AEyeCoL: Advanced Understanding of Eye Gaze in Co-Located Collaborative Learning

• Applicants: Dr. Babette Bühler; Dr. TIm Fütterer; Professorin Dr. Enkelejda Kasneci; Professor Dr. Ulrich Trautwein
• Subject Area: Image and Language Processing, Computer Graphics and Visualisation, Human Computer Interaction, Ubiquitous and Wearable Computing ; General and Domain-Specific Teaching and Learning
• Term: since 2025
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 563097140

Collaborative learning (CL) is a key component of 21st-century education, recognized for its potential to foster critical thinking, problem-solving, and lifelong learning skills. Whereas extensive research has examined the effects of CL on learning outcomes, much less is known about the non-verbal mechanisms that underpin successful collaboration. This project aims to advance our understanding of gaze dynamics in co-located, small-group CL settings using mobile eye-tracking technology and machine learning (ML) techniques. By analyzing gaze interactions such as joint visual attention (JVA), JVA initialization, and eye contact, we aim to uncover the social and cognitive processes that drive effective collaboration. The proposed project has five main objectives: (O1) To assess a comprehensive set of group-level gaze indicators relevant to CL. (O2) To develop an automated gaze analysis pipeline using computer vision methods, replacing manual annotation of areas of interest to extract gaze indicators and publish it as an open-source resource. (O3) To publish an anonymized large-scale eye-tracking dataset for future collaboration research. (O4) To predict team dynamics, such as leadership emergence and turn-taking, using ML models based on gaze indicators. (O5) To transfer and adapt our methods to real-world educational settings with school-aged children, shedding light on the interplay between gaze dynamics, group processes, and learning outcomes. Through two cascading CL studies, the first one investigating university students, and the second one elementary school children, we will address critical gaps in the field, including the lack of comprehensive gaze indicators beyond JVA, the scarcity of tools for scalable analysis, and the limited focus on naturalistic and multi-user settings in existing research. By investigating how gaze behaviors reflect team dynamics and their relationship with instructional quality and learning outcomes, this project will offer new insights into the processes underlying effective collaboration. The results will lead to the publication of a rich, anonymized mobile eye-tracking dataset, an open-source framework for gaze analysis, and actionable design guidelines for enhancing CL in both higher education and school contexts. These contributions will not only advance research on gaze dynamics but also inform the design of AI-driven tools and instructional methods to improve collaboration and learning outcomes in diverse educational and professional settings. Furthermore, our interdisciplinary approach, integrating research on education and computer science, aligns with UGaze by advancing the understanding of gaze dynamics in collaborative interactions, bridging Gaze Sharing and Gaze-Based Interaction through the development of scalable tools and an open-source dataset, fostering community-driven research and enabling applications across educational and diverse multi-user contexts.

• DFG Programme: Priority Programmes
• Subproject of: SPP 2481: Understanding Gaze
• Co-Investigator: Professor Dr. Peter Gerjets