Rajat posted a link in subinterest 'Education'

As MOOCs continue to experiment with delivery models to achieve better student outcomes in order to bridge the gap between user intent and user behavior ( as reflected in low completion rates). One could perhaps think of two broad approaches:

1)Data/AI driven: As laid out in this Berkley piece on a machine learned approach "(by).... both tracking learners’ behaviors in real-time and dynamically adapting content based on each learner’s individual clickstream history." A variant of this approach is already seen in likes of Khan Academy.

2) Human driven : By populating an online classroom with adequate number of mentors and close involvement of professors or TA's along with cohort driven peer evaluations to effectively deal with the fundamental problem of lack of feedback and collaboration in an asynchronous learning environment.