In this thesis we study and identify processes of socially shared metacognitive regulation within collaborative group work, with the intent of attaining knowledge of how such processes can be measured and quantified, and in turn be utilized with learning analytics to afford better collaboration within the group. The study started with a broad focus on understanding the practice of collaborating and was, through an extensive literature, narrowed down to focusing on socially shared regulation of learning, and from then further down do socially shared metacognitive regulation which is part of the former.
To study the processes to be quantified, we have coded, transcribed and analysed four hours of video recordings of an international group of five students working on their semester project under the educational approach of problem-based learning. In order to analyse the video data, we utilised different methods of inductive analysis and selection of the materiel. To narrow down a large corpus of video data, macro-level coding was employed to assess which parts of the recordings were relevant to us. From the coded clips we then picked three that we found to be representative of different regulatory processes. Two of the chosen clips were transcribed textually with a focus on two separate modalities of interaction, namely bodily and verbal interaction. This transcription was carried out using parts of the Jeffersonian notation system, which was appropriated to fit our research goal. Lastly, a small planning sequence was transcribed using a graphical transcript in order to properly represent the visual aspects of interaction.
During the analysis of the recording which was conducted, it became clear that the identified regulatory processes could be looked at from different levels of abstraction. At a macro-level we were able to observe the overall processes of socially shared metacognitive regulation, planning, monitoring and evaluation. At a meso-level we identified further mechanisms within planning, such as negotiating the purpose and nature of tasks. When looking at a micro-level, we further discussed the meso-level mechanisms through different modalities of interaction: verbal and nonverbal language, eye gaze and bodily gestures.
This analysis allowed us to infer that insights gathered on different levels of metacognitive regulation of learning can be used for assessment of the quality of collaboration. We further discuss how our observations related to modalities of interactions could be used for creation of indicators of successful socially shared metacognitive regulation (SSMR). The main contribution of this thesis comes with the reflections related to the application of multimodal learning analytics to measure aspects of SSMR. Our work describes opportunities and challenges that come both from using data from single modalities and from combining multimodal datasets.