I am guessing that could be both:

(1) about  prediction issues which are the basis for Intelligent tutor systems.

(2) about student performance definitions and measurements.

I've just finished a literature review about student performance and some of my findigns are that the definitions and measurements of student performance can be several and not standardized or consensually defined. The thing about prediction models is that they depend precisely on definitions, measurements and the possibility of replicability. They are useful just if you have same conditions as where they were generated and where they are going to be applied. Finally I would say that I have not found enough research about the 'human factor' 

all learning environment have to consider :

learner caracteristic,to providence adaptabilité