To me learning has a lot to do with self-determination and the freedom of choice. That is the freedom to choose your own destination in life. Learning also has to do with trust, namely the trust that you choose the right learning path and that you are competent in what you are doing. Competence here is meant in terms of the autonomy to achieve things and being confident about your own choices (not necessarily that you are also good at it).
I believe that Learning Analytics can help you in this, but it can also achieve quite the opposite. If data is used to exercise control, it can lead to mediocrity and conformism. What I have seen in forms of Learning Analytics tools so far, has more to do with Educational Business Analytics than with learning. Tools to analyse grading behaviours of teachers or truancy trackers for pupils, for example, stimulate reflection about the appropriateness of such behaviour. If a teacher is shown to mark too harshly or too leniently, this is (implicity or explicitly) flagged up. It is this flagging that triggers a Pavlov effect, because it must lead to re-adjust your outlying behaviour. It's a virtual slapping tool!
If we take Learning Analytics not as a cognitive technology, but for behavioural learning, most people will aim for the middle ground, so as to avoid standing out of the crowd. It simply is the safest place to be in a statistical analysis! But this also means to give up the self-determination and freedom of choice and be patronised by an algorithm.
While there seem to be clear commercial visions for the production of tools to support Educational Business Analytics, what I feel lacking is the vision for learning. What is the greater good of Learning Analytics? Should we not worry to be steamrolled over by commercial business monitoring systems that may impact on our learning, and our self-determination? I would like to see a shining path to follow that will lead us to better learning and teaching, not to more administrative pressures.