AI-Assisted Academic Writing in Arabic Education: Causal Evidence and Ethical Implications from MAN 4 Hulu Sungai Tengah

Abstract

This multi site study investigates both the effectiveness and ethical implications of AI-assisted academic writing in Arabic Language Education at MAN 4 Hulu Sungai Tengah. Employing a mixed-methods design, the study involved 90 students divided into two groups: an AI-assisted writing group and a conventional learning group. Quantitative findings indicate a statistically significant improvement in writing performance among students in the AI-assisted group, along with a notable difference in overall achievement compared to the conventional group. The large effect size suggests that properly guided AI support can accelerate drafting and revision processes while substantially enhancing writing quality. Qualitative results further highlight stage specific patterns of AI utilization. Generative AI tools were primarily employed during the early stages of writing, such as idea generation, outlining, overcoming writer’s block, and improving academic tone. In contrast, editing tools were more frequently used for grammatical correction and stylistic refinement. From an ethical perspective, the findings reveal uneven risk distribution. Lower-performing students demonstrated a higher tendency toward overreliance on AI, often accompanied by a diminished sense of authorial voice. Conversely, higher-performing students were more likely to retain control over their writing and exhibited stronger awareness of academic integrity. Additionally, limitations of AI in handling culturally and disciplinarily specific terminology within Arabic academic discourse emphasize the continued necessity of instructor feedback to ensure conceptual accuracy and rhetorical appropriateness. Overall, this study proposes an effective and ethically grounded hybrid instructional model, positioning AI as a targeted scaffolding tool while integrating transparency standards and context-sensitive pedagogical safeguards.