arxiv: 2605.07069 · v3 · submitted 2026-05-08 · 💻 cs.MA · cs.CY

Recognition: 2 theorem links

· Lean Theorem

Social Theory Should Be a Structural Prior for Agentic AI: A Formal Framework for Multi-Agent Social Systems

Adrian Xuan Wei Lim, Iain J. Cruickshank, Kathleen M. Carley, Lynnette Hui Xian Ng

Authors on Pith no claims yet

Pith reviewed 2026-05-13 07:54 UTC · model grok-4.3

classification 💻 cs.MA cs.CY

keywords agentic AImulti-agent systemssocial theorystructural priorsMASS frameworkemergent dynamicsdynamical systems

0 comments

The pith

Agentic AI systems must incorporate social theory as a structural prior to model emergent behaviors in multi-agent environments.

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

The paper argues that when AI agents interact in groups, such as on social platforms or in robotic teams, their collective behaviors arise from interactions studied extensively by social scientists in human settings. Rather than learning these patterns solely from data, the authors propose building social theory directly into the model as a foundational constraint. They formalize this idea through the Multi-Agent Social Systems framework, which treats agent interactions as a dynamical system shaped by four specific priors drawn from social theory. This approach matters because ignoring these structures could lead to AI systems that produce unexpected or unstable group-level outcomes in real deployments.

Core claim

Agentic AI systems must be modeled with social theory as a structural prior, formalized as the Multi-Agent Social Systems (MASS) framework representing a dynamical system of information generation, local influence, and interaction structure, defined by four structural priors: strategic heterogeneity, networked-constrained dependence, co-evolution, and distributional instability.

What carries the argument

The MASS framework, a dynamical system class defined by information generation, local influence, and interaction structure, anchored in the four social theory priors of strategic heterogeneity, networked-constrained dependence, co-evolution, and distributional instability.

If this is right

AI agents with these priors allow formal propositions about how individual actions generate system-level social outcomes.
Evaluation of multi-agent AI should incorporate tests for adherence to these social structural priors.
Governance strategies for AI systems must address co-evolution and instability to manage emergent risks.
Modeling and simulation of AI societies should start from these priors rather than from scratch.

Where Pith is reading between the lines

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

If the framework holds, AI training methods would need to enforce these priors as constraints rather than relying on post-hoc alignment techniques.
This connects to challenges in multi-agent reinforcement learning where emergent norms often deviate from intended behaviors.
Testable extension: Deploy MASS-constrained agents in a simulated social network and compare their interaction patterns to unconstrained agents against real human data.

Load-bearing premise

The structural priors identified from human social theory transfer directly and sufficiently to artificial agents without modification or additional AI-specific validation.

What would settle it

A controlled simulation of multi-agent AI where agents without the four social priors produce identical emergent system behaviors to those with the priors, or where the priors fail to predict observed dynamics in actual AI deployments.

Figures

Figures reproduced from arXiv: 2605.07069 by Adrian Xuan Wei Lim, Iain J. Cruickshank, Kathleen M. Carley, Lynnette Hui Xian Ng.

**Figure 2.** Figure 2: MASS dynamics in MoltBook. Top row: Co-evolving reply-network. Bottom plots: [PITH_FULL_IMAGE:figures/full_fig_p007_2.png] view at source ↗

**Figure 3.** Figure 3: Statistical summary of MoltBook analysis [PITH_FULL_IMAGE:figures/full_fig_p020_3.png] view at source ↗

read the original abstract

Agentic AI systems are increasingly deployed not in isolation, but inside social environments populated by other agents and humans, such as in social media platforms, multi-agent LLM pipelines or autonomous robotics fleets. In these settings, system behavior emerges not from individual agents alone, but from the multi-agent interactions over time. Emergent dynamics of individuals in a social group have been long studied by social scientists in human contexts. \textbf{This position paper argues that agentic AI systems must be modeled with social theory as a structural prior, and formalizes a Multi-Agent Social Systems (MASS) framework for how agents interact and influence to generate system-level outcomes.} We represent MASS as a class of dynamical system of information generation, local influence and interaction structure, formulated by four structural priors anchored in social theory: strategic heterogeneity, networked-constrained dependence, co-evolution and distributional instability. We demonstrate the importance of each structural prior through formal propositions, and articulate a research agenda for how MASS should be modeled, evaluated and governed.

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.

Desk Editor's Note private letter to a colleague

This position paper argues for social theory as structural priors in multi-agent AI and names a MASS framework, but offers no new derivations or tests to show the priors are required rather than optional.

read the letter

The main thing here is a clear call to treat four ideas from social theory—strategic heterogeneity, networked-constrained dependence, co-evolution, and distributional instability—as fixed priors when modeling multi-agent AI systems. The authors formalize this as the MASS framework and sketch propositions showing how each prior shapes system-level outcomes in settings like platforms or robot fleets. That framing is useful for people already working at the AI-social science boundary, because it gives a compact way to talk about why uniform rationality assumptions often fall short in interactive environments. The paper does a decent job laying out the research agenda that would follow if the priors were adopted, including how to evaluate and govern such systems. What is missing is any demonstration that these priors cannot be replaced by simpler, AI-native rules such as explicit communication protocols plus uniform update dynamics. The argument rests on the transfer from human social theory without showing where that transfer breaks or what would falsify it. There are no derivations, simulations, or data checks, so the central claim stays at the level of advocacy. The propositions illustrate importance but do not test necessity. This is the kind of paper that belongs in a venue that publishes position pieces, because it surfaces a coherent set of questions even if the answers are not yet supplied. I would bring it to a reading group to see whether the four priors generate testable distinctions or just restate known difficulties with multi-agent coordination. It does not contain results I would cite directly, but the framing could be referenced when discussing design choices in agentic systems. A serious editor should send it to referees so the community can decide whether the analogy from human social theory is load-bearing or merely suggestive.

Referee Report

3 major / 1 minor

Summary. This position paper argues that agentic AI systems deployed in social environments must be modeled using social theory as a structural prior. It formalizes a Multi-Agent Social Systems (MASS) framework as a class of dynamical systems of information generation, local influence, and interaction structure, defined via four structural priors drawn from social theory (strategic heterogeneity, networked-constrained dependence, co-evolution, and distributional instability). The paper demonstrates the importance of each prior through formal propositions and outlines a research agenda for modeling, evaluation, and governance of such systems.

Significance. If the central claim holds, the MASS framework could offer a principled interdisciplinary approach to predicting emergent dynamics in multi-agent AI deployments such as social media platforms or robotic fleets, potentially improving system robustness and informing governance by importing established insights from social theory into AI design.

major comments (3)

[Abstract] Abstract and central claim: The assertion that agentic AI systems 'must be modeled with social theory as a structural prior' is not accompanied by any argument or counterexample showing why the four priors cannot be replaced by AI-native alternatives (e.g., homogeneous rationality plus explicit communication protocols); this necessity claim is load-bearing for the thesis but remains unestablished.
[MASS Framework] MASS framework definition: The manuscript describes MASS as a dynamical system formulated by the four priors but supplies no explicit equations, state-transition rules, or formal specification of how information generation, local influence, and interaction structure are mathematically encoded, preventing assessment of whether the priors function as structural constraints.
[Formal Propositions] Formal propositions: The propositions are presented as demonstrating the importance of each prior, yet the text contains no derivations, proofs, or concrete examples illustrating how violation of any prior alters system-level outcomes within the claimed dynamical system.

minor comments (1)

[Introduction] The term 'agentic AI' is used without an explicit definition distinguishing it from standard multi-agent systems; a brief clarification in the introduction would improve accessibility.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the constructive feedback on this position paper. We clarify that the manuscript is intended as a conceptual bridge between social theory and agentic AI rather than a complete mathematical treatise, and we address each major comment by committing to targeted revisions that strengthen the arguments without altering the core thesis.

read point-by-point responses

Referee: [Abstract] Abstract and central claim: The assertion that agentic AI systems 'must be modeled with social theory as a structural prior' is not accompanied by any argument or counterexample showing why the four priors cannot be replaced by AI-native alternatives (e.g., homogeneous rationality plus explicit communication protocols); this necessity claim is load-bearing for the thesis but remains unestablished.

Authors: We acknowledge that the necessity claim is central and currently rests on the accumulated evidence from social science rather than explicit contrasts with AI-native alternatives. In revision we will expand the abstract and add a short subsection contrasting the four priors with homogeneous rationality plus explicit protocols, using examples such as how the latter fails to generate distributional instability or co-evolution in multi-agent LLM deployments, thereby providing the requested argument. revision: yes
Referee: [MASS Framework] MASS framework definition: The manuscript describes MASS as a dynamical system formulated by the four priors but supplies no explicit equations, state-transition rules, or formal specification of how information generation, local influence, and interaction structure are mathematically encoded, preventing assessment of whether the priors function as structural constraints.

Authors: As a position paper the MASS framework is introduced at a high level to emphasize the structural priors. We agree that an explicit encoding would improve evaluability. In the revised manuscript we will insert a dedicated formalization subsection that defines the state space (agent strategies, network topology, information distribution) and sketches transition rules showing how each prior constrains the dynamics. revision: yes
Referee: [Formal Propositions] Formal propositions: The propositions are presented as demonstrating the importance of each prior, yet the text contains no derivations, proofs, or concrete examples illustrating how violation of any prior alters system-level outcomes within the claimed dynamical system.

Authors: The propositions are currently stated as direct implications from social theory. We accept that derivations and concrete examples are missing. We will revise the propositions section to include brief logical steps for each claim together with one illustrative example per prior (e.g., violation of co-evolution in robotic fleet coordination), thereby demonstrating their effect on system-level outcomes. revision: yes

Circularity Check

0 steps flagged

No circularity in MASS framework formalization

full rationale

The paper defines the Multi-Agent Social Systems (MASS) framework explicitly as a class of dynamical system formulated by four structural priors drawn from external social theory literature (strategic heterogeneity, networked-constrained dependence, co-evolution, distributional instability). It then articulates formal propositions to demonstrate the importance of each prior. This is a standard position-paper framework construction that adopts and illustrates external concepts rather than deriving outputs that reduce to the inputs by construction. No equations, fitted parameters, self-citations, or uniqueness theorems are shown to create a closed loop; the central claim is an advocacy position for adopting these priors, supported by external literature, without self-referential reduction or renaming of known results.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 1 invented entities

The central claim rests on the direct applicability of human social theory to AI agents and the sufficiency of the four listed priors as structural foundations.

axioms (2)

domain assumption Social theory from human contexts supplies valid structural priors for AI agent behavior
The paper states that emergent dynamics studied in human groups apply to agentic AI systems.
domain assumption Strategic heterogeneity, networked-constrained dependence, co-evolution, and distributional instability are the key structural priors
These four are presented as the anchors of the MASS framework.

invented entities (1)

MASS framework no independent evidence
purpose: To represent multi-agent interactions as a dynamical system of information generation, local influence, and interaction structure
New named framework introduced to organize the four priors.

pith-pipeline@v0.9.0 · 5494 in / 1308 out tokens · 38024 ms · 2026-05-13T07:54:28.711099+00:00 · methodology

Review history (3 revisions) →

discussion (0)

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

IndisputableMonolith/Foundation/RealityFromDistinction.lean reality_from_one_distinction unclear
We represent MASS as a class of dynamical system of information generation, local influence and interaction structure, formulated by four structural priors anchored in social theory: strategic heterogeneity, networked-constrained dependence, co-evolution and distributional instability.
IndisputableMonolith/Cost/FunctionalEquation.lean washburn_uniqueness_aczel unclear
Proposition 1 (Heterogeneity Non-Reducibility)... Proposition 4 (Non-existence of a Stationary Distribution)

Reference graph

Works this paper leans on

106 extracted references · 106 canonical work pages · 7 internal anchors

[1]

The theory of planned behavior.Organizational behavior and human decision processes, 50(2):179–211, 1991

Icek Ajzen. The theory of planned behavior.Organizational behavior and human decision processes, 50(2):179–211, 1991

work page 1991
[2]

Danial Amin, Joni Salminen, and Bernard J. Jansen. How to model ai agents as personas?: Applying the persona ecosystem playground to 41,300 posts on moltbook for behavioral insights, 2026. URLhttps://arxiv.org/abs/2603.03140

work page arXiv 2026
[3]

Position: Llm social simulations are a promising research method

Jacy Reese Anthis, Ryan Liu, Sean M Richardson, Austin C Kozlowski, Bernard Koch, Erik Brynjolfsson, James Evans, and Michael S Bernstein. Position: Llm social simulations are a promising research method. InForty-second International Conference on Machine Learning Position Paper Track, 2025

work page 2025
[4]

Distinguishing influence-based contagion from homophily-driven diffusion in dynamic networks.Proceedings of the National Academy of Sciences, 106(51):21544–21549, 2009

Sinan Aral, Lev Muchnik, and Arun Sundararajan. Distinguishing influence-based contagion from homophily-driven diffusion in dynamic networks.Proceedings of the National Academy of Sciences, 106(51):21544–21549, 2009

work page 2009
[5]

Modeling the diffusion of complex innovations as a process of opinion formation through social networks.PloS one, 13(5):e0196699, 2018

Valentina A Assenova. Modeling the diffusion of complex innovations as a process of opinion formation through social networks.PloS one, 13(5):e0196699, 2018

work page 2018
[6]

Princeton university press, 1997

Robert Axelrod.The Complexity of Cooperation: Agent-Based Models of Competition and Collaboration: Agent-Based Models of Competition and Collaboration. Princeton university press, 1997

work page 1997
[7]

The role of social networks in information diffusion

Eytan Bakshy, Itamar Rosenn, Cameron Marlow, and Lada Adamic. The role of social networks in information diffusion. InProceedings of the 21st international conference on World Wide Web, pages 519–528, 2012

work page 2012
[8]

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
[9]

Network science.Philosophical Transactions of the Royal Society A: Mathematical, Physical and Engineering Sciences, 371(1987), 2013

Albert-László Barabási. Network science.Philosophical Transactions of the Royal Society A: Mathematical, Physical and Engineering Sciences, 371(1987), 2013

work page 1987
[10]

Understanding incentives: Mechanism design becomes algorithm design

Yang Cai, Constantinos Daskalakis, and S Matthew Weinberg. Understanding incentives: Mechanism design becomes algorithm design. In2013 IEEE 54th Annual Symposium on Foundations of Computer Science, pages 618–627. IEEE, 2013

work page 2013
[11]

Assessing the impact of planned social change.Journal of MultiDisci- plinary Evaluation, 7(15):3–43, 2011

Donald T Campbell. Assessing the impact of planned social change.Journal of MultiDisci- plinary Evaluation, 7(15):3–43, 2011

work page 2011
[12]

Evolutionary information diffusion over heterogeneous social networks.IEEE Transactions on Signal and Information Processing over Networks, 2(4):595–610, 2016

Xuanyu Cao, Yan Chen, Chunxiao Jiang, and KJ Ray Liu. Evolutionary information diffusion over heterogeneous social networks.IEEE Transactions on Signal and Information Processing over Networks, 2(4):595–610, 2016

work page 2016
[13]

An approach for relating social structure to cognitive structure.Journal of Mathematical sociology, 12(2):137–189, 1986

Kathleen Carley. An approach for relating social structure to cognitive structure.Journal of Mathematical sociology, 12(2):137–189, 1986

work page 1986
[14]

Computational modeling for reasoning about the social behavior of humans.Computational and Mathematical Organization Theory, 15(1):47–59, 2009

Kathleen M Carley. Computational modeling for reasoning about the social behavior of humans.Computational and Mathematical Organization Theory, 15(1):47–59, 2009

work page 2009
[15]

Social cybersecurity: an emerging science.Computational and mathemat- ical organization theory, 26(4):365–381, 2020

Kathleen M Carley. Social cybersecurity: an emerging science.Computational and mathemat- ical organization theory, 26(4):365–381, 2020. 10

work page 2020
[16]

Simulation of stance perturbations

Peter Carragher, Lynnette Hui Xian Ng, and Kathleen M Carley. Simulation of stance perturbations. InInternational conference on social computing, behavioral-cultural modeling and prediction and behavior representation in modeling and simulation, pages 159–168. Springer, 2023

work page 2023
[17]

Statistical physics of social dynamics

Claudio Castellano, Santo Fortunato, and Vittorio Loreto. Statistical physics of social dynamics. Reviews of Modern Physics, 81(2):591–646, 2009

work page 2009
[18]

Why Do Multi-Agent LLM Systems Fail?

Mert Cemri, Melissa Z Pan, Shuyi Yang, Lakshya A Agrawal, Bhavya Chopra, Rishabh Tiwari, Kurt Keutzer, Aditya Parameswaran, Dan Klein, Kannan Ramchandran, et al. Why do multi-agent llm systems fail?arXiv preprint arXiv:2503.13657, 2025

work page internal anchor Pith review Pith/arXiv arXiv 2025
[19]

The spread of behavior in an online social network experiment.science, 329 (5996):1194–1197, 2010

Damon Centola. The spread of behavior in an online social network experiment.science, 329 (5996):1194–1197, 2010

work page 2010
[20]

A theory of heuristic and systematic information processing.Handbook of theories of social psychology, 1:246–266, 2012

Shelly Chaiken and Alison Ledgerwood. A theory of heuristic and systematic information processing.Handbook of theories of social psychology, 1:246–266, 2012

work page 2012
[21]

ChatEval: Towards Better LLM-based Evaluators through Multi-Agent Debate

Chi-Min Chan, Weize Chen, Yusheng Su, Jianxuan Yu, Wei Xue, Shanghang Zhang, Jie Fu, and Zhiyuan Liu. Chateval: Towards better llm-based evaluators through multi-agent debate. arXiv preprint arXiv:2308.07201, 2023

work page internal anchor Pith review arXiv 2023
[22]

Diversity Collapse in Multi-Agent LLM Systems: Structural Coupling and Collective Failure in Open-Ended Idea Generation

Nuo Chen, Yicheng Tong, Yuzhe Yang, Yufei He, Xueyi Zhang, Zou Qingyun, Qian Wang, and Bingsheng He. Diversity collapse in multi-agent llm systems: Structural coupling and collective failure in open-ended idea generation.arXiv preprint arXiv:2604.18005, 2026

work page internal anchor Pith review Pith/arXiv arXiv 2026
[23]

Examining identity drift in conversations of llm agents.arXiv preprint arXiv:2412.00804, 2024

Junhyuk Choi, Yeseon Hong, Minju Kim, and Bugeun Kim. Examining identity drift in conversations of llm agents.arXiv preprint arXiv:2412.00804, 2024

work page arXiv 2024
[24]

The two-step flow of communication in twitter-based public forums.Social science computer review, 33(6):696–711, 2015

Sujin Choi. The two-step flow of communication in twitter-based public forums.Social science computer review, 33(6):696–711, 2015

work page 2015
[25]

Deep reinforcement learning from human preferences.Advances in neural information processing systems, 30, 2017

Paul F Christiano, Jan Leike, Tom Brown, Miljan Martic, Shane Legg, and Dario Amodei. Deep reinforcement learning from human preferences.Advances in neural information processing systems, 30, 2017

work page 2017
[26]

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
[27]

Agenda setting

Renita Coleman, Maxwell McCombs, Donald Shaw, and David Weaver. Agenda setting. In The handbook of journalism studies, pages 167–180. Routledge, 2009

work page 2009
[28]

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
[29]

Improv- ing factuality and reasoning in language models through multiagent debate

Yilun Du, Shuang Li, Antonio Torralba, Joshua B Tenenbaum, and Igor Mordatch. Improv- ing factuality and reasoning in language models through multiagent debate. InForty-first international conference on machine learning, 2024

work page 2024
[30]

Llm-botguard: A novel framework for detecting llm-driven bots with mixture of experts and graph neural networks.IEEE Transactions on Computational Social Systems, 2025

Jinglong Duan, Weihua Li, Quan Bai, Minh Nguyen, Xiaodan Wang, and Jianhua Jiang. Llm-botguard: A novel framework for detecting llm-driven bots with mixture of experts and graph neural networks.IEEE Transactions on Computational Social Systems, 2025

work page 2025
[31]

Collective honesty? experimental evidence on the effectiveness of honesty nudging for teams.Frontiers in Psychology, 12:684755, 2021

Yuri Dunaiev and Menusch Khadjavi. Collective honesty? experimental evidence on the effectiveness of honesty nudging for teams.Frontiers in Psychology, 12:684755, 2021

work page 2021
[32]

Social role theory.Handbook of theories of social psychology, 2(9):458–476, 2012

Alice H Eagly and Wendy Wood. Social role theory.Handbook of theories of social psychology, 2(9):458–476, 2012

work page 2012
[33]

Quantifying the effect of sentiment on information diffusion in social media.PeerJ Computer Science, 1:e26, 2015

Emilio Ferrara and Zeyao Yang. Quantifying the effect of sentiment on information diffusion in social media.PeerJ Computer Science, 1:e26, 2015. 11

work page 2015
[34]

Measuring coordinated versus spontaneous activity in online social movements.New Media & Society, 25(11):3065–3092, 2023

Camille Francois, Vladimir Barash, and John Kelly. Measuring coordinated versus spontaneous activity in online social movements.New Media & Society, 25(11):3065–3092, 2023

work page 2023
[35]

Social positions in influence networks.Social networks, 19(3):209–222, 1997

Noah E Friedkin and Eugene C Johnsen. Social positions in influence networks.Social networks, 19(3):209–222, 1997

work page 1997
[36]

Tiktok and the algorithmic transformation of social media publics: From social networks to social interest clusters.New Media & Society, 28(3):1019–1036, 2026

Paolo Gerbaudo. Tiktok and the algorithmic transformation of social media publics: From social networks to social interest clusters.New Media & Society, 28(3):1019–1036, 2026

work page 2026
[37]

Structuration theory: past, present and future

Anthony Giddens. Structuration theory: past, present and future. InGiddens’ theory of structuration, pages 201–221. Routledge, 2014

work page 2014
[38]

Threshold models of collective behavior.American journal of sociology, 83(6):1420–1443, 1978

Mark Granovetter. Threshold models of collective behavior.American journal of sociology, 83(6):1420–1443, 1978

work page 1978
[39]

Economic action and social structure: The problem of embeddedness

Mark Granovetter. Economic action and social structure: The problem of embeddedness. American journal of sociology, 91(3):481–510, 1985

work page 1985
[40]

The strength of weak ties.American journal of sociology, 78(6): 1360–1380, 1973

Mark S Granovetter. The strength of weak ties.American journal of sociology, 78(6): 1360–1380, 1973

work page 1973
[41]

Topological measures for identifying and predicting the spread of complex contagions.Nature communications, 12(1):4430, 2021

Douglas Guilbeault and Damon Centola. Topological measures for identifying and predicting the spread of complex contagions.Nature communications, 12(1):4430, 2021

work page 2021
[42]

Opinion dynamics and bounded confidence models, analysis, and simulation.Journal of Artificial Societies and Social Simulation, 5(3), 2002

Rainer Hegselmann 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
[43]

Measuring Massive Multitask Language Understanding

Dan Hendrycks, Collin Burns, Steven Basart, Andy Zou, Mantas Mazeika, Dawn Song, and Jacob Steinhardt. Measuring massive multitask language understanding.arXiv preprint arXiv:2009.03300, 2020

work page internal anchor Pith review Pith/arXiv arXiv 2009
[44]

Modeling information access in construct

Brian Hirshman, Michael K Martin, and Kathleen M Carley. Modeling information access in construct. Technical report, Technical report, Carnegie Mellon University School of Computer Science, 2008

work page 2008
[45]

Metagpt: Meta programming for a multi-agent collaborative framework

Sirui Hong, Mingchen Zhuge, Jonathan Chen, Xiawu Zheng, Yuheng Cheng, Jinlin Wang, Ceyao Zhang, Zili Wang, Steven Ka Shing Yau, Zijuan Lin, et al. Metagpt: Meta programming for a multi-agent collaborative framework. InThe twelfth international conference on learning representations, 2023

work page 2023
[46]

Evolution of social norms in llm agents using natural language.arXiv preprint arXiv:2409.00993, 2024

Ilya Horiguchi, Takahide Yoshida, and Takashi Ikegami. Evolution of social norms in llm agents using natural language.arXiv preprint arXiv:2409.00993, 2024

work page arXiv 2024
[47]

Nomiclaw: Emergent trust and strategic argumentation in llms during collaborative law-making

Asutosh Hota and Jussi PP Jokinen. Nomiclaw: Emergent trust and strategic argumentation in llms during collaborative law-making. InProceedings of the AAAI/ACM Conference on AI, Ethics, and Society, volume 8, pages 1278–1289, 2025

work page 2025
[48]

Simulating rumor spreading in social networks using llm agents.arXiv preprint arXiv:2502.01450, 2025

Tianrui Hu, Dimitrios Liakopoulos, Xiwen Wei, Radu Marculescu, and Neeraja J Yad- wadkar. Simulating rumor spreading in social networks using llm agents.arXiv preprint arXiv:2502.01450, 2025

work page arXiv 2025
[49]

Tracking china’s cross-strait bot networks against taiwan

Charity S Jacobs, Lynnette Hui Xian Ng, and Kathleen M Carley. Tracking china’s cross-strait bot networks against taiwan. InInternational conference on social computing, behavioral- cultural modeling and prediction and behavior representation in modeling and simulation, pages 115–125. Springer, 2023

work page 2023
[50]

SWE-bench: Can Language Models Resolve Real-World GitHub Issues?

Carlos E Jimenez, John Yang, Alexander Wettig, Shunyu Yao, Kexin Pei, Ofir Press, and Karthik Narasimhan. Swe-bench: Can language models resolve real-world github issues? arXiv preprint arXiv:2310.06770, 2023

work page internal anchor Pith review Pith/arXiv arXiv 2023
[51]

Habermas on strategic and communicative action.Political theory, 19(2): 181–201, 1991

James Johnson. Habermas on strategic and communicative action.Political theory, 19(2): 181–201, 1991. 12

work page 1991
[52]

Social bots and their coordination during online campaigns: a survey.IEEE Transactions on Computational Social Systems, 9(2):530–545, 2021

Tuja Khaund, Baris Kirdemir, Nitin Agarwal, Huan Liu, and Fred Morstatter. Social bots and their coordination during online campaigns: a survey.IEEE Transactions on Computational Social Systems, 9(2):530–545, 2021

work page 2021
[53]

Semi-Supervised Classification with Graph Convolutional Networks

Thomas N Kipf and Max Welling. Semi-supervised classification with graph convolutional networks.arXiv preprint arXiv:1609.02907, 2016

work page internal anchor Pith review Pith/arXiv arXiv 2016
[54]

A survey of llm-driven ai agent communication: Protocols, security risks, and defense countermeasures.arXiv preprint arXiv:2506.19676, 2025

Dezhang Kong, Shi Lin, Zhenhua Xu, Zhebo Wang, Minghao Li, Yufeng Li, Yilun Zhang, Hujin Peng, Xiang Chen, Zeyang Sha, et al. A survey of llm-driven ai agent communication: Protocols, security risks, and defense countermeasures.arXiv preprint arXiv:2506.19676, 2025

work page arXiv 2025
[55]

Evaluating behavioral alignment in conflict dialogue: A multi-dimensional comparison of llm agents and humans

Deuksin Kwon, Kaleen Shrestha, Bin Han, Elena Hayoung Lee, and Gale Lucas. Evaluating behavioral alignment in conflict dialogue: A multi-dimensional comparison of llm agents and humans. InProceedings of the 2025 Conference on Empirical Methods in Natural Language Processing, pages 16377–16391, 2025

work page 2025
[56]

Large language models miss the multi-agent mark

Emanuele La Malfa, Gabriele La Malfa, Samuele Marro, Jie M Zhang, Elizabeth Black, Michael Luck, Philip Torr, and Michael Wooldridge. Large language models miss the multi- agent mark.arXiv preprint arXiv:2505.21298, 2025

work page arXiv 2025
[57]

The “panel” as a new tool for measuring opinion.Public Opinion Quarterly, 2(4):596–612, 1938

Paul Lazarsfeld and Marjorie Fiske. The “panel” as a new tool for measuring opinion.Public Opinion Quarterly, 2(4):596–612, 1938

work page 1938
[58]

The logical and mathematical foundation of latent structure analysis.Studies in social psychology in world war II Vol

Paul F Lazarsfeld. The logical and mathematical foundation of latent structure analysis.Studies in social psychology in world war II Vol. IV: Measurement and prediction, pages 362–412, 1950

work page 1950
[59]

Llm generated persona is a promise with a catch

Ang Li, Haozhe Chen, Hongseok Namkoong, and Tianyi Peng. Llm generated persona is a promise with a catch. InThe Thirty-Ninth Annual Conference on Neural Information Processing Systems Position Paper Track

work page
[60]

Camel: Communicative agents for" mind" exploration of large language model society.Advances in neural information processing systems, 36:51991–52008, 2023

Guohao Li, Hasan Hammoud, Hani Itani, Dmitrii Khizbullin, and Bernard Ghanem. Camel: Communicative agents for" mind" exploration of large language model society.Advances in neural information processing systems, 36:51991–52008, 2023

work page 2023
[61]

Does socialization emerge in AI agent society? A case study of Moltbook

Ming Li, Xirui Li, and Tianyi Zhou. Does socialization emerge in ai agent society? a case study of moltbook.arXiv preprint arXiv:2602.14299, 2026

work page arXiv 2026
[62]

Measuring fine-grained negotiation tactics of humans and llms in diplomacy.arXiv preprint arXiv:2512.18292, 2025

Wenkai Li, Lynnette Hui Xian Ng, Andy Liu, and Daniel Fried. Measuring fine-grained negotiation tactics of humans and llms in diplomacy.arXiv preprint arXiv:2512.18292, 2025

work page arXiv 2025
[63]

Individual and shared meanings.Research on Language & Social Interaction, 10(3-4):341–373, 1977

Bonnie Litowitz. Individual and shared meanings.Research on Language & Social Interaction, 10(3-4):341–373, 1977

work page 1977
[64]

AgentBench: Evaluating LLMs as Agents

Xiao Liu, Hao Yu, Hanchen Zhang, Yifan Xu, Xuanyu Lei, Hanyu Lai, Yu Gu, Hangliang Ding, Kaiwen Men, Kejuan Yang, et al. Agentbench: Evaluating llms as agents.arXiv preprint arXiv:2308.03688, 2023

work page internal anchor Pith review Pith/arXiv arXiv 2023
[65]

Network studies of social influence.Sociological Methods & Research, 22(1):127–151, 1993

Peter V Marsden and Noah E Friedkin. Network studies of social influence.Sociological Methods & Research, 22(1):127–151, 1993

work page 1993
[66]

The landscape of emerging ai agent architectures for reasoning, planning, and tool calling: A survey.arXiv preprint arXiv:2404.11584, 2024

Tula Masterman, Sandi Besen, Mason Sawtell, and Alex Chao. The landscape of emerging ai agent architectures for reasoning, planning, and tool calling: A survey.arXiv preprint arXiv:2404.11584, 2024

work page arXiv 2024
[67]

The agenda-setting function of mass media.Public opinion quarterly, 36(2):176–187, 1972

Maxwell E McCombs and Donald L Shaw. The agenda-setting function of mass media.Public opinion quarterly, 36(2):176–187, 1972

work page 1972
[68]

The role-set: Problems in sociological theory.The British journal of sociology, 8(2):106–120, 1957

Robert K Merton. The role-set: Problems in sociological theory.The British journal of sociology, 8(2):106–120, 1957

work page 1957
[69]

The social construction of meaning: An alternative perspective on information sharing.Information systems research, 14(1):87–106, 2003

Shaila M Miranda and Carol S Saunders. The social construction of meaning: An alternative perspective on information sharing.Information systems research, 14(1):87–106, 2003. 13

work page 2003
[70]

Springer, 2015

Amiya Mukherjee et al.Differential topology. Springer, 2015

work page 2015
[71]

Identifying cross-platform user relationships in 2020 us election fraud and protest discussions.Online Social Networks and Media, 33:100245, 2023

Isabel Murdock, Kathleen M Carley, and Osman Ya ˘gan. Identifying cross-platform user relationships in 2020 us election fraud and protest discussions.Online Social Networks and Media, 33:100245, 2023

work page 2020
[72]

Pro or anti? a social influence model of online stance flipping.IEEE Transactions on Network Science and Engineering, 10(1):3–19, 2022

Lynnette Hui Xian Ng and Kathleen M Carley. Pro or anti? a social influence model of online stance flipping.IEEE Transactions on Network Science and Engineering, 10(1):3–19, 2022

work page 2022
[73]

A combined synchronization index for evaluating collective action social media.Applied network science, 8(1):1, 2023

Lynnette Hui Xian Ng and Kathleen M Carley. A combined synchronization index for evaluating collective action social media.Applied network science, 8(1):1, 2023

work page 2023
[74]

A global comparison of social media bot and human characteristics.Scientific Reports, 15(1):10973, 2025

Lynnette Hui Xian Ng and Kathleen M Carley. A global comparison of social media bot and human characteristics.Scientific Reports, 15(1):10973, 2025

work page 2025
[75]

Recruitment promotion via twitter: a network- centric approach of analyzing community engagement using social identity.Digital Govern- ment: Research and Practice, 4(4):1–17, 2023

Lynnette Hui Xian Ng and Iain J Cruickshank. Recruitment promotion via twitter: a network- centric approach of analyzing community engagement using social identity.Digital Govern- ment: Research and Practice, 4(4):1–17, 2023

work page 2023
[76]

Building bridges between users and content across multiple platforms during natural disasters

Lynnette Hui Xian Ng, Iain J Cruickshank, and David Farr. Building bridges between users and content across multiple platforms during natural disasters. InProceedings of the 17th ACM Web Science Conference 2025, pages 499–503, 2025

work page 2025
[77]

Aurasight: generating realistic social media data.arXiv preprint arXiv:2509.08927, 2025

Lynnette Hui Xian Ng, Bianca NY Kang, and Kathleen M Carley. Aurasight: generating realistic social media data.arXiv preprint arXiv:2509.08927, 2025

work page arXiv 2025
[78]

the enormous turnip

Hiroko Oe and Junkichi Mochizuki. An examination of nudge policy through the lens of" the enormous turnip": Fostering collective action and persistence.International Journal of Academic Research in Progressive Education and Development, 14(2):1746–1772, 2025

work page 2025
[79]

Cambridge university press, 1990

Elinor Ostrom.Governing the commons: The evolution of institutions for collective action. Cambridge university press, 1990

work page 1990
[80]

Under- standing the trolling phenomenon: The automated detection of bots and cyborgs in the social media.Journal of Information Warfare, 15(4):100–111, 2016

Jarkko Paavola, Tuomo Helo, Harri Jalonen, Miika Sartonen, and Aki-Mauri Huhtinen. Under- standing the trolling phenomenon: The automated detection of bots and cyborgs in the social media.Journal of Information Warfare, 15(4):100–111, 2016

work page 2016

Showing first 80 references.