Replication Data and Scripts for: The mechanism of generative AI's construction of cultural identity

This repository contains the completely anonymized, text-mined composite indicators (N=11,628) and the corresponding R execution scripts used to validate the parallel mediation model for the manuscript under review. To comply with the highest standards of computational social science transparency and reproducibility, this repository provides all necessary materials for exact replication: Finalized & Neutralized Dataset (S1_Dataset.csv): The finalized, strictly neutral dataset containing all core continuous variables. The categorical variable dummy coding is strictly defined as 0 = UGC and 1 = AIGC to perfectly align with the structural path coefficients. Comprehensive Replication Script (Replication_Script_R1.R): The complete R script providing the full lavaan package workflow. It includes precise 5000-resample Bias-Corrected (BCa) bootstrapping outputs for testing parallel mediation, addresses endogenous residual covariances for accurate model fit, and incorporates a Multi-Group SEM (MG-SEM) architecture to rigorously verify the "Authenticity Paradox" and formally rule out Simpson's Paradox. Exhaustive Lexicons (Supplementary Materials.docx): To ensure complete computational transparency, the full custom "Technology Acceptance Lexicon" and "Sentiment Lexicon" utilized for the baseline semantic extraction of the Douyin comments are publicly available here.