Menno Nijboer - Applying Presentation Scheduling to Realistic Conditions

afstudeercolloquium

Learning facts is a large part of any education. Earlier research has shown that learning can be improved by carefully scheduling the order of fact presentations. Concretely, these schedules balance two important memory effects: spacing and the testing effect. In this thesis the latest evolution of a scheduling algorithm, which is based on a cognitive model, is tested under realistic (classroom) conditions. These conditions include regular, fixed length study moments. The model is a refinement of earlier work, and uses the memory equations found in ACT-R to model how participants forget information. This model is continuously adjusted based on reaction time feedback. Two studies were performed: the first experiment had a single study session, while the second experiment had three sessions spread over an entire week. Both studies were concluded with a test. The model was compared against other scheduling algorithms and a standard learning method, and was found to increase performance significantly. Learning behaviour of the participants was analyzed, and it was discovered that test performance can be predicted from the model, which opens up possibilities for tailoring the length of study sessions. Finally, analysis of the test results showed that participants using the model performed significantly better than participants in all other conditions. Thus, this work shows that scheduling using a cognitive model improves learning when used in typical educational circumstances.