Hunch.com articulated "taste graph" and what do you get? Interesting personalized restaurant recommendations, for one thing.Cross reference a person's Twitter friendships with their Foursquare favorites with their
Taste-gathering startup Hunch is experimenting with a recommendation service that cross references social graph connections on other services with the large set of unusual questions its users have answered. Questions like "do you like facial hair on men? Yes? Well, 48% of our users have said that." The end result is a simple prototype website where you enter a city and your Twitter username and Hunch will show you Foursquare venues it thinks you'll like. Or at least it thinks that people on Hunch who are like your friends on Twitter tend to like those places, on Foursquare. Crazy? Maybe not.
Restaurant recommendations are just the beginning. Hunch knows a lot about a lot of people. The company recently said that the average Hunch user has answered 152 personal questions about themselves. Now that data and our corresponding friend connections are going to be the basis for personalized recommendations. Want to see how well the company thinks it understands you? Check out the recently launched Hunch Twitter predictor game. It's downright eerie.
Hunch co-founder Chris Dixon explained (vaguely) what's going on by email.
We developed the technology to project and propagate our taste data using graph-like connections via public APIs. In this case we propagate our taste profiles to Twitter by projecting the subset of Hunch users connected with twitter onto all Twitter
users. Then we propagate this taste data to Foursquare by projecting the subset of Twitter users checking in on foursquare onto all Foursquare venues. With our collection of taste profiles, in real time we can calculate affinities between any Hunch user, Twitter user, and Foursquare venue. As we project and propagate across all the web's entities, we will enable crazy data mashups. It's going to be cool!
In other words, if Hunch doesn't know about you well enough to make Foursquare recommendations via a Twitter account that's tied to both Foursquare and Hunch, then it will assume you are like those Twitter friends of yours who are on Hunch, and Foursquare.
That's the kind of data-driven value that making all these connections explicit will allow. The future will look like a big algorithm and interface war between companies battling it out to better serve you based on commonly, publicly available user data. Or data you selectively expose in return for recommendations.