Hunch: A Cure for Indecision?

Continued from page 1

Question time: Hunch asks users several questions to refine their original query. Credit: Question time: Hunch asks users several questions to refine their original query.

In addition to rating whether a result was useful, users can suggest other recommendations or improvements to surveys. New query topics can be added to a "workshop" area to be developed until users' votes indicate that they are accurate enough to be released as questions on the wider site.

John Riedl, a professor of computer science at the University of Minnesota, who studies recommendation systems and online collaboration, says that Hunch is tackling a fascinating problem, and one that academic researchers have been working on for a long time. Computers are great for creating games like 20 Questions, he says; Hunch, however, is dealing with much more amorphous problems. A single question might have dozens of possible results, and the information that a user provides may not help a computer distinguish clearly among the options.

A key issue, Riedl says, will be whether the site can build up a base of volunteers who are willing to contribute. While projects that rely on user-generated content--such as Wikipedia--represent "some of the extraordinary accomplishments of our time," Riedl says, Hunch is asking a lot of its users.

Hunch has yet to answer the question of how it could make money, but Fake says that revenue will likely come from sponsored links that appear alongside results. Or if Hunch suggests a particular brand of laptop, the site might get a referral fee if the user goes ahead and buys the device. However, Fake stresses that advertising would only appear after a result is generated and wouldn't influence the products that the site suggests.

As the site improves, Fake hopes that the results will feel increasingly intuitive, appropriate, and even a little mysterious. She says, "I want it to feel like the Magic 8 Ball experience."