Adaptive Learning Environments using Cross-Subject Classification

According to Cognitive Load Theory (CLT) (Sweller et al. 1998) the type and amount of cognitive load during learning is crucial for successful learning. CLT recommends that cognitive load should not exceed working memory capacity at any time during learning. Therefore it should be held within an optimal range of learners’ memory capacity. A continuous online measure of cognitive load during learning is necessary to investigate whether real-time learning environments fulfill this criterion. Hence, we aim to developi computer-based learning environments adapting to learners’ individual cognitive load online and assisting individuals to increase their learning success. Antonenko et al. 2010 suggest that electroencephalogram (EEG)- data might be used to measure the current cognitive load of a subject instantaneous.

In order to justify this approach, a learning environment has to be tested with subjects. In an adaptive learning environment it is not feasible to use data from the same subject and the same task for classifier training and testing. Therefore we demonstrate how to cope with these challenges and illustrate why a cross-subject workload prediction based on EEG-data is a promising approach to develop a computer-based adaptive learning environment which individually supports learners in an optimal way.