I would choose option a. The most effective way to test the hypothesis that Facebook impacts happiness is to assign subjects to either Facebook or no Facebook groups at random so that all relevant variables can be assumed equal except for the one you control - Facebook exposure. (Of course, you would also have to be sure the subjects do as they are told; i.e., the no Facebook group cannot use Facebook for the duration of the study.) You then have to gather data on happiness ... which means finding a valid measure of happiness.
