Thus far, the model for scaling xMOOCs (MOOCs in the Udacity and Coursera tradition) has been fairly simple and universally applied: build distribution systems that easily accommodate large numbers. Stream lectures on YouTube where anyone can view them, develop grading and correcting systems that can provide automated feedback or quickly distribute peer feedback, and host discussion forums where participants questions can be answered by other participants or highly up-voted questions can be dealt with by instructors. All of these systems leverage mostly pre-existing platforms, and dealing with 1 million students is, except for server costs, not much different than dealing with 1.
It's also not much more difficult to teach 100 courses than to teach 1, as evidenced by the 211 courses now hosted on Coursera.
These MOOCs are scaling fast. What would it look like to scale them slow?
What if the distribution mechanism of the course material wasn't just the technologies of dissemination, but a community built around a body of learning?
Let's say you believed that high-quality legal education required some kind of facilitated seminar experience. High-quality seminar facilitation cannot be automated, so it's impossible to scale quickly. But perhaps it could be scaled slowly.
What if you got 21 law school students to agree to be teaching fellows for a course accepting about 500 or so participants, on a topic of broad interest, like copyright or something. What if the following year, you ran a similar course, once again borrowing the time and talents of law school students, but adding a twist: you also borrow the cognitive surplus and good will of previous course participants. Basically, you identify another 10-20 teaching fellows from the most accomplished participants in the previous year's course.
With 40 teaching fellows, you could now admit 1,000 students. Out of these 1,000 students, you could probably find another 50 or so people willing to help facilitate the following year, and with the 20 from the law school and perhaps another 10 or so returning participants, you can now host about 2,000 students in the same high-quality facilitated learning experiences.
If these high quality learning experiences have significantly less attrition than the typical xMOOC, then you probably only have to hit this size, 2,000 or 4,000 enrolled students, to have the same number of course completions as a typical xMOOC.
And over the years, with continued doubling that draws on a constantly growing pool of people capable of facilitating high-quality seminar experiences, it would be possible to get to the point of having over 100,000 people enrolled within the decade. But you might actually get many of those people to stick it out through the whole course.
This would be scaling slow, a pyramid scheme of learning.
It would be very interesting if someone ran that design experiment.