Web & communication software

Call It a Hunch (or Bing), But No One Is That Indecisive

Apparently Hunch uses this information to group you with other folks who answered the same way. Later, when you ask Hunch a question (Should I marry this person? Should I eat Chinese food for lunch? Why are we born to suffer and die?) it will weight responses from this group higher, assuming they also asked these questions.

Does your head hurt yet? Because mine does. Reading the site's FAQ is only slightly more illuminating. Ultimately it all comes down to "Trust us, we're from MIT and a lot smarter than you."

In actual practice, though, Hunch isn't really all that smart. In fact, in some cases it seems to have gotten hit upside the head with a stupid stick.

Ask Hunch "Should I get a divorce," for example, and the first thing Hunch will ask you is "Are you married?" Answer "No," and it continues on with another question. (Houston, I think we have a problem.) Answer "Yes," and it will run through another seven or eight questions, then provide you with an answer, as well as the percentage of like-minded people who answered the same way. Ask it what you should have for lunch, and it will guide you through a similar series of questions, then provide you with four options.

When the answer involves a product (like, What running shoes should I buy?) Hunch might link you to an external site selling that product, which is how it makes its money.

The problems? The answers aren't exactly rocket science. Unless you have zero friends, you're likely to get as good (and probably better) advice from an actual human. Worse, you can't just type any question into Hunch and get an answer. You either have to pick from questions someone has already submitted or create a new one from scratch -- tediously, step by step, question by question. And if you already know the answer ... you don't need to ask the question in the first place, do you?

The other problem is what Hunch could do with all that bizarre-yet-highly-personal information it has gathered about users' alien abduction theories and clown preferences. The company says it won't share this info with any other humans, and I believe it. But if in a year or two Hunch goes down the toilet, that data may be its only real tangible asset. The temptation to sell it could be overwhelming. And then? The next thing you know Kang and Kodos will be knocking on your door, looking to run a few tests.

If enough people ever use it, Hunch might one day provide moderately useful insight into the (alleged) wisdom of crowds around a particular topic. At this point in time, though, Hunch is really more of a massive experiment in machine learning -- and we're the guinea pigs.

Meanwhile, it's lunch time. Gee, I wonder what there is to eat....

Would you rely on a Web site to help you make big (or even little) decisions? Post your thoughts below or e-mail me: cringe@infoworld.com.

Subscribe to the Daily Downloads Newsletter

Comments