How GenAI shapes STEM EFL students' willingness to communicate: an exploratory interview study using the T-CADS framework

Abstract

This study explores the mechanisms underlying changes in the Willingness to Communicate (WTC) of STEM undergraduates following a GenAI-assisted English as a Foreign Language (EFL) course. Drawing on semi-structured interviews with 13 first-year STEM students at a university in Northeast China, this study adopted a qualitative analytical framework based on the Topic Modeling-Corpus Assisted Discourse Studies (T-CADS) hybrid approach, integrating NLP-based topic modeling and corpus-assisted discourse analysis to analyze and visualize the interview data. The findings indicate that students' perceptions of GenAI were characterized by three interrelated dimensions: enhanced learning efficiency, low-anxiety interaction, and concerns about technological dependency. Changes in WTC were primarily reflected in greater confidence during classroom communication and increased transfer to authentic out-of-class interactions, although a small number of students demonstrated only context-dependent improvements. The findings further suggest that GenAI does not directly enhance students' WTC but instead facilitates changes through the combined effects of emotional safety, resource availability, and the reconfiguration of communicative contexts. These findings provide empirical support for shifting STEM English classrooms from technology-oriented implementation to communicative task design and suggest that English teachers should place equal emphasis on learner confidence, communicative context design, and the critical use of GenAI.