Learning analytics- new idea for curriculum review
Learning analytics may be a quite fresh word to most of us. First let us think about a simple question: how can a professor know their students’ daily engagement in his subject? Maybe someone will say: the final exam may reflect whether a student works hard. But, what about a student works hard in whole semester but plays under par in final exam? And what about a student never attends the lecture and read the resources but performs really well and pass the exam?
In high educational field, most learning resources are available online. Thus, engagement in learning system becomes the most essential part in whole students’ learning procedure. Different from conventional teaching method in small class, teachers (or lecturer) in university who impart knowledge to hundreds students are hard to get instant feedback from most of them. Hence, in most time, teacher cannot make a correct judgement on their teaching quality. The good thing is many learning systems such as University’s learning management system (Blackboard) and other related systems (e.g. YouTube, Turnitin, Echo360) provide sufficient data (times of access certain material, student’s connection and etc.) to reflect each students’ online behavior.
So this is what learning analytics do. When massive access records from learning system are collected, we need to implement various techniques to extract the value information from the datasets. Actually, learning analytics is just one part of the whole procedure. To provide a good feedback, three phases are required: gathering, analysis and visualization. In each phase, some limitations still exist thus impact the performance.
In first step, the main problem is the scalability of the datasets. Which size of data is suitable? If the size is too big, the later analysis procedure will be impacted a lot such as spending time is too long, the ability of typical database software tools cannot handle this size of data. On the other hand, if the size is too small, some meaningful dataset will be discarded and some valuable thing may be ignored.
In second part, learning analytics should have strong connection with the science. On the one hand, advanced data mining techniques should be applied to ensure more valuable thing can be analyzed. On the other hand, teachers who are non-expert in the computer science (or science) field should also benefit from the results even they don’t have any idea of these technologies. Thus the input and output should be as simple as possible. The requirement on encapsulation of this part will be very high.
In final step, visualization of data need to be improved. As the output part, it converts the obscure data to variant and obvious charts to help every user can benefit from them. However, which format should be chosen to fit different teachers? (because various teachers focus on different part of datasets) How to combine multiple charts to gain the best performance and convey most information? This part has great connection with arts and psychology.
Besides the previous questions, other problems cannot be ignored.
- How can we ensure the data from the learning system is accurate? For example, the count of a student access one resource is often miscalculated.
- How to build common framework which suit different disciplines?
- Sometimes the data cannot reflect the true story: Bob doesn’t access any online resource this semester, however he obtains all of them from his roommates Alice.
Learning analytics is a new research field and encounter its bottleneck during these years. As explained before, it is a valuable research direction which deserve putting more effort on it. Internet brings the convenience to the high-education but isolates students and teacher. No one hopes education in university is pair-blind.