<![CDATA[Most classroom assessments of Indonesian-language proficiency still rely on a single outcome measure—whether the selected answer is correct or incorrect while largely overlooking how a learner's attention moves before arriving at that answer. This study aims to identify and compare eye-gaze scanpath patterns between high- and low-performing learners across three Indonesian-language task types: interpreting figurative language (majas), extracting the gist of a short reading passage, and applying spelling and punctuation conventions (Ejaan yang Disempurnakan/EYD). The goal is to reveal how cognitive load manifests differently across these task types. Unlike most prior eye-gaze research that treats cognitive load as a single, undifferentiated construct, this study tests the hypothesis that cognitive load varies systematically with the underlying language skill being exercised: figurative-language tasks emphasize semantic inference, reading comprehension emphasizes main-idea extraction, and orthography tasks emphasize rule-checking. To this end, gaze heatmap data were collected via real-time eye-movement recording through each participant's webcam while they completed the three tasks on an Intelligent Tutoring System (ITS) application. The article concludes by discussing the implications of this task-sensitive approach for designing an adaptive, AI-assisted Intelligent Tutoring System capable of responding to cognitive load in a task-specific manner.]]>
Copyrights © 2026
