BEYOND SURFACE ERRORS: AN ARTIFICIAL INTELLIGENCE-ASSISTED SYSTEMIC FUNCTIONAL LINGUISTICS APPROACH TO ESL WRITING ASSESSMENT
This study examined the use of artificial intelligence (AI) in facilitating the examination of student writing errors through the Systemic Functional Linguistics (SFL) framework. The objective was to create an AI-assisted assessment tool proficient in detecting faults associated with ideational, interpersonal, and linguistic metafunctions in writing. This method was compared to conventional error analysis techniques, emphasizing grammar, vocabulary, and punctuation. A mixed-methods strategy was utilized. A qualitative study investigated the categories and origins of errors, whereas a quantitative analysis documented the prevalence and distribution of these errors. The dataset consisted of 50 argumentative essays written by Pakistani university students studying English as a second language. Each essay varied in length from 200 to 400 words. Errors were initially detected utilizing OpenAI tools (GPT-4o mini model) and subsequently validated manually by SFL expert for precision. A total of 475 errors were detected: 194 (41%) interpersonal errors, 240 (51%) textual errors, and 41 (8%) ideational errors. The findings indicate that SFL facilitates a profound comprehension of writing challenges, transcending superficial faults to uncover students' conceptualization and organization of their ideas. The SFL methodology revealed more profound insights into student cognition and expression than traditional methodologies. This work emphasizes AI's capacity to integrate conventional and functional methods in error analysis. It proposes practical applications for instructors seeking to provide more substantive, formative feedback to enhance students' writing abilities.