The role of epistemic beliefs in predicting deep learning strategies in an AI-assisted English approach

Abstract The rapid integration of AI chatbots into English language education raises critical questions about their influence on learners’ epistemic beliefs, conceptions of learning, and strategic approaches to knowledge acquisition. While prior research has examined AI’s role in language skill development, little is known about how chatbot-mediated learning environments reshape foundational cognitive constructs. This mixed-methods study addresses this gap by investigating 425 university students’ interactions with AI chatbots. Quantitative results demonstrated that epistemic uncertainty regarding English knowledge negatively predicted both application-oriented and testing-oriented conceptions of learning, indicating that AI tools may exacerbate skepticism rather than foster knowledge internalization. Dominant exam-oriented paradigms further amplified students’ reliance on testing as a proxy for competency, reflecting systemic assessment pressures on epistemic development. Qualitative analysis revealed that chatbot features promoting immediate feedback and authority validation reinforced transactional learning strategies, while limited guidance for critical evaluation increased cognitive load. Thematic findings underscored students’ tendency to conflate chatbot efficiency with epistemic reliability, prioritizing expediency over metacognitive engagement. These results highlight the dual role of AI chatbots as both scaffolds for personalized learning and potential inhibitors of deep cognitive processing. The study advocates for pedagogically structured AI integration frameworks that align tool functionality with explicit instruction in epistemic self-regulation. We argue that teacher-mediated critical interrogation of chatbot outputs is essential to balance cognitive efficiency with intellectual autonomy in AI-enhanced language education.