Using Local and Global Self-evaluations to Predict Students’ Problem Solving Behaviour

Abstract

This paper investigates how local and global self-evaluations of capabilities can be used to predict pupils’ problem-solving behaviour in the domain of fraction learning. To answer this question we analyzed logfiles of pupils who worked on multi-trial fraction tasks. Logistic regression analyses revealed that local confidence judgements assessed online improve the prediction of post-error solving, as well as skipping behaviour significantly, while pre-assessed global perception of competence failed to do so. Yet, for all computed models, the impact of our prediction is rather small. Further research is necessary to enrich these models with other relevant user- as well as task-characteristics to make them usable for adaptation.