MOOC Stats and Trends in 2014

I have also taken programming MOOCs (algorithms 1 and 2, databases, and computational investing), and the programming assignments are a good fit for MOOCs - it's easy to check the result from running a program (for example, the sizes of the five largest strongly connected components of a directed graph consisting of almost one million nodes).

In the fall, I took Financial Markets (Yale, from Coursera). I don't know if that qualifies as humanities or not, but it did have a peer-graded component that worked better than expected. Each week there were multiple choice quiz questions on the videos. But one week there was also a writing assignment. You had to read a fairly technical paper on market efficiency, and then answer three questions on it with max 300 words each.

After the deadline, you have to grade 5 other students answers (as in "Did they summarize the main points correctly?"). This worked much better than I expected. Summarizing the main points, and then reading through and grading 5 other answers really made the content stick. So I was actually pleasantly surprised at how well it worked.

But I think it the peer reviewed writing works best when the assignment is a fairly focused question to answer.

My reviews of the courses:

http://henrikwarne.com/2012/05/08/coursera-algorithms-course...

http://henrikwarne.com/2013/02/18/coursera-algorithms-course...

http://henrikwarne.com/2011/12/18/introduction-to-databases-...

http://henrikwarne.com/2015/01/01/coursera-course-review-com...

I've taken a bunch of MOOCs, mostly technical. Only one involved writing "proper essays" (The Modern and the Postmodern, Wesleyan University) but it was one of the best courses I've ever taken.

There was a lot of peer assessment, but

a) Peer assessment is actually amazing for the graders (reading other students' work, and providing feedback), and

b) It can work well enough for the gradees if you're careful.

The tricks to making it work depend on

- Having enough people grade each paper (involves making your grade depend on your grading),

- Having some TAs grading or having some sample papers of known quality to provide "ground truth" values,

- Trusting the grades of "trustworthy" graders (an eigenvalue problem),

- Trusting the grades of "good" students.

That said, I've looked at some other humanities classes and been scared away by poorly constructed multiple-choice quizzes, so it does take effort to get it right.

(Might only work for introductory/survey courses, though.)