In March this year we officially launched Project Inspire – our open source solution to provide learning analytics features natively within Moodle. Inspire supports a wide range of predictive models and machine learning backends, all securely on the institution’s Moodle site, without exposing personal information to an outside agency.
Since then Phase I of the project has been collecting historical log data from a number of organisations and institutions using Moodle, using the Anonymise plugin, which enables any institution using Moodle 2.7+ to easily (and safely) contribute data. Project Inspire uses the data to train analytical models. The more data we collect, the better our models become.
In Moodle 3.4, Inspire System will become part of Moodle Core, making it automatically available to all Moodle sites, whether locally managed, Partner hosted or hosted on Moodle Cloud.
Following that, we will be expanding the scope of the Inspire System to include other aspects of Moodle use and you can read about them and also add your own aspirations, thoughts and ideas for the project on our roadmap.
When we first introduced Project Inspire we talked to Moodle HQ Research Analyst and a member of our new Education team, Elizabeth Dalton.
So what better way to get an update of where Project Inspire is at and how you also can be inspired and join the movement for the next phase, than to once again talk to Elizabeth?
Moodle HQ: Thanks Elizabeth for taking the time to chat to us about the progress of Project Inspire.
Can you give us a quick summary of what’s happened since we launched in March?
Elizabeth: In May, the first public release of Inspire as a third-party plugin was made available, coinciding with the release of Moodle 3.3. The predictive model included with this version can be applied to any Moodle course in Moodle 3.3.
Reception has been very positive. We’ve received a lot of great feedback from Moodle users and from the learning analytics research community. For example, I was asked to give a short presentation about Inspire at the Learning Analytics Summer Institute, in Ann Arbor, Michigan. This is a workshop event attended by researchers in learning analytics from around the world. The learning analytics research community is very focused on ethics and accountability of analytics tools, and we are right in line with those concerns in our designs for Inspire. I’ve also been asked to give short webinar presentations to various universities and Moodle users groups around the world. I’m happy to provide these presentations on request.
We’ve also received a few more offers of site data to analyse, which is really great, because we need a lot more data if we are going to ship “trained” models. The current plugin requires the site administrator to train the model on their own data before it can be used to make predictions.
Moodle HQ: So the Inspire plugin for Phase 1 is now available, at the same time as Moodle 3.3 release. Can you tell us a little bit what the plugin does and if a Moodle site would like to use it, how they can do that?
Elizabeth: As with all regular Moodle plugins, the Inspire plugin can be downloaded and installed from the Moodle Plugins database. It does require at least Moodle 3.3 and PHP v7. It ships with one built-in prediction model, “Students at risk of dropping out of courses,” which will probably be most useful to higher education institutions who are delivering fully or mostly online courses, but there is also an API allowing anyone with php skills to build additional models. The existing model provides proactive notifications to instructors by using the Events architecture, and the tool makes it easy for instructors to send messages of encouragement to students identified by the model.
Because this model is trained on the site’s own data, predictions should be quite accurate for each site, but the model does require a certain quantity of data to train. Inspire will check during the model training process and let the site administrator know if there is enough high-quality data on the site to be able to make accurate predictions.
More information is available in the documentation.
Moodle HQ: When Inspire is incorporated into Moodle 3.4 can you describe to us how it will help Moodle sites and users?
Elizabeth: We’re planning several major enhancements, in addition to shipping Inspire automatically as part of Moodle Core. Some of these changes will be to other parts of Moodle, so they will be visible even to those not using Inspire yet.
For example, we’re adding fields to the Course Description to specify the proportion of the course intended to be completed online and the number of hours of student effort the course is expected to take, which can really help improve predictions based on student effort. At the site level, we’ll be tracking what kind of education is being offered on the Moodle server, e.g. K-12, Higher Education, Corporate Training, etc.
We are preparing more models that will be appropriate to these different contexts, and we’d like Inspire to be able to recommend the best models for a site based on what the site is being used for. A full list of enhancements being worked on is here.
One of the most exciting features we are working on is a way to determine how much Inspire is helping learners and teachers by tracking the effects when messages are sent about Inspire predictions. We are not aware of any other system, commercial or otherwise, that builds in this kind of evaluation feature. As I mentioned before, we already check each model to see how well it will work with the site’s data, but this will go beyond that– site administrators will be able to see whether having Inspire installed and used by teachers is making a difference to students.
Moodle HQ: What else is still happening in Phase 1 and can Moodlers still get involved or contribute to this phase or for next steps of the project?
Elizabeth: Phase 1 is over, but people can certainly still get involved! We are still collecting data sets– we need a lot more data from universities, K-12 schools, corporations using Moodle for training… everyone, really. We have a very aggressive anonymisation utility that protects user identities in the data collection process. Please see here for full details.
We also want to encourage everyone to look at the models and indicators we are using and offer suggestions about what you need to help learners be successful on your own Moodle– or suggestions about other kinds of analytics you’d like to see included in Inspire.
Thanks Elizabeth for taking the time to chat to us about the progress of Project Inspire.