Enhancing Arabic writing skills using Chat GPT-based AI learning models: A tridimensional human-AI collaboration framework
Ahmad Zubaidi, Abdul Munip, Sembodo Ardi Widodo, Taha El-Amine Zerrouki
Abstract
Writing Arabic, particularly at the Insya’ Ric stage, is difficult due to its grammar complexity. This study employed artificial-intelligence technology, namely ChatGPT, via customized prompts to enhance and assist in Arabic writing instruction. The research and development model used was Borg and Gall (1983), and the limited trials of this study were conducted with 40 students from the Islamic Religious Education program at an Islamic university in Indonesia. Results showed 89.6% of experts rated the AI tool as excellent, 88% of students exceeded the minimum learning standard, and writing proficiency increased by 12.5%. However, when tasked with generating Indonesian words alongside their corresponding Arabic translations, the AI exhibited significant errors, producing inaccurate translations and frequently repeating the same words or phrases instead of generating distinct counterparts in each language. To counter these problems, the research used the Tridimensional Human-AI Collaboration model, with AI as a tool and humans as conceptors, validators, and executors in writing. Tools like sarf.one integrated with ar-radif were utilized for better verb morphology and noun phrase accuracy, thus improving overall lexical accuracy. The findings suggest that AI can be helpful to create structured content, but human input is necessary to achieve correct grammar and context. This research contributes to the growing field of AI-assisted language learning, particularly in Arabic writing, by offering an advanced framework for collaborative writing education.