Personalization in Online Learning Platforms
The art of recommendation systems
Introduction
I once taught a group of kids Python programming. It was one of the most insightful and joyful experiences I ever had. Kids are super smart! They are creative and fast learners — there is no limit to their imagination.
Going into this experience, I had expected all students to learn in the exact same manner. This was far from true. While several students picked up the topic quickly, others didn’t. While some were eager to learn, others were barely interested.
That’s when it occurred to me to adjust my teaching methods. After conducting some research, I decided to adopt a well-known learning methodology, known as “peer learning”. whereby students interact with and learn from one another. The result would be a pace-driven and personalized learning process that is fun, engaging, and tailored to every student’s needs.
I don’t believe there is one optimal way of teaching all students within a class, for every one of them has different strengths, capabilities, and interests. Nevertheless, personalizing the learning experience remains one of the biggest challenges in education. But by providing courses and learning materials for students to study at their own pace, online learning platforms are leading the way towards personalized learning as opposed to traditional offline learning.
What is Personalized learning?
Personalized learning refers to instruction in which the pace of learning and the instructional approach are optimized for the needs of each learner. Learning objectives, instructional approaches, and instructional content (and its sequencing) may all vary based on learner needs. In addition, learning activities are meaningful and relevant to learners, driven by their interests, and often self-initiated.
— Office of Educational Technology
There are different definitions for personalized learning but I would highlight it in the following 4 principles:
- Learning pace is adjustable for each learner
- Learning objectives and approaches are tailored for each learner
- Learning is driven by learner interest
- Learners have the choice in what, how, and where they learn
In this article, I will talk about one way of personalizing the learning experience for the learners in online learning platforms..
Recommendation Systems
We live in a world that values data as much as, if not more than, oil. Technology has made our lives easier, particularly artificial intelligence (AI), which we can leverage to personalize the student learning experience. To this end, recommendation systems play a pivotal role.
A recommendation system is an intelligent system designed to propose items to users based on numerous factors, such as their usage history on a certain platform and the behavior of similar users. Within the online learning context, the recommended items could be learning materials or paths that resemble the student’s learning history and the behavior of similar students.
Recommendation System Types
There are three types of recommendation systems, which I will discuss within the online learning context:
1- Course-Based Recommendations (content-based)
In course-based filtering, the system recommends courses similar to those in which the student is enrolled in.
2- Student-Based Recommendations (collaborative-based)
Unlike course-based filtering, student-based filtering recommends courses taken by other similar students. The similarity is measured based on previously taken courses.
3- Hybrid-Based Recommendations
Hybrid-based filtering is a combination of course-based and student-based filtering, offering the best of both.
There are many applications for recommendation systems in online learning platforms. We can achieve our goal of personalizing students’ experiences by:
Recommending learning paths
Suggesting or placing a student on an optimal learning path based on his/her behavior, preferences, and learning activities.
Recommending learning materials
- Suggesting courses based on previous course enrollments.
- Suggesting a specific learning object (lesson, video, exercise, etc.) based on the student’s learning progress.
Therefore, Learning objectives and approaches will be tailored for each learner and the learning process will be driven by learners' interests.
Recommendation System in Edraak Platform
In Edraak (an Arabic MOOC platform), we use course-based filtering to recommend courses that a student might find interesting.
The list of recommended courses is shared with students via email, on the courses landing page, or — ideally — on the dashboard, which is the platform’s frequently-visited entry point.
Take a look at the following example. Here, the user is enrolled in the “Internet of Things” course under the “Technology” category.
Accordingly, the user will receive the below recommendations:
Thus, the learner has gained a personalized experience based on his/her interests.
Cold Start Problem
A common issue faced with recommendation systems is the cold start problem. Basically, when new students join the platform we lack usage history from which to recommend courses.
At Edraak, we resolved this issue by introducing an onboarding process. Instead of showing an empty dashboard, we ask new students a few questions to capture their preferences and interests. Doing so helps us build an initial student profile that enables us to recommend suitable courses.
Recommendation Engine Model
The recommendation model is built using Term Frequency-Inverse Document Frequency (TF-IDF), a numerical statistic that highlights how important a word is to a course in a collection of courses.
The model scans the course content and highlights its most important parts. It then uses the cosine similarity measurement, which gauges the similarity between two non-zero vectors of an inner product space. This measurement evaluates the similarity between two courses and builds a matrix (2D array), where each cell contains the similarity score between course A and course B. For example, similarity[A][B] = 0.897 means the similarity score between course A and course B is 89.7%.
This model aims to predict user interests, given that it is more likely for a student to enroll in a course that is similar to the one they are currently enrolled in (i.e. courses with a high similarity score).
Effect on Edraak’s Platform
The introduction of the personalization features (recommendation engine and onboarding process) positively impacted user behavior:-
- User enrollments increased by 35%.
- The dashboard exit rate decreased by 53%.
Conclusion
There are many ways to personalize a student learning experience on an online learning platform. However, recommendation systems remain a personal favorite! Their ability to predict what we will most likely prefer — even if not 100% accurate — makes them highly valuable within the education field, as it motivates online students to learn more via a personalized experience.
I would like to thank you for going through this article, don’t forget to give it a few claps if you like it.
