This is a great thread, and what I'm interested in as well. While "Learner Analytics" are a great idea to help students modify their behavior (or at least understand what the predicted outcome is), I'm interested in how we can use Analytics to design environments that improve learning outcomes for all students - or those targeted, such as those at-risk.
We have the most potential control/influence on the design of courses and strategies, so it seems like Analytics would be a great tool to help us evaluate which designs work, and which don't, for selected purposes.
Is anyone aware of literature around the use of Analytics to evaluate course and programs? I haven't found any myself.
We're planning this kind of work in the California State University system as an outgrowth of auditing/reporting metrics (such as those previously mentioned, around time on task, last login, etc.).