AI as a debate coach: A mixed-methods analysis of student self-efficacy and perceptions in an AI-assisted debate

Debate is an effective pedagogical approach, yet it presents significant challenges for EFL learners. While Artificial Intelligence (AI) chatbots are increasingly integrated into education, their application in multi-skilled tasks like debate preparation remains underexplored. This study investigated the impact of a two-phase AI-assisted intervention on 48 EFL learners' debating self-efficacy and perceptions. Phase 1 involved traditional debate preparation without AI assistance, focusing on foundational skill development through instructor-led instruction. Phase 2 introduced AI chatbots for refinement of arguments, rebuttals, and delivery practice, allowing students to enhance their debates through AI-powered scaffolding. Data were collected via self-efficacy questionnaires at three time points (pre-intervention, post-Phase 1, and post-Phase 2), a post-intervention perceptions questionnaire, written reflections, and focus group interviews. Repeated-measures ANOVAs revealed significant stepwise increases in students’ debating self-efficacy across the three time points, with the most substantial gains observed in debate skills and language use. The perceptions questionnaire corroborated these findings, demonstrating that students rated AI as most effective for refining speeches, locating evidence, and developing arguments, while perceiving it as least helpful for oral delivery practice. Furthermore, qualitative analysis yielded nuanced and contextualized insights regarding both the benefits and limitations of AI-assisted debate preparation.