Abstract: The robust growth of computational social science is reshaping the research paradigms and methodological systems of political science. As a generative explanatory approach, agent-based modeling reconstructs multi-level interaction processes within complex systems by relying on computer platforms and it enjoys unique advantages in solving the cross-level inference problem of “micro-level individual behavior—macro-level social emergence”. However, existing research predominantly focuses on Western institutional environments or single-policy scenarios, failing to adequately capture the complex vertical game logic under China’s party-state system. Based on Chinese political practice, The research selects “Central Authority Intensity” and “Local Action Space” as core dimensions, and constructs an agent-based modeling analytical framework for “central-local interactions” under the party-state system. By setting behavioral rules and interaction strategies for heterogeneous agents, this framework attempts to replicate dynamic processes in virtual space, including policy implementation deviations under the pressure-based system, the diffusion of local innovations under the tournament system, and governance involution under the logic of blame avoidance. The research translates the core characteristics of China’s party-state system into a computable experimental design and provides a dynamic evolutionary perspective for understanding political operations within complex institutional environments. Consequently, it opens a new methodological path for constructing an autonomous knowledge system in China’s political science.

Key words: party-state system, computational political science, Agent-based Modeling (ABM), central-local interactions