One of my tasks next week is to present Cohort 2's experience of learning sets in the online environment. This is one area where (unusually) we did not build on Cohort 1's experiences in the first iteration of the fully online BA (Hons) Learning, Technology, Research degree, as Cohort 1 delayed the introduction of formal sets until the start of Level 3. In comparison, Cohort 2 researchers will already have had experience of learning sets for at least two modules at Level 2.

I'm drawing a distinction here between formal and informal sets because its quite probable that before the introduction of sets, small groups of researchers might have decided to share assignments and offer critical review. These groups might now recognise that they were in fact forming learning sets - but there should be a distinction between these and the conscious commitment to learning sets.

"The limits of my language are the limits of my mind. All I know is what I have words for."
(Wittgenstein)

The function of our learning sets is to enable more efficient peer review, though many learning sets do far more than this - and some have forgotten the point of the set in the rush towards the end of a module. Peer review was introduced much earlier than learning sets - partly because the Cohort 2 team wanted to see if the learning sets could be larger - the size of the existing sub-communities (from 6 to 20 researchers). This was the first iteration - though this was not made explicit to researchers. The results were mixed, and one researchers' request to try smaller groups coincided with the team's implementation of smaller sets. I did a review of the literature before even the first iteration, but I seem to have lost track of this as I didn't think of this as an aspect of my research. However, the general concensus in research papers seems to be that, for both face-to-face and online environments, groups of about six would be appropriate.

What I could not locate was any advice about how to construct learning sets. My experience as a very minor cog in the pilot stage of the Cognitive Acceleration in Mathematics Education (CAME) research with Kings College London has influenced my view that the learning sets should be set up to optimise the experience - but as I couldn't persuade the team, we agreed instead that for the next iteration the groups should be self-selecting. An editable document was set up in FirstClass, and researchers opted into learning sets, some of which were based on shared research topics.

The third iteration (or second if you didn't recognise the first one) was the almost accidental introduction of learning sets in WebCT. Again, researchers were given the opportunity to self-select (I still couldn't persude my colleagues to take a different approach). WebCT has the advantage over FirstClass in that there is a tool to create groups with a choice of levels of privacy. This proved to be a success - but as with the previous iterations, the outcomes were far from uniform success.

Both qualitative and quantitative data is available, though I've had little opportunity for proper analysis so far. However, I can draw some general conclusions and identify areas for improvement - analysis of the data might offer more action strategies.

Problems and Challenges

Self-selection works imperfectly - factors include the level of confidence to start or join a group, late selection, uncertainty about shared interests, doubt about the ability to form an effective set, issues of trust.

Peer review skills need to be developed further - its easy to say what is right about another person's ideas, very hard to be usefully critical especially in an online environment.

Team work and the ability to take resonsibility for one's own learning - graduate skills - need to be developed. Much could be gained by analysing the qualitative data - researchers' evaluations of their experiences, examining what works in a successful learning set in order to find out what's missing in a less successful set.

Final comments (for today)

For the most part (how do I know this?), learning sets are not only an essential part of the research process, but they work well. We are working towards improving the experience and look forward to finding out how this has varied between the different cohorts.

---

Some links to my earlier thoughts:

Lydia's blog entry on peer review was an early inspiration

Building on Cohort 1's experience of Action Inquiry, forming sets

Some researchers' comments on the second iteration

Setting up in WebCT

A note on setting up Gill's access

A different perspective on peer review

What researchers did

... more of what they did

Data on discussions

Researchers' comments on size

More protocols needed