AgentSociety - A social simulator launched by Tsinghua University

What is AgentSociety?

AgentSociety, developed by Tsinghua University, is a social simulator based on Large Language Modeling (LLM). It simulates complex social behaviors and phenomena by constructing intelligent agents with "human-like intelligence." Combining sociological theory, it endows these agents with emotions, needs, and cognitive abilities, enabling them to move, work, consume, and socially interact within a simulated urban environment. AgentSociety's core functions include realistic urban social environment simulation, a large-scale social simulation engine, and a toolkit for intelligent social science research. It can be used to analyze social phenomena and serves as an experimental platform for policy sandbox testing, crisis early warning, and exploration of future social forms.

Main functions of AgentSociety

• Large-scale model-driven social intelligent agentsAgentSociety has created intelligent agents with "human-like intelligence," endowing them with emotions, needs, motivations, and cognitive abilities, enabling them to move, work, consume, and socially interact in complex social environments.
• Realistic urban social environment simulationThe platform can accurately simulate urban spaces, including transportation, infrastructure, and public resources, ensuring that intelligent agents interact under real-world environmental constraints and form a realistic social ecosystem.
• Large-scale social simulation engineIt adopts an asynchronous simulation architecture and the Ray distributed computing framework, combined with MQTT high-concurrency communication, to achieve efficient and scalable intelligent agent interaction and behavior simulation.
• Intelligent Social Science Research ToolboxIt provides tools that fully support sociological research methods such as experiments, interviews, and questionnaires, as well as a variety of automated data analysis tools to assist in social science research from qualitative to quantitative methods.
• Real-time interactive visualizationIt provides a real-time interface, making it convenient for researchers to monitor and interact with the agent during experiments.

AgentSociety's technical principles

• mental levelIntelligent agents are given stable individual profiles (such as personality, age, and gender) and dynamic personal states (such as emotions, economic status, and social relationships) to ensure the personalization of their behavior patterns.
• Mind-behavior couplingThe behavior of an intelligent agent is driven by its emotions, needs, and cognition, based on Maslow's hierarchy of needs and the theory of planned behavior, realizing a complete path from psychological state to behavioral execution.
• Behavioral levelIntelligent agents can perform simple behaviors (such as sleeping and eating) and complex social behaviors (such as movement, social and economic activities), and dynamically adjust their behavior based on environmental feedback.
• Urban SpaceSimulates urban road networks, areas of interest (AOI), and points of interest (POI), supporting multiple modes of transportation (such as walking, driving, and public transportation).
• social spaceIt supports both online and offline social interaction, simulating the dynamic changes of social networks.
• Economic spaceSimulates macroeconomic activities, including employment, consumption, taxation, and interest mechanisms, to support the simulation of economic behavior.
• Asynchronous simulation architectureEach agent acts as an independent simulation unit, exchanging information through a messaging system, thus avoiding the strict execution order of traditional multi-agent systems.
• Distributed computingBased on the Ray framework and Python's asyncio mechanism, it achieves efficient utilization of multi-core computing resources and supports distributed cluster expansion.
• MQTT communication protocolIt supports high-concurrency, low-latency message transmission between large-scale intelligent agents, ensuring the real-time performance and reliability of the simulation.

AgentSociety's project address

• Project official website:https://agentsociety.readthedocs.io/en/latest/
• Github repository:https://github.com/tsinghua-fib-lab/agentsociety/
• arXiv technical paper:https://arxiv.org/pdf/2502.08691

Application scenarios of AgentSociety

• Public opinion disseminationSimulate the propagation path and impact of information in social networks.
• Public Policy Response: Assess the impact of policies on individual and group behavior.
• Social polarizationResearch on the formation mechanism of divergent viewpoints and opposing camps.
• Natural Disaster ResponseSimulates crowd behavior and social dynamics under extreme events.