Caterina Fake personalizes the Internet (again)

Caterina Fake is back. The Co-founder of Flickr, the landmark photography site that helped launch the era of user-generated content and popularized group dynamics that influenced services like Facebook, Twitter and Foursquare is on a new mission. "One of the overarching goals of my career has been to make technology more human," she says. "You should be able to feel the presence of other people on the Internet."

With Hunch, the DLD friend is using these social skills to tackle one of the most vexing problems in computer science: online recommendations.

The Fakeian solution is to get people talking about themselves - their opinions, tastes, beliefs, idiosyncrasies. Then, once they have shared enough information, mine that data for correlations that provide precisely tailored recommendations for each user. It is a quietly radical premise, implying that our tastes are defined not only by what we buy or what we've liked in the past but by who we are as people.

Since the site launched in June 2009, it has collected 55 million answers to the "Teach Hunch About You" questions from its 1 million active users. Once Hunch's algorithm collects enough data, it can start finding surprising correlations. For instance, People who believe in alien abductions are more likely than nonbelievers to drink Pepsi.

Hunch - which Caterina cofounded with entrepreneur and tech investor Chris Dixon- is at the forefront of a new wave of so-called social search engines that combine user-generated content with powerful algorithms. But for Hunch to truly succeed, it needs more data. So Hunch is scouring the Web for information, combing the databases of social sites like Facebook and Twitter for anything that's publicly available - opinions and allegiances, likes and dislikes, followers and friend requests. Hunch is learning whether people who identify themselves as liberals on Facebook share similar dining preferences. It is learning whether people who follow Glenn Beck on Twitter like the same kind of cars. The idea behind this treasure hunt is fairly simple - the more sophisticated and thorough Hunch's taste data becomes, the better its recommendations. Like Google, Hunch's business model relies on sponsored links, which appear alongside its recommendations; the company gets paid whenever a user clicks on one.

"The ultimate goal of the company is to map every person on the Internet to every object on the Internet, be that a product, a service, or a person," Caterina says.

Read more about Hunch in an older DLDnews entry and discover the primary article in the Wired Magazine here. Additionally, watch her in an interview at DLD 2007 below.