From Thrun’s blog, explaining why the Udacity approach didn’t work well for SJSU:

The traditional-semester pacing of the classes didn’t work well with the lifestyles and time-demands of the students in the program. In fact, 30% of our students worked 30+/hours per week in addition to coursework. Another 40% worked at least part-time. Work, families, other classes, and high school schedules demand a more innovative approach to pacing and we’re committed to figuring that out with SJSU over the fall.

Thrun has been telling us how he is going to revolutionize education for a couple years now. Yet last week was the first time he learned that students work substantial hours while going to school. His education on this issue isn’t quite complete, because he still thinks this is some weird “nontraditional” behavior. Eventually one hopes he’ll go into the NCES data to find out his “nontraditional” student is not that far from the median student at a state institution.

Interestingly, his reaction to finding out students work is to complain that you can’t possibly teach working students in math in as short a time as a semester. (!!)

My sense is that xMOOCs have a community problem.

Sure, you can get an answer to a math problem at 2 a.m. from a student in the Czech Republic, and that’s pretty cool. But whereas cMOOC communities persist and do meaningful things in the world, in general xMOOC communities are less robust. They don’t persist. They connect students as students, but not as colleagues. It reminds me of this old diagram from Jon Mott and David Wiley comparing Open Learning Networks to class networks based on a CMS:

Which could easily be updated as

One of the reasons this happens is that too much of the conversation happens inside the proprietary system. A lot of people teaching xMOOCs have worked around that. I know Stanford Online sets up IRC channels for their courses that live long after the course is gone, and that many professors in the CourdacityX courses push conversations to Twitter and other social media channels. That’s good, and I hope they keep moving in that direction.

But the other piece that makes the networks less robust is that the primary focus of the social interaction is often the course itself, and not the individual work or interests of the students. Communities thrive and survive when they are tied to things you do — your work, your hobbies, your music. They will survive the class to the extent that they have been built around activities that are bigger than the class itself.

That’s why I think that if xMOOCs want to truly have persistent, effective communities they are going to have to build the community not around success in the course, but around larger, more authentic applications of course content. To a certain extent, the xMOOC will have to become the chewy center of the cMOOC.

I’m not saying this is revolutionary. It absolutely already exists. You can make an argument that this is a version of ds106, or a cMOOC with a more rigid assignment core. And you can, of course, cut out the F2F layer there if you want — it could be an online subcohort layer, or you could eliminate it entirely. I work in blended learning, so this happens to be the scenario I’m interested in.

You can also argue that if the massive cohort is structured as a cMOOC, then the xMOOC is really just a collection of shared digital resources. and not a MOOC at all. Perhaps it’s a “distributed flip“. I wouldn’t argue with you there.

But I think the basic idea is sound, and the direction that the “massive” communities of xMOOCs are going to go if they are going to retain meaning. This is particularly the case now that xMOOCs are moving into blended scenarios where a global community based on xMOOC assignments is a bit redundant when students have access to a local face-to-face subcohort. That subcohort will absorb a lot of the interaction that is now being handled in the massive cohorts of xMOOCs, potentially making the larger community a bit meaningless unless it becomes somewhat more authentic.

Presentation I gave at InstructureCon on Distributed Flip types of things. I blow through the presentation in about 16-17 minutes, but stick around for the Q & A if you have time, I think it was a great discussion.

I agree with the bulk of the Salon article We Must Hate Our Children. This stat, however, is pretty egregious stat abuse.

[Y]oung American adults are less likely than older Americans to have attended college. This has to be the first generation for whom that’s true. We’re putting the history of American progress in reverse.

To think about why that stat is wrong, consider the following made up but probably completely correct stat:

Young people are having less sex than ever. In fact, older adults were less likely to be virgins than 21-year olds!

I’m guessing you get it, but briefly, if college attendance was completely level in this country older adults would still be more likely to have more education, because older adults have both the education they got when young and the education they got when older, whereas the young — I think you get it. I won’t belabor the point.

The best comparison for this sort of argument is how, say, people in their late twenties today compare in degree holding to people in their late twenties 20 or 30 years ago. There are a number of ways in which this is not completely like-to-like (compositional shifts, for example), but it’s a pretty good measure.

What does that measure look like? Via the New York Times (which is via NCES data):

That’s right, educational attainment in America is skyrocketing. That’s not necessarily for good reasons. A lot of that growth explosion is to do with the recession and the growing unemployment divide between the college educated and high school grads. And the story is not unambiguous — clearly before the recession there was a long period of stagnation that might be related to both student loan rates and college cost.

But why does it matter, if I agree with the author in theory about the crippling effect on student loans? Quite simply, because getting it right matters. To solve a problem, you have to know what the shape of that problem is. Statistics like these distract us from solutions that may be available, and ultimately hurt students.

I used to do edstat debunking on here a lot; I’m going to start doing it again I think. If you find anything fishy out there, tweet me @holden.