Umeå universitets logga

Cognitive pre-requisites should be taken into consideration when providing personalized and adaptive digital content in web-based learning platforms. In order to achieve this it should be possible to extract these cognitive characteristics based on students´ study behavior. Working memory capacity (WMC) is one of the cognitive characteristics that affect students’ performance and their academic achievements. However, traditional approaches to measuring WMC are cognitively demanding and time consuming. In order to simplify these measures, Chang et al. (2015) proposed an approach that can automatically identify students’ WMC based on their study behavior patterns. The intriguing question is then whether there are study behavior characteristics that correspond to the students’ WMC? This work explores to what extent it is possible to map individual WMC data onto individual patterns of learning by correlating working memory capacity with learners´ study behavior in an adaptive web-based learning system. Several machine learning models together with a rich context model have been applied to identify the most relevant study behavior characteristics and to predict students’ WMC. The evaluation was performed based on data collected from 122 students during a period of 2 years using a web-based learning platform. The initial results show that there is no linear correlation with learners´study behavior and their WMC.