Saturday, January 5, 2013

Coursera dreams and details

I listened raptly to Daphne Koller’s TED talk and was simultaneously inspired and appalled by the difference between its idealistic dreams and reality. One compelling aspect of MOOC was instant feedback -- to facilitate learning -- and how it relates to the way homeworks/quizzes are implemented.

However, the implementation has so far been extremely inconsistently across different courses and in some cases have deviated from the ideals envisioned. Feedback runs the gamut from too little to literally giving you the answers. This fine if the quizzes don’t count towards a grade or certificate but is totally bogus if doing so enables me to earn a perfect grade easily.

For instance, in Scott Page’s Model Thinking class the quizzes count towards the grade and gives the student multiple attempts (five) to try to get it right. This is somewhat good, I guess, but the feedback is literally the answer! I expect feedback to be something in the order of clues e.g. “the relevant concept here is …. check lecture …. etc.” Or if the answer is given then the following attempt would present a different question - but it doesn’t. In theory I could get a perfect score if I just filled everything in with the feedback. In Lada Adamic’s Social Network course, we get 100 attempts which is quite interesting given that it is somewhat balanced by very cryptic feedbacks with no indication of whether you’ve answered the question correctly. In what the Economist is reporting as the most popular Coursera course with over 180,000 students, we get 1000 attempts on exercises which do not count towards a grade and exams that do count. This course meets with the ideals - we get two attempts (I think that it might be just the right number - something like 2 or 3) with instant feedback on what we did wrong and then on the second attempt we get a different set of questions.

The problem with Coursera and its franchise approach is that very little is being done to streamline the experience - I guess in many ways this is like going to a real college. We may all get to go to Harvard online but the the course could also suck just like at the real world version of Harvard. I am hoping that this myriad approaches are some kind of randomized experiment to see which one works better but I am doubtful that this is so.

Again, Khan Academy has the right idea and adaptive testing is the way to go but one thing professors seem to hate doing is to create test banks. This is one reason why text books from companies like Pearson are so expensive but at the same time appealing to professors. The questions have already been prepared.

No comments: