Like many, my complaint with the MOOC isn’t the MOOC itself, but the surrounding hype. My complaint is simpler than the eminent devaluation of teaching in terms of personal and public investment thanks to the MOOC marketing (as Bogost has argued and others have provided evidence). My primary complaint with the hype is the hubris.
The MOOC hype grossly underestimates the computational difficulty of teaching.
Effective teaching is hard. It’s harder than driving. It’s harder than chess. I bet it’s even harder than medical diagnosis. [Or as a computer scientist, maybe I’m just not that great at teaching, more on that in a later post.]
As Larry Wall reminds us, hubris can be a virtue for programmers, along with laziness and impatience. But education is more than a programming problem to be solved – it is critical for us to keep this in mind. In fact, these programmer virtues end up being educator vices; effective teachers are tirelessly hard working, patient, and humble.
Can the Internet bring content to students with unprecedented ease? Absolutely, it already has. Is individualized tutoring better than the broadcast information delivery of a large lecture? Of course. Can machine learning help learners and teachers? Sure, I think so. Personally, will I use the online lectures in my own classes? You can count on it.
But the MOOC was not introduced as a way to help teachers. It has not been billed as an effective tool to help learners. It has been marketed as the solution to automate the task of overpriced, inaccessible higher education.
Data-intensive, Google-scale machine learning will eventually [patience, patience] help teachers and learners. But it will also have limits. Consider language translation, current machine learning approaches have a considerable ceiling. And that’s just information translation, teaching is even harder. Google is amazing: page-rank, street-view, self-driving cars, but these are easy problems compared to teaching.
Can teaching be automated? Or will there always be a substantial amount of human labor involved? Some aspects of teaching are already automated: books deliver facts, spreadsheets compute final grades and adaptive diagnostics place students into classes (for better or worse). But what about the general-purpose teacher, can she be automated?
Teaching and learning are so multi-faceted and varied that their definition is as slippery as the AI problem itself. And unlike the Turing test, a simulation of human-level behavior isn’t enough. Teaching is not concerned with what is intelligent, but rather, what should be intelligent. In an important sense, education is concerned with making human computation better, which by its latest definition, are exactly those aspects hard to automate.
To be a good teacher, being patient and hard-working are necessary, but far from sufficient. If that were the case, computer science faculty would have written programs to automate teaching drudgery years ago. Programs are tirelessly hardworking and patient. But can they be humble? Humility and empathy are the hallmarks of good teachers. Humility and empathy enable teachers to motivate students and adapt to their individual and collective variation.
Learn local, learn global.