Will Hunch Help You Make Decisions? Signs Point to Yes

Last week I wrote about Twitter, a flawed and difficult-to-grasp social media technology that nonetheless becomes addictive once you get the hang of it—so much so that it’s quickly changing the way many people communicate. This week I’m going to write about Hunch, a flawed and difficult-to-grasp social media technology that nonetheless becomes addictive once you get the hang of it—so much so that it’s bound to change the way many people make certain kinds of decisions.

The product of a New York City startup founded by Flickr co-creator Caterina Fake, Hunch is designed to help us all cope with the problem of choice. Where should I go on vacation? Should I drop out of college or get my degree? What cool new video game should I buy? What’s the best sleep aid for me? Which New York City museum should I visit?

For almost any personal decision, chances are that someone else has already thought it through and can list the leading possibilities. But while the Web offers plenty of community sites where you can solicit such advice—I reviewed a bunch of them, including Yahoo Answers, back in 2006—Hunch has added an ingenious twist. It’s the “decision tree,” an algorithm that guides you through a big choice by asking you to make lots of smaller, easier ones.

Hunch has decision trees for roughly 1,300 topics so far. Each poses a series of multiple-choice questions. Your answer at each point determines which branch of the tree you’ll follow, until you wind up at a single recommended answer. A simple decision might involve only one or two questions, while a complicated one can have a dozen or more. Decision trees are devised by users themselves—in fact, if you feel like you’re an expert on something, Hunch encourages you to build a tree yourself, or help improve existing trees by adding new questions.

A Hunch decision treeThere’s an important wrinkle, however, that makes exploring Hunch more than just a process of clicking through a bunch of mathematically preordained decision trees. The site remembers how you’ve answered other questions, and over time, it builds up a picture of your preferences. That information is factored into the final recommendation, and might even override your answer to a specific question within a tree.

As an example of all this, here are the questions you’ll see for the topic “Should I buy an Amazon Kindle?”—which, as regular readers of this column know, is a decision I’ve been struggling with myself.

• Do you have a commute that allows you to read during it?
• Do you frequently travel with more than two books in tow?
• Do you subscribe to any major newspapers in print versions?
• Do you wish you could dynamically resize the text of print publications?
• For now, photos appear in black and white on the Kindle. Is this ok?
• Are you concerned with conserving paper in order to save trees?
• Do you get a particular sense of satisfaction from storing a book you’ve read on a bookcase?
• Are you clumsy with personal electronic devices like cell phones?
• Does having quick access to a dictionary/wikipedia seem valuable?
• Do you have several books at once on your nightstand?

When I went through this tree mechanically answering “Yes” to every question, Hunch told me that there was a 92 percent chance that the right answer for me is “Yes, you should buy a Kindle.” I couldn’t find anything on the Hunch site that explains how these percentages are calculated, but I’m guessing that the other 8 percent represents the room for doubt left by my “Yes” answers to the seventh and eighth questions, which would militate against buying one of the e-book reading devices. (Or maybe Hunch knows somehow that I’ve been trying desperately to come up with reasons not to spend $350 on a Kindle.)

The Kindle question was constructed with a yes/no answer, but most decision trees on Hunch can lead to a variety of results. And if the site is missing an important result, you’re free to add it, and to specify which questions in the tree should lead to that result. (I couldn’t help adding my own name to the list of results for the question “Which technology writer would I like?”)

For certain questions, Hunch can hit surprisingly close to the target. When I played through the “Where should I go on vacation?” topic, Hunch guided me straight to the answer that was already at the top of my personal list: Florence. I wasn’t even trying to steer the answers toward Italy, at least not consciously. When I played through the topic “What’s the best dog breed for me?” I ended up, reassuringly, with Australian Shepherd—which is, of course, the breed I already own. I was less happy with Hunch’s answer to “Which superhero am I?”: Watchmen‘s Dr. Manhattan, who, while certainly a hunk, is too aloof for my taste.

But at this early stage, with so few people using the site, it would be hard to portray Hunch as a place to turn for consistently trustworthy recommendations. There just hasn’t been enough time for users to fill out the branches of the trees. The topic “What’s a good spa in Boston?” for example, has only four possible outcomes—which is embarrassingly incomplete when you consider that local review site Yelp lists more than 110 day spas around Boston. And some of the questions users have programmed into Hunch are so predictable and simplistic that they’re essentially rephrasings of common knowledge. For the topic “Where should I live in the Bay Area?” the first question in the tree is “Would you rather have: Great weather, with a subdued, suburban lifestyle, or iffy weather, but with an exciting, urban lifestyle?” To me, that’s precisely the same as asking “Would you rather live in Palo Alto or San Francisco?”

Still, the more people who use Hunch, the smarter it will get. The process may be slow—I suspect that building a truly useful decision tree is harder than it looks, and that it will take a while for Hunch to build up a community of volunteers with the requisite thoughtfulness and expertise. But it’s happened before. Just look at Wikipedia.

And Hunch’s general model feels new and exciting. My own prediction is that millions of users will be drawn to the site, which turns the potentially stressful process of reaching a decision into a fun, interactive quiz. The decision-tree format may not respect the subtlety and grayness of the real world; Hunch’s style guide insists that the answers to each question in a decision tree be mutually exclusive, which, in real life, they rarely are. But the trees do offer a convenient way to navigate through a mess of possibilities, and perhaps to reach unexpected and thought-provoking answers. And hey, it’s got to work better than a Magic 8-Ball.

Hunch was testing its system privately until March 27, when it took the lid off and started letting in a few outsiders. I got an invitation to open a beta account less than a week after requesting one, and from what I’m hearing, Hunch is responding to account requests even faster now. But if you want to try Hunch and they don’t send you an invitation right away, I’ve got a handful to give out. Be one of the first five people to write in to send your e-mail to [email protected], and I’ll send you one.

For a full list of my columns, check out the World Wide Wade Archive. You can also subscribe to the column via RSS or e-mail.

Wade Roush is a freelance science and technology journalist and the producer and host of the podcast Soonish. Follow @soonishpodcast

Trending on Xconomy

By posting a comment, you agree to our terms and conditions.

4 responses to “Will Hunch Help You Make Decisions? Signs Point to Yes”

  1. Brian Leng says:

    Does this network work?

  2. spencer says:

    Ha it sounds like the magic 8 ball of the internet.