Thursday, February 23, 2012

MOOCs and AI-Stanford Course: Examples of Connectivism and Cognitive-behaviourism Respectively.

In a recent paper Anderson and Dron (2011) describe three generations of distance education (DE) pedagogy: cognitive-behaviorist, social constructivist, and connectivist.
They conclude:
  • That all three current and future generations of DE pedagogy have an important place in a well-rounded educational experience.
  • To a large extent, the generations have evolved in tandem with the technologies
  • all three models are very much in existence today.

For each of these models of DE pedagogy they outlined a set of  conditions that characterizes each of them.

During the last years MOOCs  (Massive Open Online Courses) have been carried out with great success. Examples are  CCK08, PLENK2010, MobiMOOC (2011), EduMOOC (2011), Change11, and LAK12. 

They represent an emerging methodology of online teaching. Their structure was inspired by the philosophy of connectivism and the implementation requires conceptual changes in perspective from both “facilitators” (tutors) and learners.
Their subjects ranged from connectivism and connective knowledge (CCK), personal learning environments and networks and knowledge (PLENK), Online learning for today and tomorrow (EduMOOC) to the more technically involved on mobile learning (MobiMOOC).
They share many common features but two different delivery formats can be established. MobiMOOC and EduMOOC for example used a centralized forum and used a Google Group  which made all debates and discussions happen nearly instantly. The others utilized "The Daily", a daily newsletter that harvests new posts from blogs and other online contributions from participants. From the perspective of learning analytics for an external researcher  (non organizer) the former represents a much simpler research environment.

The 2011 AI-Stanford class on Artificial Inteligence taught by Sebastian Thrun and Peter Norvig was also a massive open online course with 160,000 registered enrollees of which  20,000 completed all coursework. In the words of their creators it was "A bold experiment in distributed education". It was offered free and online to students worldwide from October 10th to December 18th 2011. The course included feedback on progress and a statement of accomplishment. The curriculum drew  from that used in Stanford's introductory Artificial Intelligence course. The instructors offered similar materials, assignments, and exams.

If one uses Anderson and Dron to classify the DE pedagogy to these online formats it becomes clear that the AI-Stanford course falls predominantly  into the  cognitive-behaviorist category (with some small components from social constructivism) and the MOOCs  into the connectivist. This analysis is similar to that recently posted in her blog by  Anne Zelenka ( for the Stanford machine learning class by Andrew Ng and the LAK12 MOOC.

Picture 1 shows the number of visitors to the main page of EduMOOC. Although it had nearly 2000 registered participants only 100/200 on average participated after the first weeks.(from Google analytics)
Picture 2 is the daily number of participants in the AI-Stanford class. It translates to around 25,000 number of daily visitors (between 1/3 and 1/4 of those that visit daily Observe that a similar pattern to the is starting to emerge for (from analytics).

Picture 1. Number of daily unique visitors to EduMOOC.

Picture 2. Comparison of number of daily visitors to, and

It is clear that we are in the presence of very different course formats:
  • the AI-Stanford participants have totally different learners goals and preparation than those in MOOCs.
  • there exists a very different nature of the subjects studied: educational theory and engineering.
  • the retention behavior (number and pattern) in both type of the massive courses is totally different and worth researching further. 
  • the AI-Stanford course falls into the cognitive-behaviorist pedagogy category and the MOOCs  into the connectivist.
I have actively participated, mostly with research interest, in 4 different MOOCs (2 from each format described above) and completed successfully the AI-Stanford course. I now joined the 7 weeks "CS101 Building a Search Engine" course that starts 20th of February. This time, sponsored by a spin-off of the AI-Stanford class :

In August 2011 during  EduMOOC, an active discussion emerged as to whether the AI-Stanford class was a MOOC (see my previous posts). I would definitely conclude that the AI-class was not a MOOC.

These are very interesting times to apply learning analytics concepts  for researching these successful massive online courses.

tag for: #lak12 and #change11


  1. Osvaldo, this is a great post. I like figures and diagrams. It would be interesting if somebody could ask the people that did not continue the course what the reason is for that.
    regards Jaap

  2. thanks for sharing this. very interesting.