Sign in to view
member only content

When it’s time to make an important decision, technology can help consumers on the research end, but it’s humans they typically turn to for practical advice. There’s no substitute for a trusted advisor familiar with one’s tastes and preferences—or, at least, there wasn’t until recently. Led in part by Flickr cofounder Caterina Fake, Hunch is a brand-new decision-making tool that gets to know the user first and then offers customised suggestions. Users of the site—which just opened its doors to the public on Friday—can ask its help in making any decision, whether it’s “what dinner recipe should I make?” or “should I send my kids to private school?”. They begin by telling Hunch a little bit about themselves through an introductory set of questions—”Where is your home located?,” for example (suburbs, rural or city), and “Do you like bumper cars?”. Then, when it comes time to make a decision, a core algorithm based on machine learning asks the user up to 10 structured questions on the topic, any of which can be skipped at the user’s request. Using those answers—along with what it already knows about the user’s particular preferences—the system proposes a customised solution. Accompanying each decision is an explanation of how Hunch arrived at it, and users can vote on whether they agree with the result, as well as suggesting new topics and questions. In that way, the system gets smarter over time—almost Wikipedia-style—reflecting the corrections and suggestions of users. Contributions to the site earn credibility points in the form of “banjos” and badges for users. Meanwhile, as it learns more about each individual user’s personality and preferences, Hunch also further refines its decision results for that person. “It’s like a friend getting to know someone’s taste and preferences over time, so they can provide sound and trusted advice,” as the site puts it. Overall, “our long-term goal is for a user to be able to come to Hunch with any decision she is pondering, and after answering a handful of questions, get as good a decision as if she had interviewed a group of knowledgeable people or done hours of careful research online.” Some Hunch decision result pages include links to external commerce sites, in which case the site earns referral fees from the linked merchants—but such links have no effect on the decision results, the site says. Some 500 decision topics, 5,000 follow-up questions and more than 30,000 possible decision outcomes are already available on Hunch, with new ones being added every day. And while the New York-based site is currently available only to people who request an invitation and create an account, that requirement will be lifted in May. It’s early days on this one, but the potential is compelling—one to try out, partner with, or otherwise get involved in…? (Related: Private Klusters help groups make decisions.)