Speaker
Description
Abstract
This mixed-methods study examined how 152 first-year non-English-major students in a Vietnamese university’s preparatory-English program chose between generative AI (GenAI) and human peers for eight English-learning tasks. Questionnaire responses yielded 1,216 task-level choices, which were analysed alongside partner-suitability ratings, multiple-response reasons, preferred workflows, 48 learning-diary entries, and interviews with 12 students. Overall, choices were nearly evenly divided between GenAI (48.3%) and peers (51.7%), but preferences varied by task. GenAI was favoured for form-focused practice and individual rehearsal, whereas peers were favoured for interactive communication and collaborative work. Students associated GenAI with immediate feedback, accessibility, and low-pressure rehearsal, while peers were valued for authentic interaction, negotiation, and shared responsibility. Many respondents preferred sequencing the two resources rather than relying on only one. The findings support a task-sensitive account of learning-partner choice, while not establishing comparative learning effectiveness.
Keywords: generative AI, human peers, learner agency, task-partner fit, EMI preparation