From Bots to Mentors: Unveiling Dialogic Patterns in AI vs. Human Teacher Writing Instruction

• In Event: Conversing with AI: Fostering Student Inquiry and Socratic Dialogue

Abstract

With the rapid integration of generative artificial intelligence into writing instruction, understanding its effectiveness compared to human teachers in dialogic feedback is essential. The study involved 66 Chinese undergraduates, divided into AI (n=35) and teacher (n=31) groups, with 2,018 dialogues from an authentic writing task. Using an extended Initiation–Response–Follow-up coding scheme and Ordered Network Analysis, we examined dialogic patterns of two groups. Results indicate AI-student dialogues are efficient but narrowly focused, dominated by student prompts and AI direct suggestions with limited scaffolding. Teacher-student dialogues exhibit richer dialogue acts and more guided support, such as hints and reformulations. This study highlights the potential of combining AI’s efficiency with teachers’ pedagogical depth through hybrid models to enhance student thinking and deepen learning.

Authors

• Siyuan Li, East China Normal University

• Jing Leng, East China Normal University