What I have learned about on-line collaborative learning

When the report “Researching Online Education” (quoted and linked below) was released, a couple people contacted me.  “Tell them what’s really going on in collaborative learning!  Tell them what we really know from research!”  I looked at their report and concluded that the work I’ve done and am most familiar with doesn’t really have much to do with what they’re exploring.  I don’t know much about business models in on-line collaborative learning.  I do think that some of the work that I did 10-20 years ago in computer-supported collaborative learning (CSCL) is relevant for today’s MOOCs and other on-line learning experiments.

The work led us to a few hypotheses: (1) We’re skeptical a business model that charges for content will work at scale and in the long run. (2) We expect education platforms that offer vertical content and/or specific education experiences will be more successful than horizontal platforms, though we think credentials and careers offer two opportunities for horizontal aggregation. (3) Without credentialing or careers, online education seems aspirational and removed from the day-to-day of many people.

via Researching Online Education – Union Square Ventures: A New York Venture Capital Fund Focused on Early Stage & Startup Investing.

I got started working in CSCL as soon as I got to Georgia Tech in 1993.  Janet Kolodner took me under her wing and got me started on several projects developing collaborative learning activities on-line with engineers and architects around campus.  Note that this was two years before Mosaic, so we had to build our clients ourselves.  We built a system called CaMILE (Collaborative and Multimedia Interactive Learning Environment) mostly in HyperCard, but then moved it to the Web as soon as graphical browsers became available.

Relevant finding #1:  With CaMILE, we created a form of collaboration we called “anchored collaboration.”  Rather than a wholly separate forum, we could link to a particular thread in a discussion, so that we could (for example) link a homework assignment to a thread for discussion of that assignment.  Jennifer Turns and I did an analysis which showed that anchoring collaboration led to longer, on-topic discussions than having a separate forum.  It seems to me when I look around at on-line learning forums today, they’re mostly stand-alone — not integrated, not anchored.

Later, we developed the CoWeb or Swiki (which I talked a bit about in a previous post).  We had several reasons for moving to Wiki’s.  We had noticed with CaMILE that the anchors that were most effective were written by teachers.  Would they have been as effective if written by peer students?  Wikis gave us the chance to explore that.  (Unpublished finding: Nope.  The posts by the teacher are always the most interesting, generating the most traffic.)  We were also interested in moving away from a strictly threaded model, based on the work in CSILE and the Knowledge Forum on network-based representations that may lead to better student learning.  Most of our earliest work with the CoWeb or Swiki was descriptive: Teachers and students were doing all kinds of wonderful things with it, and we simply tried to catalog them.  Over those early years, the Swiki evolved rapidly, in response to the needs of teachers and students. Jochen Rick did a nice CSCW paper describing our design process and how the Swiki met the needs of different roles in an educational context.

Relevant finding #2: One of our coolest findings from back then was that collaborative learning could be better than classroom learning at lower cost.  Jochen Rick ran this study, in two English classrooms: One doing close-reading on paper, and the another doing the identical activity in a Swiki.  We tracked costs down to teacher and student time (e.g., using diary studies).  The Swiki-based learning was better and at lower cost.  Here’s a paper providing an example of a blended classroom that really did reduce costs and improve learning.

Relevant finding #3: We did CSCL research for a long time (from pre-Web into the early 2000’s), and we started to notice how and where collaboration worked and when it didn’t work.  Jochen did another nice paper on the interaction between the culture in the classroom and collaboration.  (His dissertation work explored the complicated issues of permissions, privacy, and transparency in personal webpages.)  We had one really large project where we worked on cross-disciplinary collaboration between engineers, mathematicians, and computer scientists.  It was a disaster.  Students had no interest in collaborating, and even accepted failing grades rather than participate in the Swiki.  (We called this “non-integrated engineering education.”) Our work completely changed — instead of creating collaborative learning situations, we switched to studying why they didn’t work.  These are important results for the MOOCs: Collaboration doesn’t always happen, and making it work sometimes requires changing culture, which is hard to do in an international, multiple-thousands-of-students “classrooms.”

One of the final projects I did in CSCL was with Karen Carroll.  We noticed that, in our English class study, there really wasn’t all that much use of the Swiki by each individual.  I had done a “dirty secrets” paper years earlier that got a plenary spot in a CSCL conference, showing that use of online collaborative forums, viewed from an individual level, was far too small for learning to occur. Our existing theory on collaborative learning (e.g., Roschelle, 1992) says that learning arises from the dialog between the participants — 0.5 notes/week/student (a fairly regular rate across several studies) is not a dialog.  We found a couple of similar papers in the literature that, like our English class study, showed significant learning, but without significant dialog.  How is learning occurring?  Karen did a really interesting interview study, where she explored all the ways that reading the on-line forum led to learning activities, even if there was no posting.  I wanted to follow up on that, to see how common these activities were and if they did explain the learning we were seeing.  But then Media Computation came along.