作者
Afaq Ahmed Khan,Tehmina Zafar,Ubaid ur Rahman Farooqi,Muhammad Suleman Ilyas
摘要
This mixed-methods study examined the use of human-guided artificial intelligence (AI) support in an undergraduate English as a Foreign Language (EFL) course, Writing and Reading for Impact, at the University of South Asia, Lahore, Pakistan. The study focused on a bounded classroom model in which AI tools were permitted for brainstorming, outlining, grammar checking, reading-response support, and revision planning, but not for submitting unreviewed AI-generated text as original student work. Participants were 120 undergraduate students enrolled in three course sections taught by the same instructor. Quantitative data were collected through pre- and post-course writing assessments, rubric-based assignment scores, revision logs, and a structured questionnaire. Qualitative data were collected through reflective journals, focus-group interviews, and instructor field notes. Results showed a statistically significant improvement in overall writing scores from pre-test (M = 61.84, SD = 8.91) to post-test (M = 72.91, SD = 8.37), t(119) = 18.74, p < .001, Cohen's dz = 1.71. The strongest descriptive gains appeared in revision quality, language accuracy, and organization. Sixty-eight percent of students achieved moderate or high improvement across the semester. Survey and qualitative findings suggested that students valued AI most when it functioned as a scaffold for starting, revising, and noticing language problems, while concerns remained about overdependence, voice, and acceptable use. The study argues that AI integration in EFL writing is most educationally productive when it is explicitly taught, ethically bounded, and connected to process-oriented writing pedagogy rather than treated as either a prohibited shortcut or an autonomous substitute for learning.