To ISC or not to ISC? That is the question…When Do Shared Gaze Patterns Predict Learning? The Context-Dependent Boundaries of Intersubject Correlation

Abstract

As education shifts toward digital formats, pre-recorded lectures are becoming increasingly common. Unlike live instruction, online education lacks the real-time interaction and student cues instructors use to infer student engagement and understanding. Intersubject Correlation (ISC), measured through eye tracking, offers a promising approach by quantifying how closely learners’ gaze aligns with peers. Higher ISC has been linked to better learning outcomes, suggesting it may reflect effective attentional engagement. However, ISC implicitly assumes shared gaze on instructionally relevant content. When gaze is drawn to interesting but irrelevant elements, seductive details, ISC may instead reflect shared distraction. This study tested whether seductive details undermine ISC’s predictive value, as seductive details are known to hinder learning. Using webcam-based eye tracking, students watched a knowledge clip on lightning either in a Control version or with added seductive details. The Seductive Details condition showed significantly higher ISCs compared to the Control condition, but learning outcomes did not differ across groups. Notably, ISC positively predicted recall performance only in the Seductive Details condition. These findings suggest that ISC depends not only on the visual characteristics of instructional material, but also on the specific learning outcomes assessed.