I'd like a way to find questions that might be of interest to me, that are not neccesarily marked with one of my "Interesting" tags.
Maybe something that mines viewing history like iTunes "Genius".
The idea is to correlate similar user viewing behaviours in order to provide "suggested reading". I'm presuming that the data dumps don't allow us to see who viewed what questions, due to the anonymity requirement.
We can already see which questions get the most views, fewest answers, and so on using existing features. If I want to answer questions I can browse under Unanswered (as several people have usefully suggested) for questions tagged with subjects I'm interested in. Most of the existing views are sorted lists, with the sort based on some metric (views, votes, etc.).
Perhaps there might be questions (even ones with answers) that would be interesting to me, but they don't have that many views or votes in total - they are in the long tail. But if a few users with similar viewing history to me look at those, then they would be valid items for the "suggested reading" list, especially if not in my preferred or ignored tag sets.
For the blissfully unaware, Apple's iTunes has an opt-in feature ("Genius") that uploads media play information to Apple central. Some algorithm then collates that against similar data from other users, and comes up with suggested items to purchase from the store.
What do you think?