Constructing a Dynamic Causal Inference Framework for Digital

Abstract

In digital scenarios, traditional linear causal inference models struggle to accurately depict complex causal relationships due to the breadth of data sources and intricate environmental dynamics. This study establishes a dynamic causal inference model suited to digital contexts based on dynamic systems theory. The model comprises multiple layers: data input, generation and evaluation of causal relationships, feedback and learning of causal relationships, and visualization of outcomes. This model automatically adjusts and refines causal structures, exhibiting adaptability and stability. Through time-varying modelling and machine learning optimization strategies, it achieves greater flexibility and interpretability in high-dimensional, non-stationary data scenarios. The research serves as a tool to address practical challenges in digital governance, intelligent decision-making, and social science experimentation.