Slowly, a common understanding of Learning Analytics is evolving. George Siemens' definition is the following:
Learning analytics is the use of intelligent data, learner-produced data, and analysis models to discover information and social connections, and to predict and advise on learning
My view is generally in accordance with this. To me, Learning Analytics is a way to take advantage of available educational datasets for the discovery of new insights into educational practice and the development of new educational services that support the goals and objectives of learners, teachers, and institutions. Unlike other people, I feel that Learning Analytics serves reflection as much as prediction. The careful design of Learning Analytics approaches needs to take a number of different perspectives, so I started to draw up a first framework diagramme that takes these into account. This is only a first draft, but I hope to develop it further as our knowledge and experience increases. Feel free to elaborate further, and I am, of course, happy to have feedback on this.
Critical soft issues that I perceive in the exploitation of educational datasets for Learning Analytics, are the competences to interpret, critically evaluate, and derive conclusions (and pedagogic actions) from the analysis. A holistic perspective is essential, because what's left out of the data coverage is as important as what is in. Privacy issues aside, which are a legal constraint, I see a number of ethical issues connected with accessing learner data. There's the issue that some teachers might abuse Learning Analytics as a means for policing and surveyance rather than a support tool. The same could perhaps be said when it comes to institutions gaining insights into teacher performances. The danger being that innovative and creative teaching might be ousted because it does not show up as falling into line with the 'traditional' algorithmic performance. As such the data might easily be abused to exercise certain pressures upon the data constituency.