From Agent Simulation to Social Simulator: A Comprehensive Review (Part 1)

TL;DR

This paper surveys the historical development and foundational concepts of agent-based modeling (ABM) for social systems, contrasting bottom-up ABM with top-down dynamics. It articulates three core modeling units—agent models, environmental models, and rule models—and discusses design principles that balance abstraction with realism (KISS vs. KIDS) along with causal inference challenges. The section then catalogs classic ABM cases across thought experiments (Sugarscape, Schelling, landscape theory), mechanism exploration (ASM, RebeLand, epidemic spread), and parallel optimization (Island economy via The AI Economist, Virtual Taobao, Autonomous Driving). It highlights how ABM serves as an artificial social laboratory, capable of exploration, explanation, and policy testing, while noting limitations in causal inference and calibration, and sketches a roadmap for integrating empirical data and observational methods in future work.

Abstract

This is the first part of the comprehensive review, focusing on the historical development of Agent-Based Modeling (ABM) and its classic cases. It begins by discussing the development history and design principles of Agent-Based Modeling (ABM), helping readers understand the significant challenges that traditional physical simulation methods face in the social domain. Then, it provides a detailed introduction to foundational models for simulating social systems, including individual models, environmental models, and rule-based models. Finally, it presents classic cases of social simulation, covering three types: thought experiments, mechanism exploration, and parallel optimization.

Figures (19)

• Figure 1: The core idea of social simulation.
• Figure 2: The development of ABM.
• Figure 3: The balance between model design principles and actual conditions.
• Figure 4: The balance between the design principles of the parsimony method model and the actual situation.
• Figure 5: The balanced relationship between realistic method model design principles and real situations.