| Popis: |
Procrastination and other types of dilatory behaviour are common in online learning, especially in higher education. While procrastination is associated with worse performance and discomfort, positive forms of delay can be used as a deliberate strategy without any such consequences. Although dilatory behaviour has received attention in research, it has to my knowledge never been included as an integral part of an adaptive learning system. Differentiating between different types of delay within such a system would allow for tailored interventions to be provided in the future without alienating students who use delay as a successful strategy. In this thesis, I present four studies that provide the basis for such an endeavour. I first discuss the results of two studies that focussed on the prediction of the extent of dilatory behaviour in online assignments. The results of both studies revealed an advantage of objective predictors based on log data over subjective variables based on questionnaires. The predictive performance slightly improved when both sets of predictors were combined. In one of these studies, we implemented Bayesian multilevel models while the other aimed at comparing various machine learning algorithms to determine the best candidates for a future inclusion in real-time predictive models. The results reveal that the most suitable algorithm depended on the type of predictor, implying that multiple models should be implemented in the field, rather than selecting just one. I then present a framework for an adaptive learning system based on the other two studies, where I highlight how dilatory behaviour can be incorporated into such a system, in light of the previously discussed results. I conclude this thesis by providing an outlook into the necessary next steps before an adaptive learning system focussing on delay can be established. |
