use llms as zero-shot translators.Nature Human Behaviour, 7(11):1830–1832, 2023

    Brent Mittelstadt, Sandra Wachter, and Chris Russell. To protect science, we must use llms as zero-shot translators.Nature Human Behaviour, 7(11):1830–1832, 2023

    work page 2023

  62. [62]

    Accessed: 2026-05-13

    https://www.theguardian.com/technology/ng-interactive/2026/mar/10/ai-impact-professors-students- learning, . Accessed: 2026-05-13

    work page 2026

  63. [63]

    Accessed: 2026-05-13

    https://www.insidehighered.com/news/tech-innovation/teaching-learning/2026/03/16/writing-faculty- push-right-refuse-ai, . Accessed: 2026-05-13

    work page 2026

  64. [64]

    Accessed: 2026-05-13

    https://blog.routledge.com/humanities-and-media-arts/ai-in-the-media-industry-a-miracle-or-a- minefield/ , . Accessed: 2026-05-13

    work page 2026

  65. [65]

    Accessed: 2026-05-13

    https://www.ibm.com/think/insights/ai-in-journalism, . Accessed: 2026-05-13

    work page 2026

  66. [66]

    Accessed: 2026-05-13

    https://algorithmwatch.org/en/could-ai-chatbots-influence-governments/, . Accessed: 2026-05-13

    work page 2026

  67. [67]

    Accessed: 2026-05- 13

    https://restofworld.org/2026/government-ai-hallucinations-south-africa-deloitte/, . Accessed: 2026-05- 13

    work page 2026

  68. [68]

    Accessed: 2026-05-13

    https://www.theguardian.com/technology/2026/apr/22/ai-hallucinations-found-in-high-profile-wall- street-law-firm-filing, . Accessed: 2026-05-13

    work page 2026

  69. [69]

    Accessed: 2026-05-13

    https://www.wired.com/story/hospitals-ai-transcription-tools-hallucination/, . Accessed: 2026-05-13

    work page 2026

  70. [70]

    Generative artificial intelligence in primary care: an online survey of uk general practitioners.BMJ Health & Care Informatics, 31(1):e101102, 2024

    Charlotte R Blease, Cosima Locher, Jens Gaab, Maria H¨ agglund, and Kenneth D Mandl. Generative artificial intelligence in primary care: an online survey of uk general practitioners.BMJ Health & Care Informatics, 31(1):e101102, 2024. 16

    work page 2024

  71. [71]

    Ergodic theorems for weakly interacting infinite systems and the voter model.The annals of probability, pages 643–663, 1975

    Richard A Holley and Thomas M Liggett. Ergodic theorems for weakly interacting infinite systems and the voter model.The annals of probability, pages 643–663, 1975

    work page 1975

  72. [72]

    Meet, discuss, and segregate!Complexity, 7(3):55–63, 2002

    G´ erard Weisbuch, Guillaume Deffuant, Fr´ ed´ eric Amblard, and Jean-Pierre Nadal. Meet, discuss, and segregate!Complexity, 7(3):55–63, 2002

    work page 2002

  73. [73]

    Large language models reflect the ideology of their creators.npj Artificial Intelligence, 2(1):7, 2026

    Maarten Buyl, Alexander Rogiers, Sander Noels, Guillaume Bied, Iris Dominguez-Catena, Edith Heiter, Iman Johary, Alexandru-Cristian Mara, Rapha¨ el Romero, Jefrey Lijffijt, et al. Large language models reflect the ideology of their creators.npj Artificial Intelligence, 2(1):7, 2026

    work page 2026

  74. [74]

    Aligning large language models with diverse political viewpoints

    Dominik Stammbach, Philine Widmer, Eunjung Cho, Caglar Gulcehre, and Elliott Ash. Aligning large language models with diverse political viewpoints. InProceedings of the 2024 Conference on Empirical Methods in Natural Language Processing, pages 7257–7267, 2024

    work page 2024

  75. [75]

    Hidden persuaders: LLMs’ political leaning and their influence on voters

    Yujin Potter, Shiyang Lai, Junsol Kim, James Evans, and Dawn Song. Hidden persuaders: LLMs’ political leaning and their influence on voters. InProceedings of the 2024 Conference on Empirical Methods in Natural Language Processing, pages 4244–4275, 2024

    work page 2024

  76. [76]

    Opinion dynamics and learning in social networks.Dynamic Games and Applications, 1(1):3–49, 2011

    Daron Acemoglu and Asuman Ozdaglar. Opinion dynamics and learning in social networks.Dynamic Games and Applications, 1(1):3–49, 2011

    work page 2011

  77. [77]

    Opinion dynamics: models, exten- sions and external effects

    Alina Sˆ ırbu, Vittorio Loreto, Vito DP Servedio, and Francesca Tria. Opinion dynamics: models, exten- sions and external effects. InParticipatory sensing, opinions and collective awareness, pages 363–401. Springer, 2016

    work page 2016

  78. [78]

    Simulating opinion dynamics with networks of llm-based agents

    Yun-Shiuan Chuang, Agam Goyal, Nikunj Harlalka, Siddharth Suresh, Robert Hawkins, Sijia Yang, Dhavan Shah, Junjie Hu, and Timothy Rogers. Simulating opinion dynamics with networks of llm-based agents. InFindings of the association for computational linguistics: NAACL 2024, pages 3326–3346, 2024

    work page 2024

  79. [79]

    Language-driven opinion dynamics in agent-based simulations with llms.arXiv preprint arXiv:2502.19098, 2025

    Erica Cau, Valentina Pansanella, Dino Pedreschi, and Giulio Rossetti. Language-driven opinion dynam- ics in agent-based simulations with llms.arXiv preprint arXiv:2502.19098, 2025

    work page arXiv 2025

  80. [80]

    Modeling the impact of large language models on opinion dynamics: A simulation-based study.Engineering Applications of Artificial Intelligence, 164:113353, 2026

    Chao Li, Xing Su, Haoying Han, Cong Xue, Chunmo Zheng, and Chao Fan. Modeling the impact of large language models on opinion dynamics: A simulation-based study.Engineering Applications of Artificial Intelligence, 164:113353, 2026

    work page 2026

Showing first 80 references.