SPEAKING WITH MACHINES, LEARNING WITH PEOPLE: AI-MEDIATED INTERACTIVE ACTIVITIES IN ARAB EFL CLASSROOMS

Abstract

Artificial intelligence (AI) has entered English-language education as both promise and provocation, offering individualized feedback while unsettling established models of classroom interaction. This qualitative study investigates how AI-mediated interactive activities shape speaking performance, motivation, and confidence among Arab learners of English as a Foreign Language (EFL) in Israel. Across three junior-high schools over eight weeks, we combined non-participant classroom observations (30 lessons) with semi-structured interviews of teachers (n = 6) and students (n = 90). Thematic analysis yielded three core findings. First, AI chatbots, pronunciation coaches, and gamified simulations stimulated high engagement and lowered anxiety, increasing time on task and participation. Second, iterative practice with immediate, low-stakes feedback fostered willingness to communicate, more self-initiated turns, and evidence of spontaneous self-repair. Third, persistent constraints—accent-recognition bias, connectivity issues, opaque feedback, and uneven teacher preparedness—limited consistent impact without targeted support. Framed by Krashen’s affective filter and Vygotsky’s sociocultural theory, the study argues that AI’s value lies in human-mediated orchestration: teachers contextualize and humanize machine feedback, translating scores into actionable, discourse-level guidance. We propose practical design principles (collaborative task framing, reflective debriefs, bias-aware “tool talk”) and a capacity-building agenda (infrastructure, PD, assessment alignment) to guide sustainable, equitable adoption in linguistically diverse EFL contexts.