AI-Enhanced Teacher Agency and English Instructional Resource Adoption in Chinese Higher Education: A Multi-Layer Ecological Study of English Major Teachers' Performance and Pedagogical Agency

Abstract

The use of Artificial Intelligence (AI) tools in higher education has revolutionized the teaching of English in Chinese universities, where English major teachers find themselves in a complex landscape of pedagogical agency, institutional demands, and technological integration. Although there has been an increased interest in research on AI in education, the available empirical studies have tended to focus on either teaching performance or adoption of the AI classroom resources, or ecological contextual factors, without using a shared analytical lens. This study aims to fill this gap by developing and testing the Multi-Layer Ecological AI-Pedagogical Agency Framework (ML-EAPAF), which proposes the concepts of four layers of analysis: Macro (institutional context), Meso (instructional resource adoption), Micro (teacher agency and performance), and Integration (cross-dataset ecological fusion) to thoroughly investigate the performance and pedagogical agency of English major teachers in Chinese higher education. This study uses two complementary datasets: AI Teaching Enhancement Dataset (n=1,000) and AI-Enhanced English Teaching Resource Dataset (n=300), along with machine learning models, statistical hypothesis testing, and innovative methods for constructing composite indexes. Four new measures are presented: Teacher Agency Score (TAS), Pedagogical Effectiveness Index (PEI), Ecological Interaction Index (EII), and Ecological Integration Score (EIS). Results showed very good predictive validity; the regression model of Linear Regression with R²=1.0000 and classification models of XGBoost with accuracy of 95.50% and AUC=0.9947. The EIS is found to have a strong positive correlation with overall teaching performance (r=0.729, p<0.001), thus validating the proposed framework ecologically. Four of the six ecological hypotheses are supported, with the use of AI tools (r=0.279, p<0.001), teacher agency (r=0.325, p<0.001), and years of experience (ρ=0.691, p<0.001) being key teachers' characteristics that predict teaching effectiveness. The results provide theoretical and practical implications for the design of curriculum with Artificial Intelligence, teacher training and educational policies in China's higher education.