Our paper on the danger of Bias in MOOCs Learning Analytics was published in the International Journal of AI in Education (IJAIED)

We showed that Learning Analytics in MOOCs can be *significantly* biased by the aberrant behavior of users who are 'fake learners'. Because these users have very high success rate, they can lead to false conclusions regarding what forms 'effective learning behavior'.  
We demonstrated this by replicating papers by Koedigner and Champaign, and using sensitivity analysis methodology.   See here.