A Data-Driven Approach to Exploring Learner Input Preferences in E-Learning

Abstract

Learners exhibit varied learning styles, influenced by their prior knowledge, experiences, and personal characteristics. This study focuses on the input dimension (visual/verbal) of the Felder-Silverman Learning Style Model (FSLSM) and proposes an automatic detection method based on interaction traces from e-learning platforms such as Moodle. The data, collected from several universities, includes hundreds of learners and tracks the time spent watching videos, reading text documents, viewing images/diagrams, and participating in forums. An unsupervised clustering algorithm was used to group learners according to their visual or verbal preferences, without resorting to self-reported questionnaires. In order to validate the identified profiles, a supplementary questionnaire was administered to analyze learners' behaviors on social networks (preferred content formats, modes of communication, frequency of use). This external data confirmed a consistency between inferred learning styles and personal digital practices. The results reveal that the majority of learners exhibit a preference for visual content (videos, images), while a subgroup shows an affinity for textual materials and verbal interactions. These findings confirm the relevance of adaptive learning systems for personalizing resources according to individual preferences. This research demonstrates the effectiveness of analyzing digital traces and social uses for the automatic identification of learning styles, thereby helping to improve the engagement and success of online learners.