From Agent Simulation to Social Simulator: A Comprehensive Review (Part 2)

TL;DR

This paper contests the limits of traditional ABM in causal inference and proposes a comprehensive computational experiment framework that combines ABM with counterfactual experimentation to uncover micro-to-macro causal mechanisms. It articulates a three-layer analytic structure—observational (forward causation), interventional (causal effects), and mechanism (reverse causation)—grounded in world models and generative narratives, including both external data-driven and internal agent beliefs. A full case study on algorithmic behavior in O2O platform ecosystems demonstrates involution among riders and evaluates governance interventions, validated against real-world benchmarks and supported by meta-models and SEM analyses. The work advances a practical, generative approach to studying complex social systems, with significant implications for designing fairer, more robust platform governance policies and for applying causal reasoning to open, adaptive environments.

Abstract

The study of system complexity primarily has two objectives: to explore underlying patterns and to develop theoretical explanations. Pattern exploration seeks to clarify the mechanisms behind the emergence of system complexity, while theoretical explanations aim to identify the fundamental causes of this complexity. Laws are generally defined as mappings between variables, whereas theories offer causal explanations of system behavior. Agent Based Modeling(ABM) is an important approach for studying complex systems, but it tends to emphasize simulation over experimentation. As a result, ABM often struggles to deeply uncover the governing operational principles. Unlike conventional scenario analysis that relies on human reasoning, computational experiments emphasize counterfactual experiments-that is, creating parallel worlds that simulate alternative "evolutionary paths" of real-world events. By systematically adjusting input variables and observing the resulting changes in output variables, computational experiments provide a robust tool for causal inference, thereby addressing the limitations of traditional ABM. Together, these methods offer causal insights into the dynamic evolution of systems. This part can help readers gain a preliminary understanding of the entire computational experiment method, laying the foundation for the subsequent study.

Figures (12)

• Figure 1: Fromm’s Classification of Emergence.
• Figure 2: Forward Sequential Causal Logic Tree of Simon’s Adaptation Theory and the Emergence of System Complexity.
• Figure 3: Abstract Framework of Causal Emergence.
• Figure 4: The process model of the computational experiment method.
• Figure 5: The comprehensive modeling framework based on computational experiment design.