“Social Search” Generally Isn’t

In attending this SDForum SearchSIG event tuesday night, I was keen to learn what Wikia, FriendFeed, Mahalo, and Facebook were doing about social search. As it turns out - and as I pointed out during the Q&A to lively discussion (which I think was recorded but doesn’t seem to be posted yet) - the answer is nothing much.

My premise in making that comment was that social search be defined as a search deriving its results from “a statistically meaningful sample of people meaningfully related to me”. I gave as an example Zagat’s guide to NYC restaurants, which, back in the day, was exactly that - a usefully large group of, mostly, actual foodies.

Mahalo, imho, is simply an extension of the editorial approach into semi-pro range. Instead of one food editor of the NYTimes, you could have, I don’t know, 10?, editors of the NYC pages. Wikia, simply the cult of the amateur doing the same thing. In both cases, these are curated results pages. If the curators are competent, passionate, and/or otherwise motivated, this is great and a step forward from algorithmic results. But making better results pages ain’t social search.

Neither is telling me what my friends think. Sure twitter lazyweb is a great way to get a recommendation for an indian restaurant in Palo Alto, but it’s a lousy way to get one for a dentist in missoula. Especially if you don’t live there and have a bunch of friends there.

Even del.icio.us and other “social search 1.0″ tools are still, more or less, dumb boxes of votes. Those votes are by smart people and often people like me (hence why I find delicious popular interesting), but there’s no variability on a given query on the axis of “social”. And, like mahalo and wikia, the results are mostly url’s - i.e. links to other pages where you as often as not have to execute another search or dig through socially undifferentiated data to extract value (a link to a restaurant page on yelp, for example, with a bunch of reviews from people I don’t know or trust).

There are a bunch of tools that let you ask questions of people. LinkedIn does a great job of this, again within a specific community, but they are very clever in the way things can seep out beyond the first degree network without hitting the undifferentiated population of “everyone”. That said, it’s still an expansion of “me”, on the assumption that my business colleagues and their business colleagues are to some degree usefully alike. While somewhat true, this is not nearly as useful as a larger population of people “like me,” who might or might not be related to me socially.

Lijit is doing a similarly interesting and useful job of letting me search the corpus read by my network, and with a little help from MyBlogLog and del.icio.us, my network’s network. Lijit is awesome, and I often use it to search my own stuff.
But what I really would like to see from “social search” is something that can search my network/neighborhood AND search other neighborhoods like mine, where “like mine” is pivotable based on context (friends, business, geo, special need, etc.). Like last.fm for stuff other than music maybe.

(I’m barely commenting on Friendfeed and Facebook because, to date, those are seredipitous discovery tools more than search ones, and ultimately you’re mostly finding people, not useful data derived from groups of people.)

Is anyone doing anything interestingly like this?