For one of my classes I recently had to suggest a new search engine feature then make a pitch about it. That was what started the cree.pr project. Most of the pitches weren't that interesting (no offense to anyone in the class reading this), but after hearing a whole bunch of them I had a revelation: Social search is a crappy idea.
Here by social search I mean something where the results shown are somehow influenced by what previous users have done. Now, it's not the case that social search is useless. There are entire communities built upon finding out what other people have done, and using the "wisdom of the crowd" to just take the cream of the cream.
But aggregating social information to make search results better is a crappy suggestion, because it's one of the defaults things you can do to improve search. It is the same thing as saying in a science experiement that "more data would have made the results more accurate". Well, duh. Similar suggestions would be to use more CPUs or build more datahubs.
So people, please actually think of something partially innovative next time.