A key point of this paper is that the massive open online course (MOOC) was based on an existing face-to-face MATLAB course delivered over 15 years to some 2,500 students (very similar to my experience). The authors believe this is key to the success of the MOOC; while this hypothesis cannot be tested, it seems plausible. However, they did not use all of the material from the face-to-face course online.

The second main success factor the authors cite is the autograder used for the (six) homework assignments used during the course. The authors stuck to a rigid schedule for the MOOC, enabling them to reveal (and doubtless discuss) the solutions to the homework in line with the schedule. This insistence on a schedule seems appropriate: again, face-to-face programming courses benefit from this.

The third main factor is the mentors, who seem to have appeared spontaneously from the cohort of students. This is apparently reflected in the students’ evaluations of the course.

The evaluation is largely of the 2015 (free) instantiations of the MOOC, rather than the later Coursera instantiations. Eighty thousand students watched the first video, and 18,000 the last, so by that measure the course had a 22 percent completion rate--very high by MOOC standards. Conversely, this course was seriously assessed by nontrivial homework exercises. The authors believe that 66 percent, the passing grade, is the equivalent of a B in their face-to-face course. Fifty-three hundred students achieved this, which equates to a 6.6 percent pass rate--well within the usual range, and probably good given the relatively stringent passing criterion.

In computing, reuse is the sincerest form of flattery, and I shall be recommending the early YouTube videos of this course to my MATLAB students--“early” meaning before their course diverges too much from mine.