INTEGRATING ADAPTIVE AI AGENTS INTO HIGHER EDUCATION LANGUAGE LEARNING

Abstract

This study examines the integration of Adaptive AI Conversational Agents (AACAs) in higher education language learning, focusing on their role in fostering critical thinking, intercultural competence, and academic writing. Learner motivation was analyzed as a mediating factor, while cognitive load served as a moderating variable. Grounded in Sociocultural Theory, Self-Determination Theory, and Cognitive Load Theory, a quasi-experimental design was conducted with 118 university students over ten weeks. The experimental group engaged in scaffolded AI-mediated dialogues and writing tasks adapted to individual proficiency, while the control group received conventional instruction. Data were analyzed using ANCOVA, structural equation modeling, and moderation analysis. Findings showed significant improvements across all three domains (p < .001), with motivation fully mediating writing gains and partially mediating critical thinking gains. Cognitive load moderated outcomes, with the best results at low-to-moderate levels of extraneous load. Qualitative insights highlighted perceived benefits such as cultural authenticity, scaffolded challenge, and enhanced writing confidence. The study underscores the potential of AACAs as both cognitive and motivational tools, offering practical implications for the sustainable integration of AI in higher education language pedagogy.