Hunch, which helps people make decisions and learns from their responses to questions. When Hunch opens to the public on Monday, more than 40,000 people will have already answered 7 million questions over months of private testing. Hunch thinks it knows people pretty well already based on that testing, and it's only going to get smarter - about you.Flickr co-founder, Caterina Fake, has co-founded another startup called
In addition to decision-type questions, there are lots of fun personal profile questions asked of users - and patterns emerge. Amongst the 79% of respondents who don't mind drinking from public water fountains, for example, Hunch found that they are most likely to be willing to risk their lives for a stranger, always wear bike helmets and say that ethics are more important than success. The 21% of respondents who say they are disgusted by public water fountains are most likely to say they would not risk their lives for a stranger, they rarely or never wear bike helmets, say that if they were one of the seven deadly sins they would be Lust and believe that success is more important than ethics. That's just the beginning of what Hunch says it knows.
The ostensible purpose of Hunch is to help users make decisions. Unlike traditional Question and Answer services that act more like forums for posting on and don't remember the information that's been shared in the past, Hunch is much more systematic. It uses collaborative decision-trees to tell you what decisions other people like you have made in circumstances similar to your own. Hunch tries to avoid keeping people trapped inside options they already agree with by introducing a "wild card" answer to each question.
Should you buy or lease a car? Should you call your parents right now? What kind of blue jeans should you buy? What ancient language should you learn? Should you hire an architect? Those are some of the questions that have already been well developed on Hunch.
The system has a built in feedback loop that helps build out more possible answers to questions, more details to consider in making decisions and a better understanding of people as more people interact with the system.
The site will open to the public at large on Monday - "Early," Caterina Fake says vaguely, she's not running a traditional PR campaign in need of publicity. In addition to users being able to ask and teach about more than 2000 different topics, developers will be able to draw aggregate data out of the system through an Application Programming Interface (API). That sounds like a whole lot of fun.
Here are a few interesting correlations Hunch has discovered about people so far.
People who say they eat fresh fruit daily are most likely to...
- eat breakfast during the week r = 0.229
- spend most of their money on other people r = 0.188
- consider themselves feminists r = 0.181
- not homeschool their children r = 0.172
- use four or more ring tones on their cellphones r = 0.164
People who could not stack their collection of books up higher than their TV are most likely to....
- not be in the middle of reading a book right now r = 0.315
- never buy local produce at a farmers' market r = 0.226
- not consider themselves feminists r = 0.209
- believe in the death penalty r = 0.189
- have never attended a political protest, march or rally r = 0.181
People who were the oldest children in their families are most likely to...
- prefer to have ten to 20 acquaintances than just a few good friends r = 0.278
- have ten or more browser tabs usually open at once r = 0.227
- believe in a "less well known religion" r = 0.180
- believe prostitution should be legal r = 0.159
- currently be in the middle of a book r = 0.138
Middle Children are most likely to...
- not have a cellphone r = 0.259
- have never had their hearts broken r = 0.205
- always wear a bike helmet r = 0.172
- say Ringo Starr was their favorite Beatle r = 0.131
Youngest Children are most likely to...
- prefer having one or two close friends, over a large number of acquaintances r = 0.210
- say that if they were one of the seven deadly sins, they would be Envy r = 0.175
- not currently be in the middle of a book r = 0.137
This is fun stuff, a great example of what can be done with aggregate data analysis. It's fantastic that Hunch is opening up an API right away. An iPhone app for this already compelling service would consume countless hours is someone built it on top of the data Hunch is making available.
Watch for the site to launch publicly on Monday. In the meantime, over the weekend, please remember that the strongest correlation Hunch found among people who say that "Sundays are for working" - is that they also say they have never been in love.