One fascinating thing Ross Schachter does in his decision analysis courses is he makes people put weights on multiple choice answers. In the beginning of every class he goes over the stats for the previous batch. OVERCONFIDENCE GETS SEVERELY PENALIZED. It's just about the only thing I remember from the course (goes over stuff very similar to Vaniver's sequence http://lesswrong.com/user/Vaniver/submitted/ - needs a better link), but I thought it was awesome.