What are Learning Analytics?

Lesson goals and objectives

In this lesson you will learn:

  • what Learning Analytics is
  • about different forms of Learning Analytics

Learning Analytics is the measurement, collection, analysis and reporting of data about learners and their contexts, for purposes of understanding and optimising learning and the environments in which it occurs. People have been researching learning and teaching, tracking student progress, analysing school or university data, designing assessments and using evidence to improve teaching and learning for a long time. Learning Analytics builds on these well established disciplines, but seeks to exploit the new opportunities once we capture new forms of digital data from students’ learning activity, and use computational analysis techniques from data science and AI.

The first step in talking about learning analytics is to determine what you mean and what you want. A wide spread classification of analytics, offered by Gartner, may be helpful.

Descriptive analytics produce reports. It merely offers what happened, no explanations. It is quite normal for managers that they want figures on study efficiency and drop-out rates.

Diagnostic analytics offer correlations. A fine example is a survey at Noorderpoort in the Netherlands with the conclusion that students who scored a 7 (out of 10) or higher on English in the preparatory course were more successful than students who scored lower on English in the preparatory course. No such correlation was found for any course. However, a causal relationship could not be established.

Correlations are often confused with causality. When a correlation is found, there is still no evidence that variables are interrelated. Fine examples of variables with a high correlation but obviously no interrelation at all can be found on the website Spurious correlations.

So, the message of course is trying to find out what is behind the figures. If you find correlations, try to find out how they are interrelated and what the dependent and independent variables are. It is fine to wonder why a course has low grades and why some courses have low student attendance while others are attended en masse by students.

An example of descriptive analysis is investigating the inflow, through-flow and outflow of students. In the Netherlands, secondary vocational education has had a barrier-free intake for the past eighteen months. This means that it can happen that students have insufficient prior education to be able to successfully complete a certain study. It is unknown what the results are, but the question is currently being investigated what the effects are of barrier-free intake on through-flow and outflow, but also on the extent to which students drop out of studies prematurely and switch courses.

Predictive analytics offer predictions, often based on the continuation of a trend. This is very similar to the kind of predictions my schoolmaster used to make: “boy, if you go on like this, nothing will come of you”.  Apparently, a trend can still be reversed.
In schools you can notice the same mechanism. Tutors try to reverse the behaviour of students who perform poorly by committing an intervention. Historical data, e.g. the achieved grades in the first period, are used as a prediction for grades to be achieved later on in the studies. On the level of institutional data, you might want to analyse what courses cause major problems for students, so you can customise the course to increase study success. For example, by offering more tutoring or question hours.

Prescriptive analytics offer recommendations. “Your grades so far are excellent, we advise you to study this extra module.” When these recommendations are offered automatically by the system you need an algorithm and use it. If you do, make sure the algorithm is based on the right assumptions, reliable data and relevant performance indicators. Students as well as instructors, managers and even support services can use predictive data to increase their success.

The problem with data is that it has to be accessible, reliable, complete and comparable. Data is often located in different systems and collected for different purposes. Furthermore, there are of course quantitative data and qualitative data, which are often very difficult to connect. Then there are data from internal systems but also data from external systems. So, before you can start with data analysis, it is important to identify the data sources and analyse whether they are suitable for the goals you want to achieve with your data analysis.