Wed 7 Mar 2012
This is an interesting proposition from Simon Buckingham-Shum and others in the SOLAR group (SOLAR stands for Society for Learning Analytics Research): an open Learning Analytics platform (OLA). The white paper can be found here.
The idea of an open shared platform for Learning Analytics recognises that current work in LA is mostly done in isolation which will inevitably lead to data silos. Instilling openness at an early stage, so they hope, will avoid the dilemma that many of us have experienced before with all kinds of learning technologies – the evolution of a cottage industry.
The concept definitely has its merits, anonymous datasets can be uploaded and then shared for e.g. testing different approaches, algorithms, or representations. Also analytics tools can be shared in this way with a global audience of researchers and users, thus leading to replicability and transferability of research, just as it should be. Additionally, it would lead to harmonious datasets of anonymised data from different sources.
Simon also presented some value propositions on why people would want to share their work and data in this way. For researchers, access to a large set of data which are otherwise hard to come by are further advantages to the above. Learners and educators could explore which of the analysis and presentation tools best suit their needs. Even enterprise has been included in their thinking as part of the LA community around the platform.
I like this concept of openness in general, and the next logical step in this venture now is to find funding to build the core infrastructure. It would seem that EU funding like FETOpen would be a good way to get this going, since the EU Commission is also interested in keeping things open.
I do have some hesitations and doubts too. A minor concern is fragmentation as there are already several data/tool sharing sites (silos if you like), e.g. DataCite.org, PLSC DataShop, LinkedEducation.org, or Processmining.org. Can we really avoid the proliferation of further such sites by creating the “mother of all analytics portal”?
The main issue is the unresolved ownership of data. Anonymisation may be superficially enough to publish personal data, but it doesn’t necessarily identify who is allowed to publish them and under which conditions they can be exploited. In Genetic research, to take just one example, the identification of patterns in data led to a frenzy of patents and patent wars. Just releasing learning data to a portal that can then be “looted” and patented by commercial entities is rather unappealing. What I think we need is a “Data Commons” framework that works in analogy to the Creative Commons implementations. This would certainly help clarify the commercial exploitation of data available under that condition, but would restrict access to others.