Learning Analytics In Moodle 3.4: What’s In The Box


Moodle 3.4 is giving you access to a dark box that, over time, is supposed to become better at predicting the future. Today, the forecast is binary: the student will either complete the course or drop out before it ends. You will be able to view the sources the box is using to make the predictions and you can play around with them to test the performance and efficiency of the box.

But there is one caveat: The darkness of the box is subjective. The box is, in fact, transparent. Inside the box there are several algorithms that take raw data and process it, sometimes taking the processed data and plugging it back in once again (fun fact: that is called “Backpropagation.”) These algorithms are sometimes labeled “machine learning,” “neural networks,” or “TensorFlow,” which in most cases explain the obscurity of the box. But you can always learn what they mean and poke into the box’s inner mechanisms. This is, in fact, what this box encourages you to do. (Not that it is aware… yet.)

A post in moodle.com explains Moodle Learning Analytics, the Moodle HQ project in charge of the embedded analytics engine that all users will enjoy starting in Moodle 3.4. As the importance of data becomes universally recognized, Moodle and other open systems are starting to provide users the tools to capitalize on their own information and extract value from it without intermediaries.

A new Analytics section in Moodle’s Site administration page gives access to general analytics settings, as well as the models shipped in for analysis. Currently, the only setting is the choice of the engine: the “PHP machine learning backend,” which comes as default, or the Python version, which is considered more powerful and provides graphical results, but requires the Python language installed on the server.

The “models” option is the analytics sandbox where authorized users can play around with the many attributes, particularly the inputs or indicators. As Moodle records pretty much every user transaction, this information can be plugged into the model to generate the prediction.

But of course, the engine uses these data to create other indicators before administrators even get inside. There are two types of indicators built:

  • Social: Interactions that relate to the ability to foster relationships and communication skills that increase learning opportunities.
  • Cognitive: Actions that lead to an increase in skills and subject matter engagement.

The models generate a list of students at risk of dropping out. Unfortunately, hidden among menus is the model’s confidence in its prediction, expressed as a percentage. Before the model has any previous records of past indicators and outcomes, its prediction will be as good as flipping a coin. But over the course of generations, it will identify trends and will become better at adjusting for particular cases.

The Moodle Learning Analytics project, previously named Project Inspire, is ongoing, with more predictors and tools becoming available in upcoming releases of Moodle. But the algorithms can be examined and modified. Analysts can even run several models at a time, each with few differences between them, to refine and benchmark predictions.

Finally, a “Students at risk of dropping out” page in Moodle 3.4 shows a list of students and allows teachers to send them direct messages and review their specific data.

Read the update “Support your learners using Inspire analytics in Moodle 3.4” at moodle.com.

Learn more about Moodle Learning Analytics at moodle.org.

Moonami LogoThis Moodle Technology related post is made possible by: Moonami a company that provides a full range of Moodle services that combine the flexibility, scalability, and power of Amazon’s world-leading cloud platform (AWS) with fanatical Moodle support. Click here to learn more.