Might Know Your Personal Tastes Better Than You Do

A screen grab of one of's questions. i i

hide captionA screen grab of one of's questions.'s Caterina Fake says she thinks complex algorithms can be 90 percent accurate in determining people's personal taste.
A screen grab of one of's questions.

A screen grab of one of's questions.'s Caterina Fake says she thinks complex algorithms can be 90 percent accurate in determining people's personal taste.

There's a lot of information about you floating around the Web.

Facebook knows if you're a "fan" of NPR, the city you live in, your birthday, and your relationship status. Amazon knows what kind of books you like to read and Netflix knows the movies you watch.

Tech entrepreneurs are trying to take all of this personal information and create a more user-friendly Web experience. But can complex algorithms account for personal taste? Caterina Fake thinks they can be 90 percent accurate and tells All Things Considered's Robert Siegel about how her site,, is working to make that happen.

Wired Business Conference in Partnership with MDC Partners

hide captionCaterina Fake

Larry Busacca/Getty Images North America

Fake is the chief product officer for, but she made her name in the tech world by co-founding Flickr. Yahoo acquired Flickr in 2005 for $35 million.

Because of Flickr's success, Fake is partly responsible for the phrase "user-generated content." And what Fake loves most about the Web is the fact that amateurs can populate it. That's one of the first things she did in her early Internet days. "I noticed there was no Web page at all dedicated to Vladimir Nabakov — one of my favorite writers — and I threw up a page about him," she says.

Fake's new venture tries to figure out what its users will like — anything they will like: magazines, cars, movies, religion, books, recipes. And, here's where Fake's experience with and love of user-generated content comes in: When you sign up for, YOU have to answer questions about yourself in order to create a "taste profile."

Questions like:

  • Where do you live?
  • Would you setup a new home theater on your own?
  • Which fries would you prefer to munch?
  • Alien abductions are they real or fake?
  • Are you a mac person or a pc person?
  • Do you generally prefer your sandwiches to be cut vertically or diagonally?
  • Which lettuce would you usually prefer on a salad?
  • Do you like to dance?

"People answer a great number of these questions — 120 questions is the average number of questions that a Hunch user will answer, and it turns out that a lot of these questions are predictive in ways you would not expect," Fake says. "So whether or not you wear plain socks or striped socks might be predictive of whether or not you will like the movie Napoleon Dynamite."

Fake claims that because Hunch is expanding its data set by having users answer specific questions, the site can make better predictions than Amazon or Netflix. The site isn't just mining what you buy, but who you are — your personal likes and dislikes. And Fake is confident that Hunch can do that better than social networking sites like Facebook, sites that have access to quite a bit of personal data.

"Social networking sites are much more performative — your portrayal of yourself is much more aspirational," Fake says. "In Hunch's case, you're maybe someone who is going to be recommended an Acura, but on a social networking site you might portray yourself as someone who drives a Lamborghini."

Whether Hunch will come out a leader in this next phase of the Web remains to be seen. Personal taste is just that — personal, and so far no site has been able to figure out the right recipe to predict it.



Please keep your community civil. All comments must follow the Community rules and terms of use, and will be moderated prior to posting. NPR reserves the right to use the comments we receive, in whole or in part, and to use the commenter's name and location, in any medium. See also the Terms of Use, Privacy Policy and Community FAQ.

Support comes from: