Avec les collègues du labo, nous serons dans deux semaines à la conférence eMOOCs 2015 organisée à Mons par l’Université Catholique de Louvain (en particulier par Marcel Lebrun). Ces trois jours sont l’occasion d’échanger avec les collègues à l’échelle européenne surtout, même si certains collègues d’outre-Atlantique viennent en général nous rendre visite. Les organisateurs nous ont proposé d’aligner le format de la conférence sur la philosophie des MOOC en « flippant » la conférence. On présente rapidement notre papier à travers une vidéo, afin de consacrer le temps d’échanges à des interactions plus fructueuses. Ils ont poussé le concept jusqu’à créer un vrai MOOC sur edX sur le thème « ConfX EMOOCs 2015 Open Conference« . Je suis clairement meilleur en conf que dans cette vidéo (en anglais, désolé), mais je vous la mets dans ce billet pour vous donner une idée de ce qu’on a pu faire avec Mattias, un collègue de l’ENS Cachan, et pour illustrer ce concept de conférence scientifique inversée qui est vraiment super !
One of the most striking consequences of Massive Open Online Courses openness are on one hand their low completion rates, and on the other hand the high heterogeneity of enrollees in terms of background, intentions, and motivations to register. What are the factors that impact course completion and engagement in the MOOC ? Several studies have tried so far to describe the relationship between learners characteristics and their engagement in the course, but most have failed to answer an important question: how robust are the results? Let us say we find out that employment status has a strong impact on completion rates in the first iteration of a MOOC. Maybe we should try to check whether it is still the case in the following iterations before trying to interpret such a result.
In that purpose, we analyzed two five weeks MOOCs whose data were available for three or more iterations. We would like to address the question of the evolution of learners profiles and course dynamics over time. The first is an entrepreneurship course called Effectuation, and the second is a project management course called Gestion de Projet. In both cases, participants had to submit quizzes and to pass an exam to earn a normal certificate. In the project management course, they had to hand over assignments to get an advanced certificate. Data on background, intentions and motivations were gathered through online surveys, whose response rates ranged from 40% to 60% of enrollees. We tried to assess to what extent engagement patterns and registrants profiles and the relationship between those two had evolved over time. We first categorized participants based on their level of engagement :
- Those who obtain a certificate are called “Completers”
- Those who submit at least one quiz or assignment but do not “complete” the course are referred to as “Disengaging learners”
- Those who do not submit any quiz or assignment but they view more than 10% of available lectures are called “Auditing learners”.
- Finally “Bystanders” are the enrollees who fail to submit any quiz or to watch more than 10% of available lectures
The two courses followed opposite trajectories; the number of enrollees decreased over time in the entrepreneurship course while it rose from no more than 3.500 to almost 15.000 in the project management course. In both cases, we noticed that the proportion of bystanders tended to increase over time, from 24 to 53% in the case of the project management course. In the same time we observed a sharp decrease in the proportion of completers.
Nevertheless, those completers were responsible for most of the course activity, whether we look at it from the lecture consumption or the quizz submission point of view. We noticed that auditing learners were responsible for a neglectable share of the course activity, unlike disengaging learners, whose share in the course activity was quite significant, in particular in the last iterations of the project management course.
We carried out a survival analysis in order (to assess) to better understand the behavior of those two categories of registrants. We report here the result for the video consumption of the entrepreneurship course. We notice that auditing learners explore only a fraction of the course material. Actually, in the three iterations of the MOOC, less than 10% of them are still active after the first week. The decline is less obvious among disengaging learners, but only 10% of them watch all the course material. Analyzing the different iterations of a given course could provide valuable insights into the evolutions of MOOCs audience and dynamics on the long-term, at a global level or at the course level. In that purpose, we need to design comprehensive studies focused on these long-term dynamics in order to get more reliable results and more robust conclusions. Standardization efforts in survey design need to go beyond questions about motivations to register, and focus for instance on parameters such as time constraints, experience in online learning, and intentions. Importantly, cross-correlations between survey and log data should be more systematic; common identifiers are often absent, impeding interesting discoveries.
The amount of studies providing valuable insights into MOOC complex dynamics is growing at a fast pace. The current issue is now to interpret such results. Why and how do sociodemographic variables impact engagement in the course? What does it mean from the self-directed learning perspective? Are we witnessing the rise of a new generation of self-directed learners or merely a temporary hobby of already experienced online learners? We hope to meet you at the conference to discuss with you those fascinating questions !