A literature review of implemented recommendation techniques used in Massive Open online Courses

Highlights

• Systematic literature review of implemented recommender systems for MOOCs.

• Design of a classification framework to review recommender systems for MOOCs.

• Analysis of research trends on recommender systems for MOOCs.

• Identification of future areas for recommender systems in MOOCs.

• It serves the purpose of offering an up-to-date review of the field and showing its evolution trends.

Abstract

Massive Open Online Courses (MOOCs) are receiving attention from learners because MOOCs enable them to satisfy their learning needs through an open, participatory, and distributed way. With the increased interest from learners, the number of MOOCs available is increasing which has increased options for learners. This as a result has created the need for recommendation systems that help learners select suitable MOOCs. This literature review covers analysis of recommender systems (RSs) that have been implemented in MOOCs with the goal of providing insights on the trends reported in the academic literature on recommender systems in MOOCs. The review discusses different recommendation techniques, recommendation types and evaluation techniques that have been used and reported on. This review includes research work over eight years, i.e. from 1st January 2012 to 17th November 2020. After the filtering process, 67 papers were selected from journals and conferences from four academic databases (i.e., IEEE, ACM, Science Direct, and Springer). A framework is designed that classifies literature on the basis of both design and evaluation aspects of RS in MOOCs. This review concludes by highlighting gaps found in the literature.

Introduction

Since the advent of the first Massive Open Online Course (MOOC) in 2008 (Downes, 2008), MOOCs have added a new dimension to education systems with the main attraction being access to free open education courses. The number of MOOCs and the number of students registered in MOOCs are growing every year with MOOCs using a number of platforms such as edX, Coursera, Udacity, NetEase and iCourse (Kang, 2014). By the end of $2019$, more then $900$ universities were offering MOOCs with $13500$ courses available, with around $110$ million students registered (Dhawal Shah, 2020b). After the COVID-19 pandemic of 2020 an increase in the online education trend resulted in a surge in the number of new user registrations and the introduction of new courses by different universities. Approximately 25%–30% more registered users joined online courses after the pandemic (Dhawal Shah, 2020a). By the end of $2020$, there were more than $950$ universities offering MOOCs with $16.3$k courses available online and around $180$ million new learners registered on them (Dhawal Shah, 2020c). With such a large number of courses available, learners face the problem of selecting courses without being overwhelmed with multiple learning choices (Zhang et al., 2018), which has been termed as an information overload (Raval & Jani, 2016) problem. Information filtering systems, also known as recommender systems(RSs), can overcome this problem by helping learners find suitable courses from the array of available resources.

With the increased usage of MOOCs, data produced by MOOCs is also increasing and this data contain information about the interests and behaviours of learners (Lundqvist & Warburton, 2019). Recommender systems have been used widely by commercial and social platforms (Zhang et al., 2018) and can use educational data to provide recommendations to learners (Ricci et al., 2010). One purpose of these systems is to help the learner by recommending different related learning objects or elements. In addition to improving the learners’ experience, these systems have also played a vital role in increasing the popularity of MOOCs.

Recommender systems are intelligent filtering applications that help users to filter out items or information according to their requirements or interests from a large number of services or products. Recommender systems make it easier for users to obtain related information even for items for which they have little knowledge or experience. Some RS provide related information in the form of a priority list, where products or services that are closer to user’s interest are placed in higher priority compared with resources which are of lesser interest to the user. For example, the GroupLens system (Resnick et al., 1994) is designed on the assumption that each time a user reads a Usenet News article they will give an opinion. The system will use all opinions as a rating and from these ratings, people with the same rating are considered like-minded, so they can be used to predict the ratings of each other.

RS can be divided into two broad categories (Manouselis et al., 2012) of ‘Collaborative Filtering (CF)’ and ‘Content-based (CBF)’ RS. There is a further type called ‘Hybrid (HB)’ RS that contains characteristics of both collaborative filtering and content-based RS. CF based RSs perform recommendations on the assumption that people who had similar tastes in the past will make the same choice in the future. This is similar to real life scenarios where we have to choose something from multiple available options and we consider recommendations of family and friends who have similar interests (Dakhel & Mahdavi, 2013). CBF RSs consider the profile of items and users to perform recommendations. Profiles include different characteristics of users and items, for example user (age, gender, education, residency area etc.) and item (actor, genre, category, type etc. in the case of a Movie item). These RS analyse the profile of items rated by a user and try to design a model that reflects the interests of the user. This model is used to recommend new items to the user (Lops et al., 2011).

After analysing the literature on RS in MOOCs, we concluded there is a need for a systematic synthesis of literature that summarizes research work and points towards future research options. While reviewing the literature on RS in MOOCs, we found that RSs in MOOCs is an interdisciplinary research field which is of interest to both technical and non technical researchers. After a detailed synthesis we decided to divide this literature into two fields for review, one which summarized the literature for non technical researchers (Author Ref. Khalid et al., 2020), for example researchers from distance learning and education, and the other which will discuss RS in MOOCs from a technical point of view.

To analyse the literature from a technical point of view, we have reviewed the state of the art of RS in MOOCs. The objective of this literature review is to summarize existing work in this field in order to identify gaps and areas in the design and implementation of RS in MOOCs that can help in future work. We reviewed work between January 1st 2012 and November 17, 2020 in the English language only. We chose 2012 as the starting year because it was declared as the “Year of the MOOC” by the New York Times (Pappano, 2012) and from this year onwards the publication of peer-reviewed research on RSs in MOOCs started. We have designed a classification framework which includes both design and evaluation aspects of RS in MOOCs. In addition, we have highlighted the key subject areas of MOOC datasets used in the research. To the best of our knowledge, this is the first literature review in this area that discusses the technical aspects of implemented recommender systems in MOOCs. This literature review not only highlights the trends in the existing literature in this field but also discusses the gaps and possible new research lines.

The remainder of this paper is organized as follows: Section 2 describes the data collection and methodology used to classify the literature work. Section 3 analyses the literature according to the classification framework. Section 4 presents trends found in the literature by discussing each recommendation type found, concluding remarks and research gaps are presented in Section 5.