EDUBUDDY: A WEB-BASED AI TOOL FOR DETECTING LEARNING STYLES AND RECOMMENDING PERSONALIZED STUDY STRATEGIES

Abstract

The ongoing challenge of meeting diverse learning preferences within standardized education highlights the need for more personalized learning support. Although learning styles—preferences for perceiving, processing, and retaining information—are often linked to engagement and academic outcomes, common assessment approaches rely heavily on self-report tools (e.g., VARK, Felder-Silverman ILS) that face major concerns: inconsistent reliability and validity, static profiles that overlook context shifts, and limited usefulness in producing practical study guidance. To address this gap, this study proposes, develops, and evaluates a Web-Based AI Tool for dynamic learning-style detection and personalized study-strategy recommendation. Rather than depending on questionnaires, the tool continuously and non- intrusively infers learning preferences using machine learning on authentic behavioral data within a web learning platform. Indicators include content format choices (text, video, simulations), time-on-task patterns, navigation paths, interactions with learning elements, and performance differences across assessment types. These data generate a probabilistic, continuously updated learning style profile based on observable behavior, reducing self- report bias and accommodating context-dependent changes. A recommendation engine then translates the AI-inferred profile into concrete, actionable study strategies delivered through the platform or a user dashboard. For example, visually oriented learners may receive guidance on concept mapping and color- coded summaries, while active learners may be encouraged to engage in peer discussion and to practice frequently. The system integrates user feedback (explicit ratings or implicit adoption signals) to refine both detection accuracy and recommendation relevance over time. Using a design-based research methodology, the tool is iteratively designed and tested in authentic settings. Mixed-methods evaluation examines detection validity, user acceptance, effects on metacognitive awareness and study habits, and initial links to learning outcomes, alongside ethical safeguards for privacy and bias. Preliminary results suggest the tool is feasible, well-received, and supports more intentional study practices, offering a practical bridge between identifying preferences and improving learning strategies.