SRLA: Self-Regulated Learning Analytics

Abstract

Learning analytics researchers have been diligently integrating trace data to study Self-Regulated Learning (SRL). Compared to traditionally used survey data, trace data, such as log or clickstream data designed and interpreted to understand a certain SRL construct, are considered to be more effective in capturing dynamic SRL as fine-grained events. Yet, trace data is not completely free from validity issues, since researchers’ understanding of contexts and target constructs heavily affect the validity of design and interpretation of trace data. Rather, researchers can adopt survey and trace data to complement each other. They can also compare survey and trace data on the same construct to deepen the understanding of what each type of data captures.

The aim of this dissertation is to understand the different nature of self-reported survey and trace data and to adopt context-specific indicators for particular SRL constructs. This dissertation is composed of three studies, each of which employed both survey and trace data to answer specific SRL-related questions. In Chapter 3, the first study demonstrates the importance of reflection in facilitating learning from hints through integrating complementary surveys and log data. The second study in Chapter 4 contrasts the alignment between achievement goal theory and trace data, and the misalignment between the theory and survey responses. The third study presented in Chapter 5 investigates causes of misalignments and builds theoretical interpretation from data.

This dissertation contributes to the field of learning analytics and SRL in multiple ways. First of all, the dissertation shows that effort in posing more valid methodological approaches could contribute to theories. In Chapters 4 and 5, I revealed the huge difference between learners’ goal statement before learning and goal-relevant behaviors during learning, and questioned some of the previous findings that heavily relied upon survey data. Furthermore, it emphasizes the importance of thoroughly understanding targeted constructs in specific contexts. Only then, can we select a strongly valid methodological approach.