Lee, Alwyn Vwen Yen

This paper, as part of a larger ongoing study, presents the use of a multi-layer approach to analyzing teacher noticing for the improvement of teaching practices. Situated in the field of teaching analytics, the use of multimodal sensors and analytics, especially for teacher noticing research, has provided affordances to discover deep insights for improving teaching practices. We collected data from a case study of one teacher over three lessons of science teaching in a secondary school. Multimodal sensors including an eye-tracking device, a microphone, and multiple video cameras were deployed in a classroom. The various sources of data were integrated and a multi-layer analysis was performed to uncover insights into the teaching practice. The findings show that a novice teacher in our case study was able to attend to events in her classroom, with some interpretations and sense-making of the events; some necessary actions were taken based on the teacher’s analysis but in some instances, necessary action was found to be lacking. Prior knowledge and the wealth of experiences or the lack thereof, together with visual cues in the environment, can affect the decision of novice teachers in executing certain actions in a classroom.

Instructors adopting a dialogic approach to teaching and learning (Reznitskaya & Gregory, 2013) tend to design learning environments where there is shared control over classroom talks for collaborative meaning making and dialogic inquiry among learners. In this thesis, the dialogic pedagogical approach follows Scardamalia and Bereiter’s model of knowledge building (2003), which leverages a learner’s natural curiosity in questioning and inquiry. In such an environment, learners engage in social collaborative inquiries to contribute and advance communal knowledge. However, working to improve one’s ideas requires considerable support (Scardamalia & Bereiter, 2014). For example, during the initial phase of the inquiry, learners could raise many competing ideas. Recognising and identifying promising ideas is an approach to support such processes and is also an important component of expertise and creativity (Bereiter & Scardamalia, 1993; Gardner 1994). Technological tools, such as online forums (e.g., Knowledge Forum), can provide conducive platforms to the sharing of ideas, but learners still encounter potential difficulties in identifying promising ideas that would be relevant, communally interesting, and of impact to the community. The goal of this research is to identify and analyse promising ideas that can sustain idea improvement in knowledge building discourse, and investigate the effect of these ideas on the collective advancement of communal knowledge. This thesis encapsulates the development and implementation of a methodology called Idea Identification and Analysis (I²A), which uses network analysis, temporal analytics, and machine learning techniques to define the attributes of promising ideas, recognise various idea types, and determine the effect of such ideas within online knowledge building discourse. The methodology is guided by an Idea Pipeline framework that is modelled after a classic innovation process. Two studies were conducted in this thesis to search for promising ideas and to determine the feasibility and scalability of the I²A methodology. The first study was exploratory in nature and its results determined the attributes of idea promisingness using the betweenness centrality (BC) network measure. A classification system was developed to determine idea types in discourse based on the recognition of BC trends. A temporal analysis was applied to identify promising ideas in discourse and the findings were qualitatively validated. Building upon the first study, the second study further improved the I²A methodology using text mining, new visualisation, and clustering techniques. In addition to confirming the presence and nature of promising ideas from the first study, learners’ input in discourse were examined to determine the mobility of ideas in discourse and various idea types could be automatically recognised. These processes also helped to confirm and uncover new promising ideas in discourse. In essence, this thesis contributes to the field of Computer-Supported Collaborative Learning (CSCL) with the development of a framework and methodology that will guide and assist future research and development in closing the knowledge gap in idea analysis within knowledge building discourse.

The widespread adoption of personal computers and mobile devices has enabled learning analytics to become more pervasive among teachers, school administrators, students and parents. While the past decade has marked notable advancements in learning analytics, less attention has been paid to the unique characteristics of learning analytics that necessitate the notion of learning analytics literacy. Although researchers have documented the common use and misuse of learning analytics in education, there is still limited research that highlights the importance of cultivating literacy around the use of learning analytics for better understanding of teaching and learning practices. This paper describes how a Micro-Meso-Macro (3M) analytical approach can be used to support and enhance learning analytics literacy among education stakeholders, while raising the prospects of how a systematic implementation of raising learning analytics literacy can be done through two interacting themes: raising awareness and raising criticality.

This paper reports an empirical study on the use of a teacher noticing approach to investigate how two teachers managed students’ classroom behaviours. We examined the integration of data from an eye-tracking device and video cameras, focusing on what the teachers paid attention to in classrooms with their corresponding managing practices. Our findings show that the experienced teacher was able to advise her students calmly and smoothly resume the lesson to preserve the welcoming environment for the students. The novice teachers constantly scanned for misbehaved students and at times used strong words and a stern voice that betrayed her emotions. The awkward silence of the class ensued, suggesting a break in the flow of the instruction.

Grasping students' emotions, especially those relating to learning, in a collaborative setting is no easy feat for teachers. The quality of collaboration comprises both visible behavior and emotion and the less visible emotional traits relating to engagement and motivation. Teachers often rely on their experience and intuition when it comes to these invisible traits. In this study, we collected multimodal data from a collaborative knowledge building classroom to analyze when and how students' emotions transpire during the working and improvement of ideas. Data included textual data, self-reports from surveys, interviews, and physiological data from face-to-face and online knowledge building discourse of 17 students in a 2.5-hour Social Studies lesson. We found shifts in epistemic emotions during idea improvement activities, and the students explained these shifts in understanding the discussion and engaging in idea-centric processes. We discuss findings for ongoing work to develop multimodal analytics for knowledge building practice.

Unprecedented global issues have caused large-scale disruption to preschools, requiring a rethink of existing teaching and learning approaches. This study examines the inception of a knowledge building approach in preschools using a case study method and a micro‑meso-macro (3M) analytical framework to understand impact and implications. We found that teachers and students, as micro units (individual agents) collaborated with each other at the meso level, where knowledge creation is carried out within communities, while interacting with macro units (school leaders). Analysis shows the knowledge building approach aided stakeholders in navigating organizational, technological, and pedagogical challenges. Elevating awareness and acknowledging impacts of the knowledge building approach can inspire critical discourse to understand and handle future disruptions in the educational landscape.

The assessment and understanding of students’ ideas in discourse have often been a difficult problem for teachers to tackle. Recent innovations and technologies such as text mining can provide a partial solution by generating an estimated count of important keywords which are representative of ideas within discourse. However, investigating idea development and flow within discourse is a much more challenging task, and requires elaborate processing and analysis. In this study, a method for analysing idea flow was proposed and tested: (a) text mining and network analysis are employed to identify and validate ideas from textual discourse; (b) identified ideas are grouped together and mapped to relevant learning objectives; (c) groups of ideas are then aggregated using a binning method and scored; (d) a flow diagram is generated using the aggregated scores to visualize idea flow within discourse. By understanding how ideas flow within discourse, discussed key ideas can be monitored and any lapses in student understanding can be identified so that teachers will have information to provide timely interventions to support and scaffold learning.

Existing research has substantial progress in uncovering outcomes of collaborative learning in recent years, but more attention can be directed towards the better understanding of collaborative learning processes via quantitative frameworks and methods. Through the use of knowledge building as a collaborative learning pedagogical approach, it is possible for researchers to glean deeper insights into aspects of students’ collaboration within authentic learning environments. In this paper, the multimodal approach of data collection and analysis was conducted with a proposed conceptual analytical framework that can characterize constructs of collaborative activities in a knowledge building classroom using machine learning methods. The application in a pilot is discussed along with how this conceptual development can offer a summary of new insights into students’ individual and group collaborative trajectories during learning tasks.