SHIH, YU-TING
E-mail: cms112103@gm.ntcu.edu.tw
LI, CHENG-HSUAN
E-mail: chli@mail.ntcu.edu.tw
Abstract
The rapid advancement of artificial intelligence (AI) has positioned data science and programming as essential competencies in higher education. However, students without a computer science background often struggle with the abstract reasoning required for programming. This study explores how 103 undergraduates interacted with an AI-based virtual learning partner during a general education course, analyzing 710 dialogue records through lag sequential analysis.
Results identified seven behavior types and revealed that task submission (TS) was the most frequent, reflecting a highly task-oriented learning style. A key finding was the bidirectional transition between information seeking (IS) and cognitive/metacognitive interaction (CM), forming a “learning exploration loop” indicative of reflective and integrative learning strategies. Transitions from IS to task-oriented requests (TO) further showed students applying knowledge to concrete tasks. Social engagement (SE) patterns, including greetings and thanks, revealed that students treated the AI as an interactive partner.
This study provides empirical insights into student behavior in AI-supported learning environments and suggests directions for enhancing interaction design and instructional strategies in programming education.
Keywords: Lag Sequential Analysis, Generative AI, Programming Education, Learning Analytics, Human-Computer Interaction