Research begins with observations. “explain” the observations using a new or existing theory; making hypotheses and testing them (ideally with controlled experiments). Students spend the majority of their time outside class, and it is has been very difficult to collect such data. For example, imagine that I proposed to mount cameras over the desks of all our students so RELATE could measure the time they spent on each problem, and observe which page of the textbook was helpful in solving each problem. Obviously, would be challenging to implement even if approved by the institutional review board.
But when a student uses an online learning system, they create a time-stamped log of their path through the resources they access, providing the data mentioned above “for free”. In addition, their answers to problems (both incorrect as well as correct) as well as any contributions to discussion forums, etc. are also logged. Typically a student would generate most of a megabyte in a semester – several hundred pages of individual logged clicks. This creates an enormous amount of data for an entire class. Our big research challenge is to interpret it to provide informative or actionable conclusions.
To start making sense of these data, we start with Data Mining of the logs of student interactions with the online system. The first thing we study is how students allocate their time; in particular how much time is spent on homework, watching videos, reading e-text, etc. Then we seek to understand student strategies and work habits; do they read e-text before or during homework – or possibly not at all.
As scientists, we always seek to understand the foundations of our work. This has caused us to conduct surveys and interviews to determine What should students learn, and perhaps more important, What is remembered at graduation?