Saturday, March 16, 2013

Automation of teaching

For my first post, I am not going to talk directly about MOOCs as such.  Rather, I want to highlight a recent study concerning a language-learning web site called Duolingo. It is claimed that "A team of independent researchers recently found that students on Duolingo take 34 hours to learn as much as a one-semester university course". (link to study)

What I find pertinent about this, relative to MOOCs, is the following.  MOOCs are touted as beneficial because they can automate instructional tasks previously undertaken by human beings.  The MOOC, then, is the latest in a long line of technological innovations that reduce or even eliminate human effort.

That said, it is arguable that there are aspects of teaching that are not amenable to automation.  My own hunch is that automated teaching is essentially an application of artificial intelligence.  Since AI is not (at least not yet) coextensive with human intellectual capability, it seems that certain tasks remain that are not amenable to automation, where a human teacher still has a valuable role to play.

We might welcome, then, a technology that allows the aspects of teaching that can be automated to be automated so that we might have more time and energy for the aspects where a human still ought to be in the loop.  Considering again Duolingo, while it might be able to replace human instruction for the first semester (or year) of reading and writing practice, what it does is open up an opportunity for such courses to increase the emphasis on conversation, discussion, creative writing, and even tentative exploration of literature.  I suspect the situation might be similar with MOOCs across a variety of courses.


  1. I agree... there are various aspects of teaching that require greater intelligence than can (currently) be automated. There's also room for current A.I. techniques to automate more aspects of teaching than we currently are.

    The Duolingo project seems like something to keep an eye on -- most of the projects of Luis von Ahn (famed for CAPTCHAs/reCAPTCHA) are interesting. For the quick scoop, here's a Ted talk where he plugs Duolingo among other things:

    By the way, I tried to follow the "independent researchers" link, but received a 404 not found. Can you update the link?

  2. Sorry about the broken link; it should now be fixed.

  3. It makes me think about what sorts of automation could be helpful in my own field, philosophy. So much of what we do happens in conversation between people articulating their thoughts with each other, the main promise is in some way of doing that across distance. Language learning involves, at least in part, picking up a set of discrete skills. I don't really see how philosophy can be similar--there are skills, like careful reading and writing and thinking, but they are fostered and nurtured through very thorough and thoughtful interaction with others. I still remember learning from Gabe about Pascal. I could read Pascal on my own, of course, but what could automate what I learned about his work *from* my conversations *with Gabe*?

  4. Chris, I find your question intriguing, and of course I'd like to thank you for the flattering example. With philosophy, as you say, the most one can hope for is to use technology to enable a "thoughtful interaction" with an otherwise inaccessible interlocutor. There is nothing that a MOOC can do to make those interactions "massive." A computer as such is not (and I believe cannot be) human enough to make the interaction meaningful or thoughtful; a computer cannot be the interlocutor.

    I believe that what you are describing here is the core of what we can do as educators in the liberal arts college setting. There could be some real added value to identifying those components explicitly, in every discipline, and structuring courses accordingly.