You are on page 1of 8

Personalization Criteria for enhancing learner

engagement in MOOC platforms


Sara Assami, Najima Daoudi, Rachida Ajhoun
Smart Systems Laboratory
National School of Computer Science and Systems
Analysis (ENSIAS) –Mohammed V University
Rabat, Morocco
saraassami@gmail.com
ndaoudi@esi.ac.ma
r.ajhoun@um5s.net.ma

Abstract— The use of ICT (Information and Who takes massive open online courses and why? What
Communication Technology) revolutionized the education personal data do MOOC platforms hold about their
sector with MOOCs (Massive Open Online Courses) becoming learners? What criteria can be obtained from a MOOC
a major focus for open education actors and top universities to learner profile to recommend suitable MOOCs?
resolve the limited resources issue and make education
To answer all the above, the second section defines our
accessible worldwide. Our study focuses on the problem of
dropouts in MOOCs and how we can reduce the phenomenon research main argument about the need for personalization
by adapting learning according to criteria that enhance to ensure quality. In the third section, the paper gives an
MOOC learners’ motivation. It relies on enhancing the overview of literature about personalization in MOOCs and
recommendation process of Courses according to the learner the fourth section summarizes the main personalization
personal expectations, preferences and cognitive learning style criteria that apply to MOOC platforms. The last section
among other personalization criteria. Furthermore, it suggests investigates the adaptive recommender system that we
a new approach to profile construction for a MOOC propose to increase learner’s motivation.
recommendation system by considering external data to define
a learner’s interests. II. QUALITY ASSURANCE IN MOOCS
Keywords— MOOC; Personalization; Recomender System; In general, MOOCs are “courses designed for many
Learner profile; Personalization criteria. participants, that can be accessed by anyone anywhere as
long as they have an internet connection, are open to
I. INTRODUCTION everyone without entry qualifications, and offer a full course
The use of ICT (Information and Communication experience online for free” [1]. Nonetheless, most of the
Technology) revolutionized the education sector in terms of articles about this new phenomenon in online learning
the methods of teaching, evaluating and enhancing student’s environments criticize the low completion or retention rates
performance. In this context, MOOCs (Massive Open of MOOCs. Indeed, the high drop-out rate attains 97% in
Online Courses) have become a major focus for open certain MOOCs [2]. However, each stakeholder analyzes the
education actors and top universities to resolve the limited phenomenon by his own motivations and role in MOOCs.
resources issue and make education accessible worldwide. A. Dropout rate critics in MOOCs
Our study focuses on the problem of dropouts in MOOCs
and how we can reduce the phenomenon. The approaches Some researchers criticize the method of calculating the
used are, first, the study of MOOCs performance assessment drop-out rate for several reasons [3], [4], [5], like integrating
and the criticisms about their quality as well as the different registrants who have not actually participated in the MOOC
requirements to ensure this quality. The second approach is dropout rate [4], [5] and the poor comparison between the
the adaptation of learning according to a MOOC learner retention rate in MOOCs and the success rate of traditional
profile that relies specifically on enhancing the courses, due to a poor understanding of the openness of
recommendation process to increase learner’s engagement MOOCs where enrollees are free to follow or quit a course
in MOOC platforms. Hence, it’ll contribute to the [5].
improvement of its quality. Furthermore, it investigates a Furthermore, [6] states that quality depends on its type
new approach to profile construction for a MOOC (open or for profit) and its goal and is limited to satisfying
recommendation system by considering external data to marketing goals of financing institutions and not the
describe a learner’s interest. openness of higher education. However, it does not
In order to name these interests and learner needs, in our necessarily influence the dropout rate in MOOCs since the
literature review we examined many questions, such as: MOOC completion rate does not always show its quality [6]

978-1-5386-2957-4/18/$31.00 ©2018 IEEE 17-20 April, 2018, Santa Cruz de Tenerife, Canary Islands, Spain
2018 IEEE Global Engineering Education Conference (EDUCON)
Page 1265
Authorized licensed use limited to: University of Peshawar. Downloaded on September 17,2020 at 11:28:40 UTC from IEEE Xplore. Restrictions apply.
but it is still an aspect to consider when approaching the European MOOCs initiative launched in April 2013 by the
issue of dropout rates. European Commission and the European Association of
In this context, a study about the quality of teaching in Distance Teaching Universities (EADTU). Therefore, a
76 randomly selected MOOCs [7] was developed by quality label has been established to guarantee the quality of
reviewing the course description, resources, learning MOOCs produced by institutions interested in becoming
activities and learners' interactions and reactions. Ten members of this partnership. The label is based on the
principles have been suggested to analyze their quality: five existing quality assurance approaches in e-learning and
for quality of teaching and five criteria for quality of derives from the Excellence label applied to e-learning and
learning resources [7].The study showed that the degree of
blended learning [10]. This quality label uses a comparison
this quality is low. Regarding the criteria for personalization
process between the institution's performance and the best
(collaborative learning and differentiation), only 21 MOOCs
practices in the field of MOOCs and open education. On a
out of the 76 evaluated offered one of three personalization
options: the choice of activities, the choice of learning practical level, the new institutions seeking the label carry
resources or the choice of the content media. Additionally, out a self-assessment and a review process adapted to
few of the 76 MOOCs had activities requiring collaboration MOOCs using benchmarks at the level of the institution on a
and none promoted the diversity in the learning group regular basis and at the level of the course for each MOOC
members [7]. by 8 features of assessment [10]. We have identified the
Another study of [8], about 379 learners enrollment in benchmark elements to consider for a better quality that are
popular MOOC platforms, explored the effect of MOOC used in learning personalization : the definition of learning
features on retention rates as well as the impact of objectives and workloads, the use of social networking, the
participant demographics. The study showed that MOOC statement of learning outcomes, the level of interaction, and
Content has a significant effect on retention rate, along with the coherence of learning.
“the interaction with the instructor” factor; it explains 79% In general, we couldn’t find standards or specifications
of the MOOC retention cases [8]. These studies show that related to MOOCs other than the quality label of OpenupEd
the quality in MOOCs depends on the quality of teaching but we can deduce that quality requirements related to
and the quality in course material. The latter is the focus of MOOCs and open online education could be met by the
our paper since it could be adapted according to learner’s personalization of learning depending on learner’s
needs and interests through customization tools. characteristics and course content.
B. Quality assurance initiatives for MOOCs III. LEARNING PERSONALIZATION IN
Quality is a concept used in all areas where a certain MOOCS:STATE OF THE ART
level of performance is required to confirm that an approach As personalization is an attractive feature for any service
or a system or a product is useful. This prompts us to rendering platform, it becomes pertinent to consider it as a
examine the different standards, practices and decisive factor in MOOCs to favor learning for all types of
recommendations about the quality of online learning, and learners.
especially for open learning like the quality code of the UK
Quality Assurance Agency (QAA) and the quality label of A. The concept of personalized learning
OpenupEd. The personalization of learning is an education adapted
to the needs of the learner, according to his pace, his
1) The quality code of the UK QAA (Quality Assurance learning preferences and his specific interests [11]. It
Agency) is meant for higher education and developed by the includes two facets: the configuration of resources (or
UK Quality Assurance Agency for education providers to educational software) and the personalization of the
ensure to students a high quality educational experience [9]. pedagogical resources [12]. However, it should be
The code consists of three parts. Regarding the second part distinguished from: individualized learning, differentiated
on quality, it is presented by eleven chapters for the different
learning and adaptive learning:
aspects of academic quality including chapter B3 "Learning
and Teaching" [9]. This chapter focuses on learning
x Individualized learning: “refers to instruction that is
opportunities for students and staff. From this code, we have paced to the learning needs of different learners.
identified four elements to consider for quality that could be Learning goals are the same for all students, but
used as personalization criteria which are: course description, students can progress through the material at different
learning outcomes, learner progress, and learner feedback. speeds according to their learning needs.”[11].
2) The OpenupEd quality label elaborated by the OpenupEd x Differentiated learning: “refers to instruction that is
who is an open, non-profit partnership offering MOOCs to tailored to the learning preferences of different
open access to education. Through its portal, it is one of the learners. Learning goals are the same for all students,
largest providers of MOOCs for higher education by but the method or approach of instruction varies
indexing courses from institutions around the world and according to the preferences of each student or what
mainly in Europe [10]. OpenupEd was the first pan-

978-1-5386-2957-4/18/$31.00 ©2018 IEEE 17-20 April, 2018, Santa Cruz de Tenerife, Canary Islands, Spain
2018 IEEE Global Engineering Education Conference (EDUCON)
Page 1266
Authorized licensed use limited to: University of Peshawar. Downloaded on September 17,2020 at 11:28:40 UTC from IEEE Xplore. Restrictions apply.
research has found works best for students like incidental, discovery and apprentice) and then, through the
them”.[11]. test results, intelligent feedback was provided to motivate,
x Adaptive learning: refers to a learning that guide and assist the learner to overcome his lack of
“continually takes data from students and adapts to knowledge for a concept [15]. However, “it does not take
their learning” [11]. In the context of MOOCs, it is a into account the participant’s competences, i.e. it does not
technology used to assign resources to learners based consider the prior knowledge that the learner has” [16].
on their specific needs. CogBooks is an Adapted Learning Network System
To conclude, personalized learning includes all the three where learning activities are stored in networks with
concepts since it gives learning instructions that respect the interrelationships. As the learner works on a subject, an
learner’s pace, preferences and needs or interests. algorithm is used to identify the next most suitable object in
Nonetheless, adaptive learning is more suitable for MOOC the network. The system inputs the learner's knowledge
platforms since it “takes the parameter of time into account profile, behavioral data, retention level, and demographic
meaning that when the learner’s learning is repeatedly data and uses it to output an adapted learning sequence. The
evaluated, the learner’s status adapts to the learner. So if a sequence is readapted based on the updated information
system repeatedly evaluates the personalization parameters, about the learner [17].
it will become adaptive” [11]. As for MOOC design Frameworks, [18] suggested “a
data architecture based on semantic web technologies that
B. Personalized learning in MOOCs support the inclusion of open educational materials in
In the context of MOOCs, The research on massive online courses” [18]. The study focused on the
personalization has produced a new type of MOOCs reuse, integration and interoperability of OER directories
entitled: “aMOOC (adaptive MOOC)” where the course using linked data techniques. The main goal was to help
adapts to the learning preferences of the individual. Thus, its users to find relevant information or resources relevant by
content is provided with various strategies and intelligent matching between MOOC description data and user profile
feedback [11]. Ref. [13] describes adaptive MOOCs as data (preferences and resources needs).
MOOCs that “use the power of software to personalize and On the other hand, [11] suggested an Adaptive MOOC
turbo-charge the learning experience”. It involves design Framework (AMDF) that satisfies eleven
“Adaptivity”, where the system retrieves the learners' data personalization parameters of seventeen found in his
and adapts to their learning by analyzing and evaluating the literature review (cf. section IV for more details). The
personalization parameters and “Adaptability” that offers modeled personalization parameters for MOOC design
the learner to choose learning parameters himself. were: Information seeking task, Learner’s level of
Nonetheless, personalization is more complicated given the knowledge, motivation level, media preference, language
lack of information on the learner profile and the preference, learning style, Navigation preference, patience,
massiveness of the participants. G. Paquette [14] states that location, weather, date and time [11]. The AMDF was based
“the big challenge of massively open online courses is mainly on learning styles by using Felder and Silverman
personalization, because we find ourselves in front of learning style model (FSLSM) as a reference for learning
thousands of students”. styles, and excluded the “learner goals and plans”, an
The solutions of researchers and institutions for the essential parameter for individualization of learning.
personalization of learning in a MOOC converge towards:
The same focus on one criterion is observed in the
x Adaptive learning environments to provide an framework for personalizing and adapting MOOCs of [16],
intelligent feedback according to the learner’s that designs MOOCs by “identifying each participant’s
performance. competence to generate matching personalized learning”
x Recommendation systems for learners to filter [16]. The framework took also into account the type of
MOOCs according to their needs. device used by the learner, but this is considered more as a
learning facility [19], like the media and language
In the following literature review, we examined most preferences mentioned by [11].
research work related to adaptation of MOOCs in their
design as well as their delivery to learners to deduce in 2)MOOC Recommendation systems for learners
section IV the criteria of personalization that could be used One of the first recommendation systems in the context
for learners. of adaptive learning for MOOCs is the work of [20] who
“proposed an improvised adaptive system, in which
1)Adaptation of MOOC design and/or content: adaptation is achieved by modeling the learners/participants,
For Adaptive Learning Technologies (ALT), two as per their knowledge levels” [20] using fuzzy logic for
environments stand out: AMOL and Cogbooks. AMOL knowledge modeling. Thus, according to the rules of
(Adaptive Mobile Learning) is an adaptive environment for recommendation, the learner receives a recommendation of
MOOCs, where the first aMOOC "Computational tasks at the end of each session of the course.
Molecular dynamics" was created [15]. In this MOOC, the On the one hand, some researchers like [21] developed a
learner would pass a test to identify his learning preference MOOC filter based on a single criterion: the expected
(one among five learning strategies: inductive, deductive,

978-1-5386-2957-4/18/$31.00 ©2018 IEEE 17-20 April, 2018, Santa Cruz de Tenerife, Canary Islands, Spain
2018 IEEE Global Engineering Education Conference (EDUCON)
Page 1267
Authorized licensed use limited to: University of Peshawar. Downloaded on September 17,2020 at 11:28:40 UTC from IEEE Xplore. Restrictions apply.
learning outcome. MOOCrank, the prototype of the interface, his knowledge, experience, competence and his
recommendation system of [4] combines learning outcomes learning objectives [26].
with MOOCs according to taxonomies developed by In a more recent research, [11] identified seventeen
reference institutions. On the other hand, others included personalization parameters for learning in his literature
different characteristics like [22] whose Framework of a review: eleven parameters widely used in e-learning
recommender system included the following features: personalization tools [cf. section III] and six parameters for
“cognitive level, knowledge background, personal education: Learning goals, participation balance for group
expectation, learning motivation and learning style” [22]. management, progress on task, waiting for feedback,
The oriented system was based on the administration of a cognitive traits and the pedagogical approach.
questionnaire to find learning expectations and machine In general, most of personalization parameters in e-
learning techniques for learning model construction to make learning are reused and adapted in MOOC adaptive
personalized recommendation. platforms.
Ref. [23] worked on the Model “Persua2MOOC” which
B. Personalization criteria in MOOCs
generates a list of links and learning objects in an
individualized learning path by analyzing the trajectories of The personalization criteria with identical or close
learners in a MOOC, i.e. their navigation traces. The study meaning were merged together to get a list of all the
aims for the prevention of learners’ demotivation by common personalization criteria from the literature. In the
recommending adapted learning activities (a personalized criteria description, we analyze the different aspects to
learning strategy). consider for a better understanding of these criteria. A
In the same frame, [24] suggested a framework for the summary of these criteria is presented in Fig. 1.
recommendation of instructional material using the learner’s We omitted the following criteria: motivation level of a
profile. It allows MOOC learners to choose their learning course, date and time and weather because of their
path based on recommended learning resources. The usefulness to instructional design of MOOCs not their
recommendation is based on two main criteria: learner recommendation.
preferences and objectives [24]. Nonetheless, the instant
personalized feedback and recommendation process relies
only on the knowledge level of the learner determined by
concept based quizzes after a new week course. The learner
preferences and learning objectives are retrieved using a
survey at registration only for the course design and
development.
Finally, a synthesis of the personalization criteria used in
earlier research is needed to define the main criteria to
enhance the learner’s engagement in MOOCs.
IV. PERSONALIZATION CRITERIA IN MOOC PLATFORMS
In this section we’ll look more closely at personalization
criteria that a recommender system can use to reduce
dropout rates among MOOC learners.
A. Personalization parameters before MOOCs
In the summit "the European eMOOCs 2014
Stakeholders Summit" the MOOCs were criticized for
lacking from a personalization approach, since "they could
benefit from 30 years of research in distance learning and a
minimum of 10 years in e-learning research "[25]. As e-
Fig. 1: Personalization criteria in MOOCs
learning platforms have progressed in personalization due to
their access to exhaustive information about learners, it Criterion n°1: Personal expectation and learning
included many personalization attributes such as:
motivation
demographic information, knowledge, competence, By learning motivation and personal expectation, we
experiences and preferences in terms of learning style, mean the reasons for a learner subscription in a MOOC.
language and content presentation, etc. In a study of [26], an Ref. [27] found that there are four main motives for signing
approach to normalize and personalize learning situations in up from a student perspective [27]:
e-learning identified the most relevant personalization x Learning about a certain topic, or increasing their
criteria based on standards and norms of distance learning. knowledge, to refresh what they had learned before,
The learner information selected for personalization criteria or learning a ‘‘just-in-time’’ topic that could help
were: personal information about the learner, his them in their work.
preferences in terms of language, type of device and

978-1-5386-2957-4/18/$31.00 ©2018 IEEE 17-20 April, 2018, Santa Cruz de Tenerife, Canary Islands, Spain
2018 IEEE Global Engineering Education Conference (EDUCON)
Page 1268
Authorized licensed use limited to: University of Peshawar. Downloaded on September 17,2020 at 11:28:40 UTC from IEEE Xplore. Restrictions apply.
x Being curious about taking a complete online course Broadly, the learning styles can also reflect “the
(a MOOC). difference in the learning strategies each learner develops
x Having a personal challenge, such as wanting to see if based on his/her previous learning experiences” [15]. In the
they could make it through a prestigious institution AMOL system [15], five learning styles were used:
online course. “apprentice (learning through mentor–student interaction),
x Winning as many course certificates as possible from incidental (learning through case study), inductive (learning
either the instructors or the MOOC providers. through example), deductive (learning through application),
Moreover, a recommender system should “identify and and discovery (learning through experimentation)” [15].
distinguish the learning expectations of the participants Nevertheless, if a learner is confined to a general
before they start to learn in MOOCs” [22]. Hence, it could learning style, it’ll be more of a stereotype approach that
“identify the participants by considering their different doesn’t take into account multiple learning styles that a
learning motivation and learning interest” [22]. learner has according to the field of knowledge of an online
course. Thus, the measure of the cognitive level is
Criterion n °2: Preferences and needs important to adapt the personalization parameters each time
The terms preferences and needs are used in MOOC the learner’s cognitive abilities change. “Cognition level
learning personalization as concepts that include many generally refers to the capacity of cognizing, judging and
attributes. Indeed, it refers to “the information relating to the evaluating the external knowledge, and is related to the
learner socio-cultural contexts (culture, languages, etc.), to practical experience, thinking skills, and storage of
the device preferences (text, audio, video etc.), and to the information of individuals” [22]. Furthermore, a learning
human computer interface preferences (the color choice, cognitive level “is an individual’s consistent approach to
fonts, etc.)” [26]. An extended definition for learner organizing and processing information during learning”
preferences of [27] includes the preference in time of study [32]. “Therefore, it is much more pervasive, stable and deep
to adapt to the biological time of a learner. seated than learning styles” [33].
This criterion is more related to the personalization of
MOOC design since it represents the configuration of Criterion n°5: Level of knowledge
resources not the adaptation of content. Further, it is mostly This criterion refers to “domain specific information that
related to demographic data (gender, age, location, etc.) that contains the knowledge state of the user” [20], unlike the
can be considered as learning facilitators. perspective of [15], [24] and [16] for whom it refers to the
knowledge acquired during the learning situation. In fact, it
Criterion n °3: Learning outcomes / Learning should be “used for taking the learner background when
objectives communicating learning materials” [11] either before
For an outcome-based learning, learning outcomes are subscribing to a MOOC or during the learning process.
specified for a MOOC in its description but could be used to Reference [26] states that it is “the information concerning
match them with a learner needs. Consequently, “a clear what the learner knows and does not know”. However, it
statement of learning outcomes for both knowledge and must be distinguished from competence and experience,
skills is provided” [10] to help a learner achieve specific though some researchers [23] [16] include these concepts as
learning results. Ref. [21] classifies learning outcomes as well. The criterion enables the adaptation of the course
intended learning outcomes (i.e. those that a learner wishes content to reduce the uncertainty level of a learner in a
to get), and achieved learning outcomes (i.e. those already specific domain.
acquired by the learner)”. Finally, this criterion uses “the
information related to the whole of behaviors (or Criterion n °6: Competence and experience
performances) that a learner must be able to demonstrate In a professional scope, experience and competences are
after learning” [26]. strongly linked. Professional experience is a “source of
competence for an individual and is the result of a person’s
Criterion n°4: Learning cognitive style transformation through his work” [34]. Consequently, an
Felder and Silverman’s learning style model (FSLSM) is experience in a field of work is considered a component of a
the most commonly used for personalization [29] [11] [30] person’s competence [34]. However, the competence here
in e-learning and MOOC context because it is “adaptable to refers to an operational competence related to a work
learning differences and individual needs” [11]. FSLSM has situation. In learning, the French Ministry of Higher
four dimensions where every learner has one of the two education and research discerns between three types of
characteristics in each of these dimensions: Active- competence: knowledge, behavioral competence and
Reflective, Sensory-Intuitive, Visual-Verbal and Sequential- operational Competence [35].
Global [31]. Ref. [22] used a similar model by Rayner and Hence, experience reflects operational competence and
Riding [22] for the framework of a MOOC recommender can be defined as “information related to the learner’s past
system where the learning style of a learner is one of the experiences which present the learner’s know-how and the
two cognitive styles in two dimensions: Wholist-Analytic learner’s familiarity with the system on which he/she will
and the Verbal-Imagery [22]. work”[26] [35]. Meanwhile, competence can reflect the
prior knowledge of a learner [16] or his behavioral

978-1-5386-2957-4/18/$31.00 ©2018 IEEE 17-20 April, 2018, Santa Cruz de Tenerife, Canary Islands, Spain
2018 IEEE Global Engineering Education Conference (EDUCON)
Page 1269
Authorized licensed use limited to: University of Peshawar. Downloaded on September 17,2020 at 11:28:40 UTC from IEEE Xplore. Restrictions apply.
competence but is “not related to specific knowledge about example, a learner of the MOOC platform Coursera
or experiences with the domain. It is acquired by the may give information about his profession, level of
learner’s past experiences and is associated with other experience, level of study, top skills, etc. Moreover,
indispensable knowledge related to the operation of the the learner, by consenting to the privacy policy,
system presented to the learner” [26]. So, we suggest enables the site to use Non-Personal Information for
integrating competence and experience in a single criterion performing statistical analyses of the collective
since they are complementary. characteristics and behavior and measuring
demographics and interests on specific areas of the
Criterion n °7: Pace of work site. Whereas it uses Personally Identifiable
Pace of work was in many studies of learner perspective Information to improve and customize the site and the
about MOOCs considered as a major factor for motivation products and services it offers [40]. Consequently, the
[19] [36] [37]. As a learner says about learning in a MOOC: MOOC recommendations made are limited to the
“it’s a new one because the students can progress at their data collected by the platform.
own speed and don’t need to keep up the speed of their class x The personalization criteria proven to be pertinent for
mates” [37]. It refers to “workloads” [10] and the amount of MOOCs up to date in a study of [19] are: working
time spent on a learning activity without having “a fixed pace and level of difficulty [19]. This study was about
timetable for accessing contents” [19]. improving completion rates of MOOCs by identifying
The adaptation of MOOCs by the learning pace can significant aspects of adaptive learning to help
teachers in designing adaptive MOOCs. The research
prevent dropouts directly by customizing the pace of the
was carried out with 475 participants in the first
course to a learner abilities but can also hinder learning
edition of the “Educational Innovation MOOC
since a learner pace of learning changes according to factors Campus” composed of four adaptive MOOCs [19]
that are difficult to define exhaustively (personal who had completed the initial survey about their
circumstances, levels of interest in a domain ,…). characteristics and their perceptions about possible
As a whole, the personalization criteria rely on the data adaptivity options to be included in future aMOOCs.
provided by a learner and the data retrieved from his Therefore, an adaptive recommendation system will be
navigation traces. Using a MOOC platform data such as log more interesting for the dropout problem in MOOCs since it
files, quizzes, knowledge diagnosis surveys,… gives an preserves the openness feature of MOOCs and can help
insight on a learner personal expectations and motivations, learners adapt their learning path. We suggest a
knowledge level related to the MOOC content, learning recommender system that includes different environments
preferences, learning style, etc. However, we’ll need other with a priority order recommendation for the learner [21]. In
sources of data such as the social and professional Fig. 2, a mapping of the approach we intend to develop in
environment of a learner to retrieve information about his our research.
competences and experiences and general learning
objectives.
V. PROPOSED ADAPTIVE RECOMMENDER SYSTEM FOR
MOOC LEARNERS
As we’ve seen in this paper, some researchers rely on
parameters used in e-learning for authoring platforms for
MOOCs to include different learner preferences during the
course construction. These parameters vary from: level of
knowledge and skills [11] [20] to goals and plans of the
learner [11] [21], to even including Date and time [11].
However, considering the current MOOC learner profile,
there are three components to shed the light on:
x MOOC enrollees that are either professionals or those
with a high level of study are the most motivated
learners for MOOCs [38] (72% of Harvardx Platform
and 63%, of MITx Platform: 63% have a Bachelor's
degree or higher [39]). Hence, a learner experience
and competence are important criteria for MOOC
personalization since they represent his professional
abilities and/or future possible professional interests.
x MOOC platforms that use two types of data to
recommend courses adapted to learners’ needs: the
data provided by learners about subjects of interests
and the data collected from log files about the
learners’ navigation on the MOOC’ platform. For Fig. 2: Content Recommendation for Adaptive learning in MOOCs

978-1-5386-2957-4/18/$31.00 ©2018 IEEE 17-20 April, 2018, Santa Cruz de Tenerife, Canary Islands, Spain
2018 IEEE Global Engineering Education Conference (EDUCON)
Page 1270
Authorized licensed use limited to: University of Peshawar. Downloaded on September 17,2020 at 11:28:40 UTC from IEEE Xplore. Restrictions apply.
We aim to enable the learner to follow adaptive MOOCs their profile for a personalized self-regulated learning
by recommending learning resources relevant to his profile. experience. We are working on extracting social network
The recommendation is an instant feedback before during data for the learner profile modeling and construction. This
and after a MOOC completion to decrease dropouts. In this profile will be used in the right type of recommender system
sense, the modeling of the learner profile used for a for MOOC content adaptation.
personalized feedback must exploit not only the data
collected from the learner navigation in MOOC platforms REFERENCES
but also the exponential evolution of data available in his [1] OpenupEd, “Definition Massive Open Online Courses (
social network profiles. These data can contribute to the MOOCs): Version 1.1 dated 12 March 2015”, 2015, [online]
.Available:http://www.openuped.eu/images/docs/Definition_Ma
construction of the learning profile because the behavior of ssive_Open_Online_Courses.pdf
the learner is more natural and less oriented than that on the [2] T. Brahimi and A. Sarirete, “ Learning outside the classroom
learning platforms. through MOOCs”, Computers in Human Behavior, 51, pp.
604–609, 2015.
The professional social network LinkedIn, for example, [3] M. Cisel and R. Bachelet, “MOOC: ce que les taux d’abandon
holds data about more than 500 million of its members [41] signifient”,
and has already launched its e-learning platform for job 2013.[Online].Available:http://blog.educpros.fr/matthieu-
seekers in November, 2016 [42]. In fact, the LinkedIn cisel/2013/06/01/mooc-ce-que-les-taux-dabandon-signifient/
[4] B. Nkuyubwatsi, “ Evaluation of Massive Open Online Courses
complete user profile includes information about the user’s (MOOCs) from the learner’s perspective “, In The 12th
skills, educations, certifications, interests and even the European Conference on e-Learning ECEL-2013, October
courses followed [43]. However, there are many constraints 2013, pp. 30-31.
[5] S. Porter, “ To MOOC or Not to MOOC », In « How Can
to social network data analysis such as the unstructured data Online Learning Help to Build the Future of Higher
and the availability of personal information provided by our Education? “, pp.1-7, Ed.Elsevier, Amsterdam, 2015.
targeted users. Therefore, data mining techniques applied to [6] S. Hayes, “ MOOCs and Quality : A Review of the Recent
learner profiles on social networks databases can enlarge the Literature “,2015.[Online].Available:
http://www.qaa.ac.uk/en/Publications/Documents/MOOCs-and-
scope of MOOC recommendation criteria. Our future Quality-Literature-Review-15.pdf ,Consulté le 14/09/2016.
publications will detail the learner profile modeling to use [7] A. Margaryan, M. Bianco and A. Littlejohn, “ Instructional
for our recommender system that exploits external data quality of Massive Open Online Courses (MOOCs) “,
extracted from MOOC platforms and social networks. Computers & Education, Vol. 80, pp. 77–83, January 2015.
[8] K.S. Hone and G.R. El Said, “Exploring the factors affecting
VI. CONCLUSION MOOC retention: A survey study”, in Computers in Human
Behavior, vol.98, pp.157-168, 2016.
As MOOC platforms evolve to cope with the increasing
needs of online learners, a constant adaptation is valuable [9] QAA, “UK Quality Code for higher education: Part B Assuring
for learner satisfaction and MOOC quality enhancement. and Enhancing academic quality: Chapter B3 Learning and
Teaching”, Gloucester: The Quality Assurance Agency for
Many research works involved several personalization
Higher Education , 2016.
parameters in their proposed Frameworks for MOOC design [10] J. Rosewell and D. Jansen, “The OpenupEd quality label:
and recommendation, but this contribution is to offer a clear benchmarks for MOOCs “, in The International Journal for
overview of what could be a unique gate to MOOCs. Innovation and Quality in Learning, pp. 88-100, 2014.
In this sense, we analyzed the dropout phenomenon to [11] A. Kahaei, “Design of Personalization of Massive Open
Online”, M. Thesis in informatique, Lappeenranta University of
deduce the importance of MOOC customization for a better Technology, Finlande, 2015.
quality of learning. Afterwards, a focus on personalization [12] J.-C. Marty and A. MILLE, “ Analyse de traces et
criteria in literature review was set to show the limits of personnalisation des environnements informatiques pour
l'apprentissage humain “, Traité IC2 : Informatique et systèmes
research in terms of criteria used for adaptive MOOCs and d’information, Hermès Sciences, 2009, In M. Lefevre and al.,
the richness of personalization parameters of learning in “Personnalisation de l’apprentissage : confrontation entre
general and e-learning particularly. besoins et approches”, STICEF, 19, pp.1-23, 2012.
Then, we proposed seven main criteria that summarize [13] D. Clark . “Adaptive MOOCs”, Cogbooks, 2013.
the learner’s source of motivation and choice. These criteria [14] G. paquette, “ Présentation de Mr. Gilbert Paquette – Semaine
2 “,2014.[Online].Available :http://rel2014.mooc.ca/assets/docs/
are: personal expectation and learning motivation, Semaine2_GilbertPaquette.pdf
preferences and needs, learning outcomes, learning [15] N. Sonwalkar, “The First Adaptive MOOC: A Case Study on
cognitive styles, competence and experience and at last: Pedagogy Framework and Scalable Cloud Architecture - Part I”,
pace of work. The combination of these criteria can be a MOOCs Forum, pp. 22–29, 2013.
used in our suggested recommender system model. The [16] A. Teixeira, J. Mota, A. Garcia-Cabot, E. García-Lopéz, and L.
de-Marcos, “A new competence-based approach for
proposed model intends to use MOOC data bases from personalizing MOOCs in a mobile collaborative and networked
different MOOC platforms, to match them with a learner environment” in RIED: Revista Iberoamericana de Educación a
changing characteristics and to respond to the adaptation Distancia, vol. 19 (1), pp. 143-160.
criteria for MOOC recommendation. [17] J. Thompson, “Types of Adaptive Learning”, Cogbooks, 2013.
[18] N. Piedra, J. A. Chicaiza, J. López and E. Tovar, “ An
Moreover, since MOOC enrollees are mostly interested Architecture based on Linked Data technologies for the
in career enhancement, we recommended the use of external Integration and reuse of OER in MOOCs Context”, in Open
data retrieved from professional social networks to extend Praxis, vol.6(2), pp.171–187,2014.

978-1-5386-2957-4/18/$31.00 ©2018 IEEE 17-20 April, 2018, Santa Cruz de Tenerife, Canary Islands, Spain
2018 IEEE Global Engineering Education Conference (EDUCON)
Page 1271
Authorized licensed use limited to: University of Peshawar. Downloaded on September 17,2020 at 11:28:40 UTC from IEEE Xplore. Restrictions apply.
[19] D. Lerís, M. L. Sein-Echaluce, M. Hernandez and C. Bueno, [37] K. A. Tan et al., “Etude des Massive Open Online Courses
“Validation of indicators for implementing an adaptive platform (MOOCs)”, France: Université Paris 13- Master 1 Innovations
for MOOCs”, in Computers in Human Behavior, vol. 72, en Communication, 2014.
pp.783-795, 2017. [38] Zhenghao C. , Alcorn, B. , Christensen, G. et al., “Who’s
[20] N. Bansal, “Adaptive recommendation system for MOOC”, Benefiting from MOOCs, and Why”, Harvard Business
Phd. Thesis, Indian Institute of Technology, Bombay, 2013. Review., 2015.
[21] I. Gutiérrez-Rojas, C. Alario-Hoyos, M. Pérez-Sanagustín, D. [39] A. D. Ho, J. Reich, S. Nesterko, D. T. Seaton, T.Mullaney, J.
Leony and C. Delgado-Kloos, ”Towards an Outcome-based Waldo and I. Chuang, “HarvardX and MITx: The first year of
Discovery and Filtering of MOOCs using moocrank “, open online courses”, in HarvardX and MITx Working Paper,
Proceedings of the European MOOC Stakeholder Summit, No. 1,2014.
pp.50–57,2014. [40] Coursera, “Privacy Policy”. Available:
[22] D. Fu, Q. Liu, S. Zhang and J. Wang, “The Undergraduate- https://www.coursera.org/about/privacy
oriented Framework of MOOCs Recommender System“, in [41] Lesechos.fr, “LinkedIn poursuit son ascension et franchit le cap
2015 International Symposium on Educational Technology, des 500 millions de members”, 2017. Available :
pp.115–119, 2015. https://www.lesechos.fr/tech-medias/hightech/0212012114562-
[23] F. Clerc, M. Lefevre, N. Guin and J. C. Marty, “Mise en place linkedin-atteint-les-500-millions-dutilisateurs-
de la personnalisation dans le cadre des MOOCs,” in 7ème 2082361.php#q2UexZahXPkg73RT.99
Conférence sur les Environnements Informatiques pour [42] Arobasenet, “Linkedin Learning propose des cours gratuits en
l'Apprentissage Humain (EIAH), Agadir, Morocco, 2015. ligne grâce à Lynda.com”, 2016. Available:
[24] D. F. O. Onah and J. Sinclair, “Massive open online courses: an http://www.arobasenet.com/2016/09/linkedin-learning-
adaptive learning framework”, in: 9th International Technology, formation-en-ligne-3354.html
Education and Development Conference, Madrid, Spain, pp.2-4 [43] LinkedIn Developers. “LinkedIn profile field descriptions”,
Mar 2015 (Published in: INTED2015 Proceedings pp. 1258- 2017. Available: https://developer.linkedin.com/docs/fields
1266).
[25] E. Ossiannilsson, “Lessons learned from the Europen eMOOCs
2014 Stakeholders Summit” In C. M. Stracke, U. - D Ehlers, A.
Creelman and T. Shamarina-Heidenreich, “Changing the
trajectory: Quality for opening up education”, pp109-116, Crete,
Greece, 2014.
[26] M. Abik, and R. Ajhoun,” Normalisation et Personnalisation des
situations d’apprentissage”, Phd Thesis in Informatique, Ecole
Nationale Supérieure d’Informatique et d’Analyse des
Systèmes, Rabat, 2009.
[27] K. F. Hew and W. S. Cheung, “Students’ and instructors’ use of
massive open online courses (MOOCs): Motivations and
challenges”, in Educational Research Review vol.12, pp. 45–58,
2014
[28] X. Wei and J. Yan, “Learner profile design for personalized e-
learning systems”, in International Conference on
Computational Intelligence and Software Engineering (CiSE ),
pp. 1-4 Wuhan China: IEEE, 2009.
[29] D. El-Hmoudova , “MOOCs Motivation and Communication in
the Cyber Learning Environment” in Procedia - Social and
Behavioral Sciences vol.131, pp. 29 – 34, 2014.
[30] R. Haddaji, F. Essalmi , S. Hamzaoui and A. Tlili, “Toward the
selection of the appropriate e-learning personalization strategy”,
in “ Innovations in Smart Learning”, Lecture Notes in
Educational Technology (LNET), Springer, 2017.
[31] S. Graf, S. R. Viola, T. Leo and Kinshuk, “In-depth analysis of
the felder-silverman learning style dimensions” in Journal of
Research on Technology in Education vol.40, n°1, pp. 79- 93,
2007.
[32] S. Messick, “The nature of cognitive styles: Problems and
promise in educational practice”, Educational Psychologist
19(2), pp. 59-74, 1984, in A. Kahaei, “Design of Personalization
of Massive Open Online”, M. Thesis in informatique,
Lappeenranta University of Technology, Finlande, 2015.
[33] R. J. Riding and E. Sadler-Smith, “Cognitive style and learning
strategies: Some implications for training design”, International
Journal of Training and Development vol. 1(3), pp. 199-208,
1997, in A. Kahaei, “Design of Personalization of Massive
Open Online”, M. Thesis in informatique, Lappeenranta
University of Technology, Finlande, 2015.
[34] J. Vinces, “Définir l’expérience professionnelle”, in Travail et
Emploi, vol. 85, pp. 21-34,January 2001.
[35] France - Ministère de l’Enseignement supérieur et de la
recherché, “dictionnaire des compétences”, 2011.
[36] M. Zhou, “Chinese university students' acceptance of MOOCs:
A self- determination perspective”, in Computers & Education,
vol. 92-93, pp. 194-203, 2016.

978-1-5386-2957-4/18/$31.00 ©2018 IEEE 17-20 April, 2018, Santa Cruz de Tenerife, Canary Islands, Spain
2018 IEEE Global Engineering Education Conference (EDUCON)
Page 1272
Authorized licensed use limited to: University of Peshawar. Downloaded on September 17,2020 at 11:28:40 UTC from IEEE Xplore. Restrictions apply.

You might also like