Dropout Rates of Regular Courses and MOOCs
Léon Rothkrantz
Delft University of Technology, Mekelweg 4, Delft, The Netherlands
Technical University in Prague, Konviktskastreet20, Prague, Czech Republic
Keywords: Dropout Rates, MOOCs, Flip the Classroom, Didactical Models, Blended Learning.
Abstract: Recently we observe an enormous grow of Massive Open Online Courses (MOOCs). But it proves that the
dropout rates of MOOCs are very high. One of the main causes are missing of necessary capabilities of
students, inability of students to manage their study and a missing appropriate didactic model. In this paper
we compare the dropout rates of MOOCs, regular courses and courses using new didactical approaches as
blended learning and flip the classroom. Finally we discuss possible ways how to teach 21st century skills
as cooperative working, learning, creativity, networking and how to solve real life problems in a context
sensitive approach. Our research findings are based on educational experiments at Delft University of
Technology (DUT).
1 INTRODUCTION
In a complex, globalising, unpredictable world full
of networks, worldwide communications and social
media we have to train students in acquisition of
typical 21
st
century skills such as critical reflection,
cooperating, networking, creativity, ability to handle
big data, ability to solve real life problems, ability
for life-long learning. We have to educate students
which are able to face and contribute to the future
world (Bussemaker, 2015). The question is whether
and how educational innovations as “blended
learning” and MOOCs enable students to acquire
such skills. It is not enough to offer courses as
MOOCs but it is also necessary to develop didactic
models to challenge students to take this courses and
complete them successfully (Rothkrantz, 2015).
The process of teaching and learning has an
impact in three domains (Bussemaker, 2015):
Qualification, this is about the role of teaching in
acquisition of knowledge, abilities,
competencies, and attitudes
such that after
graduation students are qualified to perform their
job or play their role in society.
Socialisation, this is the way how we teach
students the social processes in a job
environments or culture and democracy in
general.
Personal development, the impact of teaching on
the process of individualisation and subjectivity
.
The question is how to realise these goals with a
student population of increasing diversity,
differences in cultural background, interest, abilities,
learning style, speed/pacing. Most regular courses
are focussed on realisation of the qualification goal.
By introducing MOOCs and blended learning
courses the hope was that they support socialisation
and personal development. But as we will show in
common used didactic models it is assumed that
students master already 21
st
century skills. But in
most secondary schools these skills have not been
trained, so they have to be learned at University.
From a recent study in the Netherlands
(Bussemaker, 2015) it proves that the teaching
process at Universities is still focussed on
knowledge transfer and qualifications of students
and not on socialisation and personal development.
The concept Bildung introduced by Humboldt in the
19
th
century focussed on the development of all
human abilities and not only some cognitive abilities
and knowledge acquisition is a relevant item.
To realise such educational goals there is a
request of smaller learning communities with intense
interaction students and teacher. But in MOOCs the
direct interaction with a teacher is minimised and is
embedded in the teaching material in a non-direct
way. We observe also a trend of creating honours
programmes challenging and enabling students to
compose their individual self-paced programmes.
Rothkrantz L.
Dropout Rates of Regular Courses and MOOCs.
DOI: 10.5220/0006811600010001
In Proceedings of the 8th International Conference on Computer Supported Education (CSEDU 2016), pages 9-18
ISBN: 978-989-758-179-3
Copyright
c
2016 by SCITEPRESS Science and Technology Publications, Lda. All rights reserved
Concepts as freedom, responsibility, societal
relevance and innovation, are very appealing to
students. But we can observe that only a minority is
able to implement such programmes successfully.
Many students don’t know what they really want,
are not aware of requirements of individual studies
and of the required capabilities. MOOCs have a high
dropout rate because students don’t have the
required capabilities, attitudes and ability to control
their own study behaviour. Student are used to take
precooked programmes and a teacher taking the
supervision on their study. Self-paced studies require
a new study behaviour which has to be learned.
In section 4 we discuss the results of a
psychological assessment at TUDelft. First years
students were tested using the Big Five personality
test to assess if students have the abilities to learn
21
st
century skills (Rothkrantz, 2015). Then we will
discuss the results of a huge experiment performed
at TUDelft from 1953-1957 (Bottema 1959).
Students were tested using personality tests,
(cognitive) ability tests, completed with interviews
of student counsellors and surveys on assessment of
features in the teaching-learning environment. The
goal of this experiment was to research if
psychological assessment could provide additional
information next to the results of final examination
at secondary schools to predict study-success or –
failure. In 2014 an experiment has been started using
new teaching learning models in mathematical
courses for first years students at TUDelft. Again we
will report the study performance of students
compared to traditional ways of teaching
mathematics. To summarize the goals of this paper
are:
Identification of factors/reasons explaining the
high dropout rate of MOOCs.
Comparison of factors underlying different
teaching-learning models varying from
classroom teaching up to MOOCs, with respect
to study -success or –failure.
Our research findings are based on results of
experiments performed at Delft University of
Technology.
2 LITERATURE SURVEY
In (Onah et al., 2014) the authors researched MOOC
drop rates from different perspectives. They listed a
number of reasons for dropout based on literature
search. Next they researched data from a specific
MOOC provided by the University of Warwick.
Their results indicate that many participants who
may be classified as dropouts are still participating
in the course in their own preferred way, either with
a slower pace or with selective engagement .
In (Yang, 2013) the author explores students
dropout behaviour in MOOCs. As a case study they
took a special class. He developed a survival model
that allows to measure the influence of factors
extracted from that data on student dropout rate. His
study shows that specific social factors as interaction
between students, emergent sub community
structure affect dropout behaviour
In (Sinha, 2014) the author researched the
underlying interaction mechanisms which govern
students’ influence on each other in Massive Open
Online Courses (MOOCs. Specifically, they outlined
different ways in which students can be negatively
exposed to their peers on MOOC forums and discuss
a simple formulation of learning network diffusion,
which formalizes the essence of how such an
influence spreads and can potentially lead to student
attrition over time.
In (Ye and Biswas, 2014) the researchers
extended traditional features for MOOC analysis
with richer and higher granularity information to
make more accurate predictions of dropout and
performance. The results show that finer-grained
temporal information increases the predictive power
in the early phases of the Pattern-Oriented Software
Architectures (POSA) tested on a MOOC offered in
summer 2013 by Vanderbilt University.
Prediction of study success is a popular research
topic over the years. (Wilbrink, 1997) shows a
literature survey of assessment in historical
perspective with more than 200 reviewed papers.
In (Tinto, 1975) ia theoretical longitudinal model
has been introduced, demonstrating how different
personality characteristics and characteristics of the
University have their impact on the interaction
process student-University. Such a process resulted
in delay, dropout or study success. In fact Tinto
describes the interaction process between three
systems, the individual student, the academic system
and the social system. According to Tinto the
individual student will enter the University with a
specific social background, personality education
and training. He will have a special binding with the
University and goal of the study. This binding will
be expressed in motivation and expectations. The
binding will change over time, caused by the
influence of the academic and social systems and the
interaction of both and will eventually result in
dropout.
A similar study has been performed in Belgium
at the Universities of Ghent and Leuven. Researcher
Lacante and Janssen were involved in research on
prediction of study-success for a long time (Lacante,
1981). They developed special surveys to predict the
academic performance at Universities. They found
similar results as the researchers at Delft University
with some refinements.
The didactic Adagio of the famous Dutch
mathematician and didactic specialist in
mathematics Freudenthal (Freudenthal, 1973) was
“You can learn mathematics only by doing and
discover mathematics in the real world”. For him
was teaching mathematics an educational task and it
should be context sensitive and application oriented.
Students should be able to design mathematical
models and translate real world problems formulated
in natural language in a mathematical language. It is
very important to give students opportunities to
reflect on and clarify their thinking about
mathematical ideas. Most of current didactic models
fits in the discovery learning tradition developed
around 1960. Piaget, Dewey, Vygotsky and Freire
and many others support constructivist learning. Up
to then drill and practice was one of the favourite
pedagogical principles in mathematics. Now the
focus is on learning based on personal and societal
experience. Our developed didactic model FETCH
2.0 is based on similar ideas (Rothkrantz, 2015). The
question is of course how to implement this didactic
model in the developed MOOC. The oldest, and still
the most powerful, teaching tactic for fostering
critical thinking is Socratic teaching. In Socratic
teaching we focus on giving students questions, not
answers. The next step is that students themselves
learn to generate questions around a learning text
(Rothkrantz, 2015).
Mathematicians are trained to ask (critical)
questions reading a scientific journal. These
questions are stimulated by the learning material but
also by the surrounding environment and context.
Developing a critical attitude by students is not
limited to mathematics. Freudenthal writes about
“Mathematics as an educational task”. Many
mathematicians use the inquiry based methods also
during reading or reviewing a paper, or documents
or listening to a presentation. It proves that inquiry
based method is an efficient way to keep the
reader/listener alert and is a first step to processing
the presented information.
3 REASONS FOR DROPOUT
In this section we discuss possible causalities of the
high drop off rate of MOOCs. TUD launched 20
MOOCs, accessible via edX platform. In this section
we focus on the MOOC Pre-Un Calculus, visited by
more than 26.000 students. Via interviews and
questionnaires experiences of students and teachers
are assessed and reported in (Vos, 2015). Most
(early) dropouts from the course didn’t take part in
the assessment procedure, so we realise that the
results are biased.
MOOCs are in principle open for everybody. No
entrance exam has been required. But some
knowledge and capabilities are required to finish a
MOOC successfully. This is stressed in information
about the course but some students still believe they
can do it. Similar experiences we have with regular
students at DUT. No entrance exam is required to
enter TUD. The first year of study should be used
for selection, orientation and adaptation. It proves
that low grades for mathematics and physics at
secondary school are good predictors for dropout.
Over the years the failure rate of regular academic
courses has been studied. We will report about a
study at Delft University of Technology with a
psychological assessment of all freshmen. It proves
that results of students at their secondary school
predict about 40% of the study success or failure in
the first year. Psychological assessments ads about
10%.
To complete a MOOC successfully a student
should have a strong motivation, based on the
expected outcomes, interest in science and increase
of knowledge and competences. Next a student
should be able to manage his time and plan a study.
In the current MOOC a global time schedule of
video lectures, simulations, exercises and exams is
presented. To enable communication and
cooperation between students all students are part of
one group taking the course together. In more
individual based schedule students will lack support
of fellow students. Many students reported that they
prefer to manage their own course without
cooperation. This violates of course the goal to
realise 21
st
century skills. Individually based studies
are offered for many years by the open University.
But they also offer meetings for participants to
create a community. Given a strong motivation and
ability to plan their study, students should be able to
come to activities. Viewing video lectures and
simulations don’t require active participation of
students. It may give them the feeling that they have
everything under control. They are confronted with
reality if they have to make assignments or exams.
Inadequate time management, un-ability to set
themselves into action and missing of other study
skills, results in many cases to dropout.
An adapted didactic model for MOOCs is
needed. Recorded web lectures of gifted teachers or
successful lectures are no guarantee of high study
success of MOOCs. In many cases the didactic
model of regular courses in lecture halls is copied. A
teacher explains the theory, gives some examples
during the course and students are supposed to study
the learning material and to make exercises. But
making homework is a great problem. The gap
between lectures and making exercises is (too) big
and most students are not able to manage their time
and set themselves into action.
It can be observed at TUDelft that a week before
the exams students start to visit libraries, study-halls
etc. to start their preparation for exams. Observing
this study behaviour of fellow students triggers other
students to start similar behaviour. It is difficult to
implement similar triggers in the MOOC
community. But we have to realise that used didactic
models assuming regular study behaviour of
students in general don’t work for MOOCs. In
regular students are creative in finding alternatives
observing peer models what is still missing in the
MOOC community.
Personal, intensive study guidance results in
better study results. A good binding with the study
and study community is an essential prerequisite for
successful study. Unfortunately the massive
character of MOOCs makes individual tutoring by
teachers impossible. One of the assumptions of
creating a study community via social media is that
students will support each other. But it proves that
creating study communities is far from trivial. One
of the outcomes of the surveys is that many students
don’t like networking and cooperation and prefer to
study in their own individual way. Students get
stronger involvement with the study and study
community.
To be a member of a learning community has a
great impact on the motivation, emotions and study
results/achievement. (Furrer and Skinner, 2003). It is
important to give students the feeling that it is
exceptional to take these courses and that they
should be involved in learning communities. In case
of MOOCs it cannot be expected that there is much
individual support or supervision of teachers. Fellow
students can take the role of a teacher and the
community of fellow students should support and
stimulate the members of the community.
Many students following MOOCs don’t live in a
student environment. Via social media they can be
involved in learning environments. Participation in
such an environment is important to develop 21
st
century abilities. MOOCs offer the opportunity for
internationalisation, integration of research in the
learning environment. In research based education
students are involved in common research activities
and in research tutored education students perform
research activities themselves.
Students following MOOCs have to take the
responsibility for their own learning process. It is
definitely true that part of the students are able to
manage their study. But most students need group
pressure, supervision of teachers, counsellors, fellow
students to complete a MOOC successfully.
Nevertheless many educators prefer self-paced
instruction so that every student is able to define his
individual study path. This enables students to
accomplish their study at their own speed. But it
creates problems to give support from the
community. In the Netherlands some PhD students
have to follow MOOCs instead of common courses
at some University. This is highly appreciated by
PhD students and the success rate is high. After
completion of a course PhD students are supposed to
do an exam at the home University.
4 EXPERIMENTS
4.1 Big Five Personality Test
One of the research goals of this paper was to
investigate how to teach students 21
st
century skills
as networking, cooperation, creativity etc.
Traditionally teaching mathematics is rather
individually oriented. A lecturer introduces students
in lecture halls into new topics. Students listen and
make notes. Parallel to the lectures are usually lab
hours during which time students are supposed to
make exercises alone or together. Especially during
the lectures students are passive consuming
knowledge, instead of exploring new topics and
discover new knowledge and even more important it
is an individual process. It takes students less energy
to listen to lecturers and see how he discovers new
knowledge. But the challenge is to take students out
of their comfort zone and challenge them to actively
search for knowledge and problem solving.
Changing the didactic approach may change the
learning attitude of students. It can be expected that
there will be resistance leaving the comfort zone. A
second question is if students have the right
personality characteristics to learn the 21
st
abilities.
If students are not able to cooperate, socialise and
network there is a problem. To research that problem
a group of first years students in mathematics and
computer science were tested using the Big Five
Personality test in September at start of the academic
year. Students were supposed to fill in a
questionnaire of 4x5 questions. Then the score on 5
factors E, A, C, N, O listed below were computed.
Students with high scores on the factors A and O are
supposed to be open for learning the 21
st
century
skills. But it proves that students score significant
lower on A and higher on O. This supports the
hypothesis that students in technology or exact
courses have the tendency to work individually and
are open for new knowledge. The factors underlying
the Big Five Personality test can be described as
follows:
Extroversion (E) is the personality trait of seeking
fulfilment from sources outside the self or
in community. High scorers tend to be very social
while low scorers prefer to work on their projects
alone.
Agreeableness (A) reflects much individuals adjust
their behaviour to suit others. High scorers
are typically polite and like people. Low scorers tend
to 'tell it like it is'.
Conscientiousness (C) is the personality trait of
being honest and hardworking. High scorers
tend to follow rules and prefer clean homes. Low
scorers may be messy and cheat others.
Neuroticism (N) is the personality trait of being
emotional.
Openness to Experience (O) is the personality trait
of seeking new experience and intellectual
pursuits. High scores may day dream a lot. Low
scorers may be very down to earth.
Table 1: Average values of the five factors of the Big Five
test E, A, C, N, O for a cohort of TUD students computer
science and mathematics and a general cohort.
Number of respondents 10189 179
Extroversion 3.05 3.05
Agreeableness 3.84 3.69
Conscientiousness 3.38 3.06
Neuroticism 2.98 3.39
Openness 4.05 3.71
From Table 1 it can be observed students
mathematics and computerscience have the same
average score on the factor Extroversions indicating
that in principle they are open/not open for social
contacts and open/not open for networking.
The students in the TUD cohort score lower on
the factor Openness and less open for new
experience and intellectual pursuits. But we may
conclude that students in exact/technical sciences are
in principle open for learning 21
st
century skills.
4.2 Psychological Assessment
Procedure at TUDelft 1953-1957
For most academic studies at Universities in the
Netherlands there is no entrance exam. Only a
limited studies with capacity problems have a
special admission procedure. Students with low
grades at the final exam at their secondary school
have a lesser chance to pass this special entrance
procedure than students with higher grades. As a
consequence many students start a study with
insufficient intellectual capabilities. The assumption
is that Universities have the opportunities to select
the students for the higher years based on their
academic performance in the first year. It is a well-
known saying that students with limited intellectual
capabilities can compensate this shortage by hard
working. To research the impact of personality
characteristics, motivation, social environment on
academic performance and study-success or –failure
a huge assessment procedure was performed in 1953
at Delft University of Technology. All students in
the first, second and third year got a psychological
assessment. They were supposed to fill in
questionnaires corresponding with test assessing
cognitive abilities, and personality. Next all students
were interviewed by student counsellors. The
outcomes of all these assessments were correlated
with the results of exams of students and results of
students at their secondary school. It proves that
there was a significant correlation between academic
performance at the University and final results at
secondary school for the subjects mathematics and
physics. Not a big surprise taking into account that
the subjects mathematics and physics play an
important role in technical studies at Delft
University of Technology. From table 2 we can see
that students with lower grades for the subjects
mathematics and physics have a low study progress
in the first year and students with high grades have a
high chance of study success. For the middle group
of about 50 % it was difficult to predict study
success or –failure. A disappointing result was the
fact that the outcome of psychological assessment
has a very limited added value to the prediction of
study-success or –failure. Apparently the impact of
personality, social environment were already
included in the performance at secondary school.
Table 2: Study-progress/delay/dropouts in percentages
crossed with average score mathematics/physics grades at
school, after two years of study, cohort 1953.
Studyprogress
150%
0% 4% 11% 15% 7%
Studyprogress
150%
0% 8% 13% 4% 0%
Delayed students
with Incomplete
first year
2% 2% 1% 1% 0%
Dropouts during
second year
2% 6% 2% 0% 0%
Dropouts during
first year
2% 9% 9% 1% 0%
Average math/
physics grades at
school exam
5-6 6-7 7-8 8-9 9-10
Table 3: Study-progress/delay/dropouts crossed with
number of passed exams in the first exam-session.
Studyprogress
150%
0% 1% 7% 14% 16%
Studyprogress
150%
3% 8% 8% 4% 3%
Delayed students
with Incomplete
first year
2% 1% 1% 0% 0%
Dropouts during
second year
2% 6% 1% 1% 0%
Dropouts during
first year
6% 9% 5% 0% 0%
Numbers of exams
passed successfully
in first period
0 1 2 3 4
From table 2 can be observed that the dropout
rate in the first two years is 31%. After two years of
study the dropout rate was 31%. In total the dropout
rate was 43%. About half of the dropouts started
another study at TUD or at a Polytechnique School.
The function of the first year was orientation and
selection. But still 10% of the students started a
second year and dropout during that year and
additionally 12% of the students drop out later in the
study. At this moment the University spend a lot of
time and effort to stimulate students to take their
decision of dropout as soon as possible. At this
moment it is in general not possible to start the
second year before completing the first year. There
are special programs and even a special MOOC
enabling student to get a better orientation of the
study.
A special problem with presented tables is that
some students enrolled in the study but never show
up. In the early days even no-show students could
get a student loan for the whole year. So no-show
behaviour has financial consequences. A similar
problem can be observed in the cohort of MOOCs
students. In the past the group of slow starters made
a successful start in the second year. The phenomena
of restarting students is not observed in the cohort of
MOOCs students. One of our recommendations for
MOOCs designers is to take care of slow starters,
students who need a longer adaptation time or to
apply the spiral principle of in increasing
difficulties.
Students with a low respectively high average
score on the subjects mathematics/physics have a
low chance/high chance to complete the study
successfully. But students with modal scores are
rather unpredictable. The hope was that
psychological personal improved prediction of
study-success or failure for the middle group.
Unfortunately this was not the case.
From Table 3 can be observed that dropouts miss
more than half of the exams already in the first
session. Some students with bad results in the first
year give themselves second start but have again bad
results in the first exam session. The phenomena of
restarting students is not observed at MOOCs.
4.3 Pre-University Calculus
July-September 2015
TUDelft designed a special MOOC called Pre
University calculus. The goal of the MOOC was to
refresh mathematical knowledge of students before
they start their academic study at the University and
to teach students missing topics of the mathematics
at secondary schools. The teaching material was
composed of small, 10 minutes video fragment with
a lot of simulations, video lectures, gaming etc. The
focus was on applications of mathematics.
Figure 1: Screenshot of the Pre-Un. Calculus MOOC.
Students were stimulated to cooperate by
interacting via special blogs. Students are stimulated
to define questions stimulated by the lectures and the
learning material. Fellow students in the network are
invited to answer the questions and commenting
solutions in the course forum. This is common
practice in mathematics learning.
Figure 2: Topics of the Pre-Un. Calculus MOOC.
In 2015 more than 26.000 worldwide followed
the course and via data analytics the performance of
students was analysed. We will report some main
findings from participating students, first students
from TUD and after that students from all over the
world. Aspirant students from TUD were stimulated
to take part in the special MOOC. Usually lectures
were offered during the summer holidays after the
school exam and just before the start of the academic
year. The MOOC was supposed to attract more
students because there was no need to come to the
University and also no need for teachers to lecture to
lecture during the summer holidays.
In total 794 TUD students enrolled in the course,
420 (53%) of them attempted an exercise and 40
(5%) attempted all exercises. Only 46 students
passed the final exam successfully. From the
interviews it proves that most students had the
feeling that they master the topics already. Just
before the start of the academic year, other activities
got a high priority as finding housing in Delft,
increasing the income by performing a job or just
taking holidays after passing the school exam
successfully. Students didn’t feel the need to
participate in the MOOC.
From table 4 it can be observed or computed that
there is no significant difference in mathematics
grades between students that enrolled and students
that did not enrol.
In total 27.186 students enrolled in the course,
4.150 students attempted to make an exercise and
only 273 attempted all exercises.
Table 4: Distribution of students who enrolled and didn’t
enrol in the MOOC.
Grade math at school exam
5 6 7 8 9 10
# of students that
did not enrol
101 764 1017 703 332 52
Number of
students that
enrolled
32 173 200 166 63 10
All students were requested to fill out some
questionnaires. We report the results from students
who showed some activity in the course. About 59
% of the students took part in the course from begin
to the end, 18 % browsed through the topics and
videos and 4% mainly watched the videos, 7%
mainly did the exercises.
Students on average stated they rarely
participated in study groups, connected to other
students, posted a comment or a question, or looked
at the forum. They also rated their perceived support
from either other students or staff rather low.
However, they did not feel that feelings of
loneliness or missing interaction negatively affected
them in the course. Neither did they feel that lack of
feedback negatively affected them. Rather, they
rated the feedback as good. It seems, students in
general neither expect nor need a lot of interaction
with either students or staff. We stress that this only
holds for students taking part in the course. From
students who didn’t take part or dropout in an early
stage we have no information.
4.4 Experiment Mathematics at TUD
during 2015
It was decided that teaching mathematics at TUDelft
will change dramatically from 2015 on. The dropout
rate had to be reduced. There should be more focus
on applications of mathematics, self-discovery,
teaching 21
st
century skills such as networking,
cooperation via social media etc. In a first
experiment 370 students studying Civil Engineering
got mathematical courses new style. First the
didactical principle flip the class room was applied.
For many years students got their lectures
mathematics in big lecture halls followed by making
exercises in small classrooms. Now the order has
been changed. Students are supposed to study video
lectures and make exercises before they meet the
teachers and fellow students in small classrooms to
discuss problems and outcomes of the exercises. In
the video lectures there was a focus on applications
of mathematics, self-discovery activities of students.
They are stimulated to cooperate in study groups.
The MOOC Pre-Un Calculus was integrated in video
lectures. Many new videos were added especially on
simulation and applications of mathematics.
We will now report some of the results from
surveys and interviews. It proves that on average
only 20% of the students viewed the video lectures
before the classroom meeting. So the assumption
that students activate the right pre-knowledge was
not correct. That poses a didactical problem for the
teacher. Repeating or summarising the homework is
boring for students making their homework and will
not stimulate viewing video lectures in the future.
One of the critical comments of students was that the
video lectures are interesting but are not direct
integrated with the learning material/book. Viewing
the video lectures maybe will result in a better
understanding of mathematics in general but it is not
necessary to pass the exam successfully. One of the
goals of the new mathematics teaching was to focus
on real applications. Then the applications should be
part of the exam and not additional. In the
compulsory homework assignments it is possible to
make links with some topics discussed in the video
lectures.
At start about 80% of the students visited the
classroom meeting and at the end it was about 65%.
The classroom meetings were highly appreciated.
Especially the moments the teacher was lecturing or
explaining difficult topics. It proves that teachers
were better in raising questions or pointing to typical
problems. Individual or group wise making
assignments during the classroom lectures was
highly appreciated. Because there was a meeting
place for the students in the classroom there was no
need to cooperate via social media. Teachers were
asked if and how many times during a lecture they
provide opportunities for students to discuss topics
or cooperate. In the next table we summarise the
results.
Table 5: Overview of opportunities teachers offered for
discussion or cooperation.
Opportunities
for discussion or
cooperation
Never 1 time
2-3
times
4 or
more
times
11% 22% 34% 33%
Teachers like to teach and play a central role
during the lessons. Many students come to the
lessons and expect that the teacher lectures. Parallel
to the classroom meetings there was a digital Lab
room (My.MathLab). Students were supposed to
make assignments and got feedback in an automated
way. It proves that many students joined their effort
and meet each other to make assignments.
An interesting option for students was to provide
feedback during the lectures using their smart
phones or laptop. One of the start up companies
from DUT, FeedBackFruits developed a tool for
mobile devices which enables students to ask
questions during a lecture. The questions are
visualised on a display in front of the teacher. It
proves that students consider the tool as an
interesting option. But defining questions takes some
time and usual the lecture is going on. It is up to the
teacher if he introduces breaks to allow and discuss
questions. Some questions can be used by the
teacher to summarize a topic. If there are many
questions about the same topic the teacher has the
option to explain the topic in an different way or to
come up with some examples. To support the
didactical approach “flip the classroom”, the start up
company FeedBackFruits was requested to generate
a plugin to make the “questioning tool” available via
edX, one of the MOOCs consortia. A layer of new
functionalities was developed over the edX platform.
This enables students to make specific notes inline
and make digital notes out of it. The plug-in also
allows users to add new content to the course and
share a message information board way questions of
the students can be considered as an online feedback
system for the teacher.
5 CONCLUSIONS
The main goals of this paper was to study the high
drop off of MOOCs, to find possible causes and to
develop a new didactical model. Secondly we want
to invest in how far if it is possible to teach students
21st century skills. We used the data from two
experiments at TUDelft.
In the first experiment, TUDelft designed a
special MOOC called Pre University Calculus. The
goal of the MOOC was to refresh mathematical
knowledge of students before they start their
academic study and to teach students missing topics
of the mathematics teached at secondary schools.
The teaching material was composed of small, 10
minutes video fragment with a lot of simulations,
video lectures, gaming etc. The focus was on
applications of mathematics. Students were
stimulated to cooperate by interacting via special
blogs. In 2015 more than 26.000 worldwide
followed the course and via data analytics the
performance of students was analysed. Only a
minority of 5% of the students completed the course
successfully. These students stated in surveys and
interviews that they prefer to work individually and
not in groups or digital community. Unfortunately
the early dropouts were not surveyed. From the
questionnaires and interviews we found many
reasons for dropout behaviour. From the available
data can be concluded that there were only a few
network activities and cooperation between students
via social media. Teaching students 21
st
century
skills will not take place automatically by using
MOOCs but special didactic models are needed.
Secondly it was decided that teaching
mathematics at TUDelft will change dramatically
from 2015 on. There should be more focus on
applications of mathematics, self-discovery,
teaching 21st century skills such as networking,
cooperation via social media etc. In a first
experiment students studying Civil Engineering got
mathematical courses new style. First the didactical
principle flip the class room was applied. For many
years students got their math lectures in big lecture
rooms followed by making exercises in small
classrooms. Now the order has been changed.
Students are supposed to study video lectures and
make exercises before they meet the teachers and
fellow students in small classrooms to discuss
problems and outcomes of the exercises. In the video
lectures there was a focus on applications of
mathematics, self-discovery activities of students.
They are stimulated to cooperate in study groups. It
proves that less than 20% prepared the lessons by
viewing the video lectures in advance. Most students
reported that lack of time and motivation was the
main reason. And they expect that the teacher will
summarise the main topics in the lessons so that they
are able to follow the lessons. Following the lessons
is important for the students because they expect the
teacher will provide essential information needed to
pass the exam successfully. Students cooperated via
the digital Lab making homework together. In the
lessons there was less cooperation also because the
teachers didn’t provide the opportunity. During the
lessons and video lectures students were able to give
comment or asking questions. This provides
essential feedback for the teacher and information
for the evaluation of the course. Again we to
conclude that blended courses don’t guarantee that
students learn 21
st
century skills.
Students mathematics and computer science got
a psychological assessment using the Big Five
personality test. From the results it proves that
students have the abilities to learn the 21
st
century
skills but this will not happen automatically. A
special didactic model is needed.
A comparison was made between a huge
assessment study at TUDelft during 1953-1957 and
recent experiments at TUDelft. It proves that over
the years 40% of the students dropout. We reported
many causes based on surveying and interviewing
students.
As a final conclusion we state that also regular
courses have high dropout rates varying around
40%. Many attempts to reduce the high dropout rate
were not successful over the years. MOOCs show
even higher dropout rates and we conclude from the
outcomes of surveys and interviews that lack of
cooperation in networking, lack of social control of
peer students and inability to manage the study were
the main causes of high dropout rates.
ACKNOWLEDGEMENTS
We thank Ingrid Vos providing me with the results
of evaluation of the experiments at TUDelft. We
thanks the colleagues of the FETCH project for their
valuable help.
REFERENCES
Bottema, O., Bakker, H. Th., 1959. Mislukking en
vertraging van de studie. Verslag van een onderzoek
verricht aan de Technische Hogeschool te Delft 1953-
1957(Dutch). Delft: Technische Hogeschool. 106
pp.32. HOGESCHOOL DELFT.
Bussemaker, J., 2015. Strategische Agenda Hoger
Onderwijs en Onderzoek 2015-20125 (Dutch),
MINISTERIE VAN ONDERWIJS< CULTUUR EN
WETENSCHAPPEN.
Cheng Ye, Gautam Biswas., 2014. Early prediction of
student dropout and performance in MOOCs using
Higher Granularity Temporal Information, In Journal
of Learning Analytics, 1 (3), 169–172.UTS e-PRESS.
Datcu, D., Rothkrantz, L.J.M., 2007. Multimodal web
based system for human emotion recognition, In
Proceedings ISC2007, pp. 91-98. EUROSIS-ETI.
Datcu, D., Rothkrantz, L.J.M., 2010. Assessment of
emotion states during e-learning. In Communication
and Cognition, 43 (1-2).UNIVERSITY OF GHENT.
Freudenthal, H., 1973 Mathematics as an educational
task. Springer Science & Business Media.
Furrer, C., Skinner, E.A. 2003. Sense of relatedness as a
factor in children’s academic engagement and
performance. In Journal of Educational Psychology,
95, p. 148–162.APA.
Lacante, M., 1981. Van intelligentie,
persoonlijkheidstudiestrategie en studeergedrag naar
studieresultaat (Dutch). KATHOLIEKE
UNIVERSITEIT LEUVEN.
Onah, D.F., Sinclair, J., Boyatt, R., 2014. Dropoutrates of
Massive Open Online Courses. In Behavioural
Patterns, 6th annual International Conference on
Education and New Learning Technologies Edulearn
14, Barcelona. IATED.
Rothkrantz, L.J.M., 2015. How Social Media Facilitate
Learning Communities and Peer Groups around
MOOCS. In International Journal of Human Capital
and Information Technology Professionals, 6(1), pp.1-
13. IGI.
Rothkrantz, L.J.M., 2009. E-learning in virtual
communities. In Communication & Cognition 42, Nr
1&2 (1), pp.37-52. UNIVERSITY OF GHENT.
Rothkrantz, L.J.M., 2010. A training tool for non-Verbal
communication. In Proceedings of The International
Conference on E-Learning and The Knowledge
Society, pp. 88-93, UNIVERSITY OF RUSE.
Rothkrantz, L.J.M., 2015, Inquiry based learning as
didactic model in distant learning. In International
Journal on Information Technologies and Security, 7
(4). DOBI PRESS EOOD.
Wilbrink, B., 1997. Assessment in historical perspective.
In Studies in Educational Evaluation, 23, 31-
48.ELSEVIER.
Tanmay Sinha, 2014. Who negatively influences me?
Formalizing diffusion dynamics of negative exposure
leading to student attrition in MOOCs. In LTI Student
Research Symposium, CARNEGY MELLON.
Tinto, V., 1975. Dropout from Higher Education: a
theoretical synthesis of recent research. Review of
Educational research, 1975, vol. 45, pp 89-125.
SAGE PUBLICATIONS.
Vos, I., 2015, Evaluation Pre University calculus MOOC.
TUDelft PRESS.
Yang, D., Sinha, T., Adamson, D., & Rosé, C. P., 2013.
Turn on, tune in, drop out: Anticipating student
dropouts in Massive Open Online courses. In
Proceedings of the 2013 NIPS Data-Driven Education
Workshop. NIPS.