American Radio Works has a nice series of 3 reports on the inadequacies of the traditional lecture, and novel teaching methods that take advantage of recent research in cognitive science. I think this is quite different from the online teaching advocated by Daphne Koller. In particular, I like the description of Eric Mazur’s peer instruction approach (I intend to get his book, but I would appreciate recommendations for other resources): The main idea is to first get students to read the textbook before coming to class, and make sure that they actually do so. They are then asked to give feedback on what they found confusing and difficult. The instructor then develops a set of questions that the students discuss in class. There seems to be good evidence that this approach works well. It would be interesting to see some control experiments to see how much of the success is due to the students just reading the assigned material beforehand. Perhaps forcing students to prepare for class, and than tailoring a traditional lecture so that it addresses the questions students identified beforehand would work just as well.
However, there is one point that is made in the article that I wholeheartedly agree with: Most of us at universities are paid for our teaching (typically 40% or more of our effort should be devoted to teaching, even at research universities). However, scientists receive virtually no training in education either before or after our PhD. It is still fairly typical for faculty members to just be given a book and thrown into the classroom when they are given a job. After that promotion is determined largely based on our scientific output and grant dollars generated. I find it surprising that university professors overall do a good job in the classroom given this environment.