Interactive lessons and Socratic style teaching, for starters. My best courses were all relatively small (10-20 people) and based around guided discovery. Technology could replicate some of that, but there are two problems for scaling -- it requires an interactive teacher that can respond not only to right answers, but wrong ones, and the group size cannot become too large. So you could do it online, but the 'M' in MOOC gets in the way.
Then there's the less class focused aspects -- office hours, homework groups, meeting classmates, etc. Some of this is replicated online, bit it would take dozens of emails or forum posts to replicate 30 minutes of one-on-one discussion with a professor (or even classmate). Much of the value in my education came out of study sessions and being exposed to other people's understandings and interpretations and needing to learn to explain my own. Small study groups using online collaboration tools (eg, voip and whiteboard apps) help with that, but MOOCs don't have the same tendancy to encourage that interaction that real classes do (and I'm not sure I buy online is the same as in person -- and I say that having worked remote before).
I went to an engineering school that had a very difficult time with their drop out rate. One of the biggest indicators of success, meaning graduation, was the students incoming math ability. Basically, if the student came in and took anything less than calculus when they first arrived, it was more likely than not that they'd drop out. Eventually, the math department added mandatory attendance to all lower level math courses and the affect was dramatic. The graduation rate went way up for these students.
Now, graduation rate is only one possible metric. That said, many universities do have hard data and that data generally states that classroom attendance leads to better outcomes overall. Certainly, that doesn't work for all students, but it does suggest that there still is a tangible, substantial difference between being in class and not.
