What matters in AI-supported learning: A study of human-AI interactions in language learning using cluster analysis and epistemic network analysis

• Published in: Computers and education Vol. 194; p. 104703
• Main Authors: Wang, Xinghua, Liu, Qian, Pang, Hui, Tan, Seng Chee, Lei, Jun, Wallace, Matthew P., Li, Linlin
• Format: Journal Article
• Language: English
• Published: Elsevier Ltd 01.03.2023
• Subjects: Applications in subject areas; Elementary education; Human-computer interface; Learning communities; Learning communities; Human-computer interface; Elementary education; Applications in subject areas
• ISSN: 0360-1315; 1873-782X
• DOI: 10.1016/j.compedu.2022.104703

This study investigates how students interact with artificial intelligence (AI) for English as a Foreign Language (EFL) learning and what matters in AI-supported EFL learning. It was conducted in naturalistic learning settings, involving sixteen primary school students and lasting approximately three months. The students' usage data of an AI agent and their reflection essays about the interactions with the AI agent were analyzed using cluster analysis and epistemic network analysis based on the frameworks of community of inquiry and students' approaches to learning. The results suggest four clusters of students, each with its distinct way of interacting with AI for language learning. More importantly, the comparisons of the four clusters of students reveal that even in AI-supported learning, not everyone can benefit from the potential promised by AI. The deep approach to AI-supported learning may amplify the benefits of AI's personalized guidance and strengthen the sense of the human-AI learning community. Passively or mechanically following AI's instruction, albeit with high levels of participation, may decrease the sense of the human-AI learning community and eventually lead to low performance. This study contributes to and has implications for the educational implementation of AI, as well as the facilitation and graphical representation of learner-AI interactions in educational settings. •Community of inquiry and students' approaches to learning are used to analyze human-AI interactions.•Four clusters of students are identified with distinct ways of interacting with AI.•Not everyone can benefit from the potential promised by AI.•The deep learning approach may enhance the human-AI learning community, leading to high performance.•Passively or mechanically following AI's instruction may weaken the human-AI learning community.