Many discussions of education reform make reference to Baumol’s Cost Disease, an economic theory which seeks to explain why costs rise in industries, like health care and education, which are resistant to efficiency gains. Essentially it argues that costs will rise rapidly in sectors that are labor intensive because wages must rise to attract workers in a competitive labor market, but the labor intensive nature of the enterprise keeps productivity from keeping up with gains in other sectors that are more amenable to automation.

Baumol did his original research on the performing arts. pointing out in the 1960’s that it takes the same number of people the same number of person-hours to perform a string quartet as it did a century ago.  While that’s true, the actual cost to hear music is much lower than it was.  Why?…recordings.

While most people accept that listening to a recording is not the equal of the live concert experience, most people listen to more recorded music than live music.  The combination of lower cost (buy MP3’s of a string quartet once for $4 and listen as many times as you want) and convenience (even in the largest of cities there isn’t a live performance of the piece you want to hear whenever you want to hear it.)  mean that recorded music is, for most people, good enough. (See also Walter Benjamin’s “The Work of Art in the Age of Mechanical Reproduction“)

Here is a parallel to education. Most would accept the premise that a small class with lots of interactions is better than a class of hundreds in which most assignments are graded by computer.  The latter model, in particular, does not lend itself to higher level thinking skills without very careful design.  The proponents of what Lisa Lane would call Network based MOOCs Would likely counter that network effects and opportunities for learner autonomy create an environment that is in some ways preferable to a traditionally structured class, even if that class is small and interactive. However, the network based MOOC is not driven by measurable outcomes, and it’s unclear how this might fit in to a credential. When early network based MOOCs were offered for credit, there was a cohort within the class.  for that cohort, which was small, more traditional assessment methods were used (journals, projects, etc.)

Any kind of redesign of education which would significantly increase the productivity of teachers will likely involve significant automation of content delivery and, more importantly, assessment.  The key question is, will learners accept a course where most assignment feedback is machine generated as good enough?  I think this will most likely depend on whether a credential earned via low cost, massive, machine graded courses is perceived to have economic and employability benefits comparable to the traditional labor intensive approach.

Interestingly, Baumol apparently argues in a new book (I haven’t read it) that the “disease” is not , in fact, a big problem, because decreases in the cost of other things will offset the inevitable cost increases in education, health care, and other labor intensive sectors. I’m skeptical. I guess we will all see in a decade or so.