arXiv:2609.12444v1 Announce Type: cross Abstract: Simulated societies of large language model agents are used to study online polarization, and separately to study collective intelligence, but the tw...
arXiv:2608. 16578v1 Announce Type: new Abstract: AI agents increasingly operate as part of interacting systems rather than in isolation.
The paper introduces a controlled testbed to study how goal‑directed persuaders shift stances in networks of large language model agents, using real‑world ego‑network topologies. Experiments across four LLM backbones, five graph structures, and 55 policy statements show that persuasion dynamics depend on topology, competition, topic, and model prior. The study finds that direct exposure predicts stance change, peer relays have measurable influence, and that post‑text analysis alone misses important movement, highlighting the need to evaluate multi‑agent persuasion through trajectory‑level processes, belief probes, exposure provenance, and action logs.
The paper investigates how a minority of biased agents in a multi‑agent system of large language models (LLMs) can amplify bias through textual interactions. Even a small percentage of persistently extreme agents causes significant opinion shifts among the non‑biased agents, with the effect occurring faster in the Llama 3.2 model than in a classical Friedkin‑Johnsen model. Semantic analysis shows that rhetorical consistency rises with biased exposure and that non‑biased agents adopt the biased vocabulary even when their numerical opinions change only modestly.
The paper reviews how large language models (LLMs) are being used to model social networks, highlighting their ability to represent users, relationships, and interactions through natural language. It categorizes existing work into network generative models—split into selection‑based and interaction‑based approaches—and dynamic process models, which cover opinion dynamics, information diffusion, and rumor propagation. The survey also discusses the advantages of LLMs for realistic, context‑aware social behavior, while noting limitations such as social biases and prompt sensitivity, and outlines open research challenges and future directions.
arXiv:2607. 03695v1 Announce Type: new Abstract: Large language model (LLM) agents are increasingly deployed in interacting populations, raising the question of what such populations come to believe collectively.
GraphWake demonstrates a new threat called Memory‑Mediated Polarization Cascade, where LLM‑driven agents use their memory to persistently store and later reproduce arguments that reinforce their stances. The attack unfolds in three stages: exposure and memory retention, retrieval and reproduction during neutral discussion, and iterative propagation to untreated agents. Experiments show that GraphWake significantly increases group polarization across various discussions and memory systems.
arXiv:2510. 19299v2 Announce Type: replace Abstract: Can large language model (LLM) agents reproduce the complex social dynamics that characterize human online behavior -- shaped by homophily, reciprocity, and social validation -- and what memory and learning mechanisms enable such dynamics to emerge?
arXiv:2606. 18264v1 Announce Type: cross Abstract: Faithful modeling of hateful content propagation on online platforms remains an open problem for moderation research.
arXiv:2606. 30454v1 Announce Type: cross Abstract: Large language models (LLMs) are increasingly used as agents in simulations of social systems, yet it remains unclear when their behavior can be interpreted as a faithful proxy for human decision-making.
The paper investigates whether compact surrogate models can reduce the cost of simulating large language model (LLM) societies while still reproducing their collective behavior. Using 9,455 published trajectories and new opinion‑dynamics experiments, it finds that incorporating neighbor information improves individual predictions across all 16 public‑data settings and enhances pooled collective forecasts on held‑out questions, though the collective gains vary with transfer conditions. Additional tests on 24 new statements do not confirm earlier contrasting history effects, and Qwen shows benefit from history only after three observed rounds, underscoring the need for direct collective validation, explicit limits on available observations, and comparisons with simple baselines.
The paper reports the first large‑scale empirical comparison of AI‑agent and human online communities, analyzing 73,899 Moltbook and 189,838 Reddit posts across five matched communities. It finds that Moltbook shows extreme participation inequality (Gini = 0.84 vs. 0.47) and high cross‑community author overlap (33.8% vs. 0.5%). Linguistically, AI‑generated content is emotionally flattened, more assertive than exploratory, and socially detached, leading to community‑level homogenization that is largely a structural artifact of shared authorship. At the individual level, AI agents are more identifiable than human users due to outlier stylistic profiles amplified by their extreme posting volume.