Classroom collaboration analytics: designing and building automated systems for collaboration monitoring in classroom settings
This talk was given on 15 August 2024 to defend my PhD thesis on the topic of Classroom collaboration analytics: designing and building automated systems for collaboration monitoring in classroom settings at Tallinn University.
You can access the thesis here:đź“• Phd Thesis
Presentations slides are available here: Slides
Opponents
Supervisors
Abstract
Collaboration is a key skill in current education, also identified as one of the main “21st Century skills”. Moreover, the research evidence on the positive impact of collaboration on learning further underscores the need to develop this skill among students. Teaching students to collaborate effectively involves teachers monitoring each group’s activities and offering support to groups when needed. However, it is extremely difficult for teachers to be aware of how all their students collaborate, especially when multiple groups are working at the same time and their interaction is not only face-to-face but also computer-mediated. Therefore, it is essential to assist teachers in classrooms with monitoring and understanding of their students’ collaboration to effectively develop such skills among students.
Researchers have explored the use of a wide range of data sources such as audio, log data, video, etc. to understand and support learning and teaching. This field of research is known as Multimodal Learning Analytics (MMLA). MMLA researchers have combined learning traces from digital as well as physical spaces of interaction to holistically understand collaboration. The field has gained substantial traction among researchers from diverse disciplines, such as learning sciences, psychology, computer science, etc. This popularity can also be witnessed in an increasing number of research studies using MMLA with a focus on collaboration, learning and teaching. This area of research has identified data indicators for collaboration, developed tools to support teacher monitoring and understanding of collaboration processes. However, the majority of this research has been conducted in laboratory settings and there is still a lack of understanding of the feasibility of the automated estimation of collaboration measures in authentic settings.
In particular, this dissertation has identified and addressed three research gaps: first, a lack of systematisation in the evaluation of machine learning solutions developed in MMLA for collaboration; second, a lack of knowledge on building generalisable machine learning models for collaboration quality and its dimensions (e.g., argumentation) in classroom settings; and third, a lack of understanding of the generalisability of machine learning models for collaboration quality when used in authentic classroom settings.