On the precipice of learning analytics at scale in Australia

 It has been promised for a long time but with the advent of cloud computing and advanced algorithms, learning analytics is finally becoming a possibility.....and at scale in Australia. As a former educator I would have loved a dashboard for each class with 26 pictures of my students so I could drill in to how they were going, not only in my class, but in all classes as a way to compare their performance in my subject. Are they just declining in my class or all classes, is it the topic, is it my teaching or is in a more generally slide that needs to be discussed with other teachers and coordinators? I wanted to see truancy records for all classes, behavioural incidents as noted by coordinators (without all detail if required), flags that the student might need some extra support and the list goes on. While some of these are now becoming a reality, I am seeing signs that within 2-4 years this kind of insight might be a reality for educators in Australia at scale. There have in the past been blockers to this kind of information at the fingertips of teachers, but there are many examples of analytics projects happening in Australian education systems that could make this an imminent reality.

With increased use of technology in the classroom there is also a set of 'learning signal' data to be gleaned from the technology tools themselves. How long did the student spend on that document, is there a trend of completing work at the last minute, how long do students normally take to complete learning tasks that may serve to support future students? This data is basic telemetry data from a tool like Office 365 (underpinned by the Office Graph which provides a connected view of this telemetry), used by millions of students around the country, but when taken in partnership with other data can be a rich source of insight. Gartner sees the use of graph based technologies as a top 10 analytics technology trend in 2021 with over 50% of businesses looking to use AI considering this type of technology, and I think this trend will be true in education also. The more pieces of data we can get about a student, the more we can help them as teachers and school leaders.

I am just scraping the surface of the possibilities in this field moving forward, but here are some top-of-mind considerations to make sure we get the most out of what this technology can provide. Firstly, we must ensure that we don't lock ourselves into a productised view of analytics by purchasing one solution that does some of the job, without being able to add more sources of data as you see fit. In the latest analytics tech trends published by Gartner, businesses are moving toward being an "intelligent composable business" where more sources of data can be added as required, and the data will provide a user-friendly and useable experience. I think this is also true of education systems and institutions who should aim to create an open, connected and private system of data capture, not limited by the software tool or format of that data, and make sure the user is top of mind when designing representations of the data.

Secondly, we need to make sure the analytics is both broad and deep across an education system or school. Broad meaning able to ingest data from a wide range of sources (into a data lake for example), but also deep in this sense meaning who in the education system has access to it. I worry that while analytics will be very powerful for the education 'system' in terms of school health, financials, like-school comparison models etc., the most tangible value may be realised from analytics the closer we get to the classroom and the students. I hear people talking about analytics not being just about a dashboard, but I have often said that dashboards have become passé before most teachers have one anywhere near as detailed as that I described above. While learning analytics affords potential well beyond that of a dashboard, giving every teacher a dashboard about their class, and student one about their own learning, would be a wonderful place to start. We should aspire to this within the next few years as a minimum.

Lastly (for the purpose of this article anyway), we need to consider how smaller institutions or individual independent schools are going the reap the same benefits from learning analytics given their challenges around scale. The larger a data set, the more reliable, representative and useful that data set is. Take for example the possibilities of 'prescriptive analytics', where an AI system might help a teacher to determine and suggest, based on vast previous experience, what type of study techniques to recommend to a senior student to perform best on upcoming assessments or examinations. In a system of 500,000 students, the AI and analytics engine will have learned from thousands, if not tens of thousands of similar students with similar characteristics prior to make recommendations. In a school of 1000 students, machine learning models or algorithms are less likely to be as reliable given the lack of data they have to ingest and use as the basis of the model. Perhaps there is a need for pooling of these schools and smaller systems to enable them to benefit from larger anonymised data sets?

We are at an exciting tipping point in learning analytics in Australian education, and with careful consideration and a laser like focus on what matters most, we can empower school leaders, teachers and students with greater insights than is currently possible. We are already seeing collaborations such as Project Constellation (led by Microsoft Education) allowing for the creation and sharing of algorithms, machine learning models and data connectors to begin democratising learning analytics. These kinds of initiatives will ensure that all systems and all schools might be able to take advantage of the power of learning analytics in the near future because from what I see in the industry, the potential of this technology is starting to be realised, and fast.