I think the Wikipedia-type model, if evolves, is dangerous for Aardvark. It will lead them to become ChaCha, and that’s already taken and, well, not massively successful. Myself and everyone around me who also try it out, keep sending the “try” command to get questions, mostly for curiosity but also because many questions are simply not relevant (e.g. tons of irrelevant local questions).

So I would think they should proceed with the original idea and work to deepen each user’s social graph acquisition and refining, and focus on improving the experience for friend-of-friend communication (who is this person? why did I get their question??)

Search – I’m sure they will get to that soon, as FriendFeed did, first they want to accumulate data and learn its patterns. Complimentary – you’re right …

What matters is how it’s used and what the service promises to be. It’s true that friends-of-friends might not know you, so they may not provide as personalized of a response as your direct friends could. However, I’ve been told by the Aardvark crew that people answer WAY more questions than they ask, and that one of their primary requests from users is to be given more questions to answer! (This makes me think that a Wikipedia-type model is at work here.)

I also agree that one of their biggest drawbacks is that there is no way to search over previously asked questions. Even then, though, the database will likely be limited to inherently subjective questions that pertain to immediate information needs of users that might not be applicable a week or a month from when the question was asked. Obviously, this is fully dependent on the type of questions being asked and indexed.

What’s particularly interesting to me is that Delver & Aardvark have two complimentary styles of “social search” — and from my latest research, I see good reason to have both models play an active role in web search. Maybe you guys should team up!