Monday, May 8, 2017

Trouble in MOOCland

The MOOC experiment is not going well.

I know I wrote glowing reports about the potential of large-scale online courses a few years ago, but there were problems with the format all along, and far from being solved, the problems are getting worse.

It’s understandable if I was enthused about a medium that looked like it could deliver specialized education to the whole world on any level from high school to graduate school for less than the price of a book. It turns out that online courses are more expensive to deliver than books, and not just slightly more expensive, but many times more expensive. The promise originally was that the MOOC format would be, to borrow a phrase from another industry, too cheap to meter. But the full-featured MOOCs today have costs so high you can compare the tuition to the traditional approach of a classroom education. At the two biggest names in the MOOC medium, Coursera and Udemy, a student pays almost as much as they might pay for a college course, but without the college credit. What is there for an economist to get excited about? Take away the cost savings and online courses are a novelty scarcely more exciting than, well, nuclear power.

The minimal cost savings of MOOCs could be forgiven if they got the job done, but they don’t. The failure rate of online courses is a continuing embarrassment. Routinely more than 90 percent of students who enroll in any given online course fail or drop out. In many cases failing and dropping out amount to the same thing, especially when required assignments are all but impossible to complete. A mechanically graded test may be graded at random because of a software error so that it is only by luck that a student scores over 20 percent. I have seen test questions that contained words and concepts that hadn’t been mentioned in the course itself, so that I was left sifting through Wikipedia to find the most likely answer. In a programming course I took one of the programming assignments due in one week took more than a week to run — I was fortunate to be able to divide it between two computers, but what happened to the students who had only one? With most MOOCs having a failure rate above 90 percent, and with no instructor actively present while the MOOC is going on, such problems can easily go undetected.

But it not just that MOOCs are in a weak competitive position. The predominant emotion surrounding the MOOC medium is disappointment. I remember how disappointed I was when I signed up for an online course in which the content had no earthly connection to the course description that the catalog had provided. If I was disappointed while not actually paying any fees, imagine how disappointed today’s MOOC customers are to find out after they have paid fees ranging from $79 to $999 that the online course they have enrolled in will not teach them about the subject they were promised. Yet this must happen every day; course titles are routinely misleading, catalog descriptions are now only a few words long, and providers no longer make it possible for students to preview course content or even see a meaningful outline before enrolling.

Some of the disappointment is the result of the small amount of content delivered in today’s MOOCs. A “course” nominally runs for four weeks and may be the equivalent of one chapter in a textbook or three weeks in a college course. At Coursera the first “specializations” covered a substantial amount of ground, bigger than a minor field of study in college but smaller than a graduate program. But that, it turned out, was too expensive to deliver. A “specialization” has been drastically scaled back so that it involves a time commitment of roughly 140 hours. That and the amount of content presented make a “specialization” the equivalent of a single undergraduate course — or arguably a little less. I worry, though, that students may be misled by the five-course format of a “specialization.” You could easily imagine that you are paying for the equivalent of a semester of college. Just imagine how you will feel afterward when you add up how much you have paid and how little you have learned. Meanwhile, the specialization certificate I earned a few years ago with 10 courses and 500 hours of study is no longer so impressive. People who know what a Coursera specialization means now can’t easily imagine what one used to be.

Online course platforms are so expensive to operate that the instructors and content providers are not meaningfully paid if they are paid at all. Some Udemy courses, I am told, have no pedagogy at all, but are little more than books repackaged to make them look like courses. You literally could just buy the book and read it and spend a lot less time and money.

The fundamental challenge in an online course is to find a way to distinguish between those who are actually learning the material and those who are just going through the motions. This is a problem for which no one seems to have good answers. The problem with mechanically graded tests is that the students who do best at them are the ones who have stolen the answers in one way or another. It is a little too easy for a criminal to do when the test is posted online. Recognizing this problem, Coursera has gone all in with anonymous peer assessment. Other students in the course, people who know no more than you about the subject, will grade your work. When you are the student being graded, this is an excruciating experience. You have surely had a chance to see the quality of anonymous Internet commentary on YouTube and Twitter. Imagine that these same people — they are anonymous strangers chosen at random from the Internet, so as far as you are concerned, they might as well be the same people — will decide whether the essay you spent three hours writing is a success or a failure. Did I mention that some of them are 12-year old boys with an axe to grind? As uncomfortable as you might feel having your college professor go over your essay, having it graded by John from the Internet is much, much worse. Yet Coursera requires anonymous peer assessment as part of every course it offers. It is a powerful deterrent to students who know what anonymous peer assessment implies and a heartbreaker to those who do not know what they are getting into. It is a fundamental flaw in Coursera’s business model and is a problem that the company has yet to take a serious look at. But at least Coursera is trying to assess students. I am less familiar with Udemy’s practices but I am led to believe they have next to nothing that would reassure a student that they are in fact learning the material they are studying. Not knowing how well you are learning might seem like a problem, but I will gladly take that after having faced the experience of being graded by anonymous strangers.

Most of these problems can be solved eventually, but there does not seem to be a solution to the cost problem. If Coursera, Udemy, and their competitors look like they are ripping off their customers, it is certainly not that the companies are making a healthy profit. Perhaps the platform for presenting online courses is, like the nuclear reactor before it, simply far more expensive than anyone ever imagined. Whatever the explanation, the enthusiasm for MOOCs depended on their scalability and low cost. Now that we know that both advantages existed only in our imaginations, it may be time to acknowledge that the more traditional learning media are in fact more cost-effective.

I must mention the enormous cost advantage enjoyed by those who can learn from web sites and books. After learning a few days ago that it was no longer possible to enroll in single course on Coursera, I decided to compare the cost of the “specialization” to the cost of learning the same material from books. As it happened, this was a very new and specialized subject area, so there are only 18 books on the subject that bookstores are able to tell me about. I could plausibly buy and read all 18 books, and then I would possess a fair approximation of the sum of the world’s knowledge on the subject — a claim no online course could possibly make. I would pay a few dollars less for the 18 books than I would pay for the “specialization.” Besides the purchase price, it would take time to read the books and digest the ideas they presented. Maybe it would 180 hours, but that is only 40 hours longer than I estimate the “specialization” would require, and that’s for a much broader and deeper study of the same subject. Or if that was too much, perhaps I could buy the one textbook that the “specialization’s” curriculum seemed to be borrowed from and read it. My cost savings with this approach: 87 percent. Time saved: 92 percent. The catch: I have to find a way to test myself. Another catch: statistically, books have a high “failure” rate. Less than 10 percent of people who buy a nonfiction book read the whole book. In the book business, we shudder at that statistic and wonder how we can do better. The mystery, though, is why the MOOC format, with all its bells and whistles, does no better than the book format in this regard.