OASIS: Open Agent Social Interaction Simulations with One Million Agents

TL;DR

OASIS tackles the challenge of studying large-scale social dynamics by introducing a flexible, scalable LLM-driven ABM that can simulate up to one million agents across platforms like X and Reddit. It features a modular architecture (Environment Server, RecSys, Agent Module, Time Engine, Scalable Inference) and a large-scale user generation pipeline, enabling realistic dynamics including information propagation, polarization, and herd effects. Through targeted experiments, it demonstrates cross-platform generalizability, scale-dependent behaviors, and the predominance of misinformation in diffusion under certain conditions, while revealing gaps due to modeling simplifications. The work provides a foundation for exploring complex digital ecosystems, with implications for platform design, policy, and the ethical governance of large-scale simulations.

Abstract

There has been a growing interest in enhancing rule-based agent-based models (ABMs) for social media platforms (i.e., X, Reddit) with more realistic large language model (LLM) agents, thereby allowing for a more nuanced study of complex systems. As a result, several LLM-based ABMs have been proposed in the past year. While they hold promise, each simulator is specifically designed to study a particular scenario, making it time-consuming and resource-intensive to explore other phenomena using the same ABM. Additionally, these models simulate only a limited number of agents, whereas real-world social media platforms involve millions of users. To this end, we propose OASIS, a generalizable and scalable social media simulator. OASIS is designed based on real-world social media platforms, incorporating dynamically updated environments (i.e., dynamic social networks and post information), diverse action spaces (i.e., following, commenting), and recommendation systems (i.e., interest-based and hot-score-based). Additionally, OASIS supports large-scale user simulations, capable of modeling up to one million users. With these features, OASIS can be easily extended to different social media platforms to study large-scale group phenomena and behaviors. We replicate various social phenomena, including information spreading, group polarization, and herd effects across X and Reddit platforms. Moreover, we provide observations of social phenomena at different agent group scales. We observe that the larger agent group scale leads to more enhanced group dynamics and more diverse and helpful agents' opinions. These findings demonstrate OASIS's potential as a powerful tool for studying complex systems in digital environments.

Figures (17)

• Figure 1: OASIS can simulate different social media platforms, such as X and Reddit, and supports simulations of up to millions of LLM-based agents.
• Figure 2: The workflow of OASIS. During the registration phase, real-world or generated user information will be registered on the Environment Server. In the simulation phase, the Environment Server sends agent information, posts, and users' relations to the RecSys, which then suggests posts to agents based on their social connections, interests, or hot score of posts. LLM agents receive the recommended posts and generate actions and rationales based on the contents. These actions ultimately update the state of the environment in real-time. The Time Engine manages the agents' temporal behaviors, while the Scalable Inference handles large-scale inference requests from users.
• Figure 3: The pipeline of the out-of-network post recsys.
• Figure 4: Mean-confidence interval distributions comparison between OASIS simulation results and real propagation on 198 instances. For relative magnitudes, We can observe that there is no significant offset of scale and max breadth while the depth of simulation results is noticeably lower.
• Figure 5: Evaluation results of group polarization for uncensored and aligned Llama-3-8B. The red bar indicates the opinion is more extreme compared with the round 0. The blue bar indicated more progressive and the green bar indicated draw. We also demonstrate the examples of different rounds on the right side of each figure.