Tuesday, November 26, 2013

...or, How I Learned to Stop Worrying and Not Hate the Udacity Pivot

Of course it’s hagiography; it’s in Fast Company, and Sebastian Thrun is in Silicon Valley. Of course he screwed the kids at San Jose State (though more on the virtue of what Cottom does with that conclusion below); that’s what privatized higher education does when “students from difficult neighborhoods” turn out to be unprofitable. And of course he wasn’t making any money; that’s the Silicon Valley business model spoofed so well by South Park. Now if we’re all done doing our best Claude Rains impressions, I want to take a more pragmatic look at the Udacity “pivot” (which admittedly sounds like a chess strategy rather than pedagogy).

To summarize for those who haven’t followed this particular mess issue, Udacity founder Sebastian Thrun has decided that his MOOCs are a “lousy product” after the debacle of San Jose State University’s remedial math course, because, hey, maybe we should get more than 7% of students to pass a course before we declare it a success. Thrun concluded that the failure was because, as Fast Company’s faithful correspondent quoted Thrun,

"These were students from difficult neighborhoods, without good access to computers, and with all kinds of challenges in their lives,” he says. “It's a group for which this medium is not a good fit."

Instead, Udacity will move toward corporate and vocational training, which Thrun sees as a better use of MOOCs.

The Internet has been, well, atwitter since the Fast Company piece came out, with most of the response negative. A lot of that has been either unfounded or, more importantly, not useful, and I think there are ways of approaching this that move higher education forward. When I was teaching political theory, I guided students with two principles of what I called “pragmatic interpretation”:

  1. Unless you’re doing history for history’s sake, think about the ideas themselves and not the author’s intent.
  2. Give the ideas the most generous reading as is consistent with the available evidence, including the entirety of the text (but no more so).

Let me use these in understanding what’s happening with Udacity. There are a few lines of criticism that are problematic, but that I think lead in useful directions. Audrey Watters and Mike Caulfield have argued that the pivot toward the corporate world reflects a class preference on the part of Udacity, moving toward an environment in which the high failure rate is a feature that filters out the “wrong kinds” of students rather than a bug. Similarly, Rebecca Schuman argues that “Thrun’s alleged mea culpa is actually a you-a culpa,” blaming students rather than the medium for the SJSU failure.

As a matter of authorial intent these may or may not have merit; all three authors have followed Thrun quite a bit more closely that I have. I don’t see anything in that article itself that leads me to such conclusions (that generosity principle), but more context and less hagiography might easily change my mind. From a pragmatic perspective, though, Udacity’s failure and these criticisms point us to what I think ought to be the critical question about MOOCs: is there a good fit between what we build MOOCs to do and what we’re using them for?

Watters is particularly critical of the pivot to vocational education:

I am not willing to shrug off lousy educational practices simply because they occur outside the walls of formal education. Many professors have been quite vigilant about criticizing MOOCs foray into higher ed; I think it’s just as important to keep that up if MOOCs want to conquer vocational ed instead. If MOOCs – short videos, multiples choice quizzes, and robo-graders – offer bad pedagogy, then that means they offer bad pedagogy for everyone, everywhere. To ignore bad pedagogy simply because it occurs in settings outside the humanities or outside the college curriculum is elitist and wrong.

To be sure, the final point is spot-on: vocational students deserve quality education as much as anyone; to the extent that vocational students are among the least well-off they deserve it more than the rest of us. But we’re only ignoring bad pedagogy if Watter’s assertion that bad pedagogy is bad across the board is correct. I don’t think it is.

My experience continually points to the conclusion that pedagogy is very dependent on context. It doesn’t only vary across institutions; it varies across courses and even topics. I’m a big believer in conceptual thinking. When I teach the legislative process, for instance, I don’t care if students understand the difference between a standing subcommittee and a conference committee. What I care about is that they understand the concept behind it: a process with so many veto points is a process in which most bills fail absent negotiation, cooperation, and compromise. So I use an online simulation based on an extensive-form game of the process to teach that.

But I can’t do that with federalism. There’s no principled division of power between the federal and state governments, because that division was the result of negotiation rather than rationality. The only option is to memorize (or keep handy, which is why Google is on phones) Article I, Section 8 of the Constitution. So that’s how I teach that. It’s a different topic, and it requires a different pedagogy.

Perhaps some pedagogies are bad across the board, but I don’t think MOOCs are one of them, and in fact vocational programs may well be a good place to look for success. I used to teach vocational programs at the Salt Lake-Tooele Applied Technology College and then Salt Lake Community College’s School of Applied Technology, and my wife still does the instructional design for them. We saw quite a bit of success with those students that Thrun dismissed in open-entry/open-exit, computer-based courses that relied on “short videos, multiples choice quizzes, and robo-graders.” Part of that success was because they had a classroom and schedule to keep, and a teacher who could add to the content when the student needed it. These students were far from the roving autodidactics that the MOOC-everything crowd seems to assume, but they were working primarily by themselves and getting quite a bit out of it.

To me that means that, Thrun’s motivation and intent in moving to vocational and corporate training aside, the move itself may have some virtue. MOOCs may have a place in higher education, especially if we understand higher education as not just the traditional liberal arts curriculum, an attitude that I think is every bit as elitist as using MOOCs as a tool to transform privilege into job qualifications. The converse of the “if it’s not a problem for me then it’s not a problem for anyone” definition of privilege holds as well: we invoke privilege when we assume that because it’s a problem for us it’s a problem for everyone.

So what might a MOOC do if it can’t teach remedial math and it can’t deliver “anything as rich and powerful as what a traditional liberal-arts education offers”? Thrun seems to be moving toward teaching things like programming; indeed he still seems to think Udacity’s MOOC-based master’s degree in computer science at Georgia Tech is a success. (I’m sure that has nothing to do with the fact that one of his first investors said it “isn’t really a Georgia Tech deal. It’s an AT&T deal.”)

That seems the wrong direction, for a reason that Watters identified:

I’m curious, for starters, why we wouldn’t want software engineers to have that [liberal arts] background. What’s missing from CS curriculum today? Is it simply a matter of content – lessons in Hadoop, for example? I once asked Thrun’s co-founder David Stavens, incidentally, if Udacity planned to offer classes in communication or project management or documentation – three things I think a lot of engineers suck at. The answer was no, making me wonder what sort of career the Udacity classes were actually going to prepare folks for.

Programming absolutely requires something “as rich and powerful as what a traditional liberal-arts education.” And not just because they need skills in communication or management or technical writing. In my albeit limited programming experience, it seems like there are usually several ways to reach the same output, and the programmer must choose which to adopt. If computer science is like most fields, graduate work moves students beyond problems that are already solved into ones for which they can’t compare their answers to a known answer, exacerbating this challenge. Frequently my choices have been aesthetic as much as technical, and I suspect that’s true of many if not most others. Programming is thus as much a literary effort as a technical one, requiring critical thinking and interpretation. I thus doubt that MOOCs are any more suited for graduate work in computer science than for remedial work in math.

Let me suggest instead that MOOCs might be more helpful if two conditions are true. The first is if the content can be described in the lower range of Bloom’s cognitive taxonomy: knowledge and comprehension with some application rather than mainly analysis, synthesis, and evaluation. MOOCs don’t seem to teach the latter well, which may well explain why they failed at teaching remedial math. But no student needs me to help them memorize a section of the Constitution or learn the steps for creating a table in Word. Students learning that kind of material need someone to tell them what to memorize and a place to find that content, and a web page does that every bit as well as a person. MOOCs, I suspect, can be used very effectively to deliver this kind of content.

But few programs, even at the vocational level, stop at comprehension. Students have to take that content and do something with it. They have to make their own tables or decide whether an act of government is a state or federal power. That takes critical thinking, and that comes through discussion. My lectures in research methods have students struggling to test a hypothesis even though they know all of the techniques; they needed the one-on-one interaction to walk them through their own hypotheses a few times. Whether that’s an in-person instructor or an online teaching mentor like Western Governors University uses (with, I’m told by some students, limited success), it can’t be done by a computer.

Thus MOOCs need to be built into a non-MOOC structure that first and foremost gets students to show up and then gives them an environment in which to go beyond comprehension. At SLCC, that’s a classroom with computers; a set schedule for each student; and an instructor who consults individually, grades work that can’t be graded by machines, and gives students feedback on what they did, not what a hypothetical typical student does wrong. At a university maybe this means that MOOCs provide small content blocks that serve as the basis for discussions led by a professor: an expansion of the classic English tutorial model, which I’ve always thought worked better anyway.

That raises two problems for MOOCs. The first is that the key argument in their favor is cost savings, and that cost savings comes by replacing labor with capital. (You didn’t need Marx to explain that one to you, right?) But if what I suggested above is right, MOOCs aren’t likely to work without that connection to one-on-one or small group instruction, so the cost savings are at best a minor share of what Thrun et al. predicted: Universities aren’t saving much unless they replace faculty with teaching assistants, which can be accomplished through adjunctification much more easily than with a massive investment in new technologies and corporate partnerships.

All of this is, of course, speculation, which brings me to the much bigger problem. I’ve offered my best guess as to what MOOCs might do well based on what happened at SJSU and how Udacity has responded. I’d like to think I’m right, but I certainly don’t think I’ve created a magic formula. Ultimately though we’re experimenting here. And that’s the bigger problem MOOCs face.

Tressie McMillan Cottom wasn’t the only person to argue that Udacity and SJSU treated the students as unregulated research subjects. But she was the only one (that I read at least) who asked the key question in that: why did universities allow their students to be used as such? Cottom puts it in the language of IRB regulation; an academic researcher would surely have to provide an extensive set of protocols to protect the students from risk, so why shouldn’t Udacity? There were certainly risks: failing a remedial math class would put the students further behind on the road to graduation, and passing without really learning the material increases the risk of failing a future course. This isn’t an upper-level elective; it’s a foundational course.

The easy answer to Cottom is to say that this wasn’t subject to IRB regulation because it doesn’t meet the federal definition of research, an answer that is technically true but practically useless. It simply takes the question one step further back, exposing much more deeply troubling issues.

  • Why don’t tech businesses have to protect the human subjects with which they experiment when they develop or market new products, especially when those subjects are among the more vulnerable in society?
  • Why are universities—most of which are, after all, either charitable organizations or run by governments ostensibly of, by, and for the people—only concerned with technical compliance and not the kind of ethical behavior that they claim to teach their students?
  • Is it a coincidence that Udacity experimented on and then abandoned students at one of the most diverse institutions in the nation but is keeping the program at Georgia Tech, where 70% of US-resident graduate students are white and the program is sponsored by AT&T?

Hence Cottom’s conclusion: “We may need more experimentation in higher education but it should be as explicit and ethical as any other we conduct in the name of science and progress.”

As we experiment with MOOCs—and I think we should continue to experiment with them—we damn well better take seriously the idea that we are experimenting. We damn well better take seriously the idea that good pedagogy is content-dependent. And we damn well better take seriously the idea that the least well off bear the burden of risk whatever we do.

Thrun is no longer St. Sebastian (a title that has been taken over by one named Vettel). Whether he is the prodigal son returned repentant or the Devil looking for souls easier to steal isn't a question worth asking. What is asking is whether those seeking responsible education can find something in Udacity's experience that moves us toward that goal. I think we can.

No comments:

Post a Comment