Statistical Analysis of Procedural Learning (Blocked vs. Interleaved) in Humans

Studying student learning patterns, and their outcomes on learning goals is a topic which has been often studied in education related literature. There are two kinds of study patterns which humans employ: blocked and interleaved. In a blocked learning pattern, learners study materials related to the same topic at once, and move on to another topic only when the first one is complete. On the other hand, in interleaved study patterns, this order is scrambled. There have been several studies that analyze the differences in blocked and interleaved study patterns, and comparing learning outcomes in both. However, most of this work has been associated with the study of declarative learning tasks, which are memory-based. Therefore, we can argue that such studies look at memory as opposed to learning. 

In our work, we focus on procedural learning. These are the tasks which you cannot learn by memorization, but have to gradually acquire the skills. We choose computer programming as our declarative task, and study the differences in skills gained between participants in blocked and interleaved patterns. We find that interleaved study pattern works best for the programming task.

Furthermore, we also aim to identify the optimal sequence of study modules using statistical modeling techniques. We use the data collected in our study and model it using a Bayesian Network. Our results indicate that it is possible to infer an optimal sequence from performance metrics on several random sequences.