Extended Reality in Collaborative Learning Applications: Evaluating the Interplay Between Attention, Engagement, and Learning

Jury

• Mr.David PANZOLI, Professor, IRIT, team REVA, Toulouse, Rapporteur
• Mr. Pablo FIGUEROA, Professor, Universidad de los Andes, Colombia, Rapporteur
• Ms. Anastasia BEZERIANOS, Professeure, Université Paris-Saclay, LISN, INRIA, Examiner
• Ms. Stéphanie FLECK,  Professeure, Université de Lorraine – PERSEUS, Examiner
• Mr. Julien NELSON, Université Paris-Cité, Centre Borelli, Examiner

Keywords

Extended Reality (XR), Virtual Reality (VRVirtual Reality), collaborative learning, user engagement, attention, synchronous learning, distance learning.

Abstract

The integration of Extended Reality (XR) technologies into educational settings promises to reshape remote education by offering immersive, spatialized interactions. However, this transition introduces a critical challenge: the “Attentional Paradox.” While synchronous collaboration in Virtual Reality (VRVirtual Reality) heightens learner motivation and the sense of co-presence, it simultaneously shatters visual attention and introduces severe social cognitive load stemming from spatial grounding and coordination overhead. Furthermore, the current literature suffers from two significant gaps: a heavy over-representation of single-user scenarios (the “Isolation Paradigm” at 60.6%) over synchronous collaboration, and an over-reliance on subjective, post-hoc questionnaires that fail to capture real-time, dynamic cognitive strain.
This thesis addresses these limitations through a two-part, telescopic empirical program using a complex, spatially demanding molecular biology educational testbed. First, we establish a continuous, multi-modal measurement architecture—logging eye-reticle dwell time, voice activity duration, and interaction frequencies—to objectively track real-time behavioral data. Experiment 1 (Diagnosis) quantifies the cognitive cost of collaboration by comparing individual and baseline collaborative learning conditions, validating that a shared virtual environment imposes a measurable coordination tax despite high subjective engagement. Experiment 2 (Optimization) evaluates the implementation of technological attention-guiding mechanisms (gaze-contingent scaling and visual highlights). The results demonstrate a profound pedagogical success, revealing a massive, statistically significant increase in learning gains for the guided condition (57.55% vs. 38.75% in the baseline, d = 0.86), alongside an 11.68% reduction in task completion times and decreased verbal coordination overhead. Crucially, however, the findings expose a critical functional trade-off: cognitive offloading via system automation inadvertently lowered learners’ subjective sense of social co-presence.
Ultimately, this thesis formalizes the “Efficiency-Sociality Paradox,” offering a fundamental theoretical contribution to the fields of Computer-Supported Collaborative Learning (CSCL) and Human-Computer Interaction (HCI). It proposes the adaptive Collaborative XR Attention Scaffolding (CXAS) model and outlines three actionable design guidelines for developing future educational XR platforms that optimally balance task efficiency with human connection.