Exploring the mechanism-based explanation of complex social phenomena: A social simulation study of the diffusion of college students drinking behavior in friendship networks

This dissertation addresses the challenge of constructing and testing mechanism-based explanations (MBEs) for emergent social phenomena, proposing a generative research paradigm to navigate the complexities of social systems. Methodologically, this paradigm integrates a Pattern-oriented Agent-Based Model (ABM) with Reinforcement Learning (RL), is grounded in a structural individualism epistemology, and utilizes a refined "Coleman's boat" diagram as a multi-level analytical framework. The paradigm is applied to the substantive case of drinking behavior diffusion within college friendship networks to explain how social capital motives and the cultural significance of drinking generate this emergent pattern. The simulation results indicate that the diffusion of drinking behavior emerges from a co-evolutionary process between individual actions and friendship network structures. Specifically, the findings identify social capital-driven motives and social exclusion mechanisms as indispensable explanatory factors. The primary contribution of this research is the formalization of this replicable, ABM-based paradigm for testing MBEs under the principles of abductive reasoning. This work refines the generative methodology toolbox of analytical sociology and offers a robust approach to addressing enduring challenges in the study of social complexity.