Showing papers by "Dirk T. Tempelaar published in 2019"
TL;DR: These findings took a step forward towards personalised and actionable feedback in learning analytics by recognizing the complexity of how and when students engage in learning activities over time, and supporting educators to design early and theoretically informed interventions based on learning dispositions.
39 citations
07 Nov 2019
TL;DR: In this paper, the authors focus on learning gains rather than learning outcomes as the most important performance dimension and find that high levels of engagement lead to lower, rather than higher, predicted learning outcomes.
Abstract: Learning analytic models are built upon traces students leave in technology-enhanced learning platforms as the digital footprints of their learning processes. Learning analytics uses these traces of learning engagement to predict performance and provide learning feedback to students and teachers when these predictions signal the risk of failing a course, or even dropping-out. But not all of these trace variables act as stable and reliable predictors of course performance. In previous research, the authors concluded that trace variables of product type, such as mastery, do a better job than trace variables of process type, such as clicks or time-on-task, in predicting performance. In this study, we extend this analysis by focusing on learning gains rather than learning outcomes as the most important performance dimension. Distinguishing two different levels of initial proficiency, our empirical analysis into the learning of mathematics by first-year university students indicates that the lack of stability of the engagement types of process type is mainly explained by learning pattern found in students of high initial proficiency. For these students, high levels of engagement lead to lower, rather than higher, predicted learning outcomes. Amongst students with lower initial proficiency, higher levels of engagement play a different role.
1 citations