Data Science & Machine Learning Providing Engagement & Learning Analytics

Concurrent Session 3

Brief Abstract

In a world where we have no shortage of data, one thing that is present is a lack of understanding around our students and their interaction with online learning content. We talk about using data science and machine learning to help us understand more about your students.


Data is something I am passionate about, but more than that, I am passionate about education and the future of it. How we can improve the lives of all of those students who don't have access to the same quality of education, giving them the right to the same education as everyone around the world. We have a duty to ensure that our future generations are given the best chances they have of improving the world around us for generations to come.

Extended Abstract

As a teacher creating a lesson plan for your class, you know that that content is not going to engage 100% of your students. You cater for the middle ground and you are there to catch the upper and lower quartiles that do not match that lesson or learning style for that subject or topic. 

Yet, with online learning we continue to try and deliver content to our students without really understanding what is most effective for them. In fact, worst than that, with little standards around creation of online learning content, it's very hard to get any undertstanding about what works and what doesn't, for each student, and why. Not to mention the fact that you cannot catch those students who are not that middle ground and aren't engaged in that course you've created. 

So our aim is understanding the learner, the student, and how they engage in content, and how they learn. Where you have most analytics, that look at the way students have interacted with systems, we don't believe that is sufficient to understanding students and is open to statistical bias and misinterpretation.

Why? Because it lacks context. That context is the human element involved in all of these actions. Take that out of the equation and you don't understand why the event occured in the first place. Time on task, page views, submissions etc. are all open to interpretation in different ways, and while we can look for patterns in those data sets, it's still drawing impersonal, uncontextual averages, to which student success and predictive analytics is aligned. 

Using facial expression recognition to analyze the emotional response of students as they take courses online allows us to do a few things different from that. Not only can we start to look at the student as an individual; comparing their own data to themselves and how they respond (and even comparing it to the other, previously uncontextual data points like page views and submissions etc.) to add the individual context to really understand that student's interaction with the course and material. 

On top of this, we can see patterns in behaviors around certain key triggers/events, and track these behaviors to responses and content. Allowing us to really start to not only understand when a student is or isn't engaged, but what is increasing their learning.