Automated Gaze Orientation Estimation Behavioural Research: A Feasible Alternative to Lab-Based Eye-Tracking for More Accessible Remote Studies

The use of remote data collection has the potential to broaden participant accessibility,particularly for populations that face barriers to traditional lab-based studies. Recent efforts toreplicate eye-tracking research remotely using conventional webcams have been successful.However, highly involved calibration periods and issues with data quality due to headmovements risk compounding the same challenges encountered in lab-based studies for thesepopulations. This paper explores the efficacy of appearance-based gaze estimation (ABGE)techniques to automatically estimate gaze orientation without calibration, in contrast toprecision eye-tracking methods. Across two studies, this technique demonstrated high inter-rater reliability between human and automated scoring in both adults and children, withautomated scoring outperforming manual scoring in terms of precision. The second studyapplies the automated gaze estimation techniques to data collected during an experimentaltask and demonstrates comparable results to research in the current literature, confirming itsability to accurately estimate gaze orientation in a digit comparison task involving children. Theresults highlight ABGE’s potential to replicate gaze orientation estimates without the need forextensive calibration, which is typically required by lab-based eye-trackers. While challengessuch as video data security and computational resources remain, the findings support thebroader use of ABGE in data collected with conventional webcams. By increasing accessibilityfor participants and researchers alike, ABGE offers a promising alternative to traditional eye-tracking methods, ultimately advancing the field's understanding of relationships between eye-gaze and cognition.