Understanding the role of AI Chatbots in English self-regulated learning: A dual-stage structural equation modeling and artificial neural network approach

This study examines the constructs that predict Self-Regulated Learning (SRL) in English language learning through AI chatbots among low-achieving undergraduate students, focusing on the roles of social presence, trust in AI, enjoyment, and the Self-Determination Theory (SDT) constructs of perceived autonomy, relatedness, and competence. To achieve this aim, we employed a dual analytical approach that integrates Partial Least Squares Structural Equation Modeling (PLS-SEM) and Artificial Neural Networks (ANN) with Importance-Performance Map Analysis (IPMA) to analyze data from 332 valid questionnaires completed by university students in Taiwan. The results indicate that enjoyment, perceived autonomy, perceived relatedness, and trust in AI significantly predict SRL, with enjoyment emerging as the strongest predictor and perceived competence the weakest. Although social presence did not directly predict SRL, it was positively associated with autonomy, relatedness, and competence. These findings suggest that AI chatbots can support SRL among low-achieving students by providing enjoyable learning experiences that satisfy learners' psychological needs, while additional support is needed to strengthen their perceived competence. • Enjoyment was the strongest predictor of SRL in English learning. • Perceived autonomy, perceived relatedness, and trust in AI significantly predicted SRL. • Perceived competence was the weakest predictor of SRL. • Social presence predicted perceived autonomy, perceived relatedness, and perceived competence, but did not predict self-regulated learning.