Answers are grounded in this paper and cite their sources.
Webcam-based eye tracking could become an infrastructure for learning research: it runs in the browser, scales to large online samples and could supply learning analytics and adaptive learning systems with gaze data. This requires accurate gaze location on small areas of interest (AOIs), such as the parts of a graph. Seventy-seven participants solved six graph items twice, with a commercial webcam-based system (Labvanced) and an infrared eye tracker (Tobii Pro Spectrum); the webcam session took place in the laboratory (n = 45) or at home (n = 32). Because the graphs were identical, AOIs within the graph were defined identically for both systems. Webcam data had low data loss but were 2.5 times less precise. At the group level, they reproduced the main pattern: gaze concentrated on the function lines more than on a control region (enrichment 1.45 vs. 1.05), and group gaze maps correlated with infrared maps nearly as highly as their reliability allowed. The effect was, however, much weaker than in infrared data (lines 2.58 vs. control 1.20), and individual differences did not agree between systems (r ≤ .23), although they were reliable within each system. In simulations, person-specific offsets of about 3.7°, not noise, explained these limitations. Home and laboratory recordings did not differ detectably. For educational technology research, webcam-based eye tracking is ready for group-level questions about how learners attend to representations, but not yet for diagnosing individual learners, predicting their responses or driving gaze-adaptive systems, which require accuracy to be validated for each learner.
No takes yet. Share an insight, caveat, or question.
Kosel et al. (2026) studied this question.
Synapse has enriched 8 closely related papers on similar clinical questions. Consider them for comparative context: