“Hunch personalizes the internet by getting to know you and then making smart recommendations about what you might like.” — About Hunch

Yesterday I played around with Hunch for a couple of hours and by the end of the day I felt I still hadn’t realized any benefit. I’d told it a lot about me but it still didn’t seem to know anything about me. For example, given my responses on religion/philosophy, why would it show me any topics on Bible versus? Conversely, how could it miss my interest in Japan, gardening, and the Bloomsbury Group? (One of these topics is sure to come up within five minutes of anyone talking with me in life.) Hunch did detect my interests in design in general and on web usability and mid-twentieth century modern specifically. But it also seemed to think I was a gamer. Hunch knew I was more interested in photography than music. Even so, it recommended a topic covering Opera, the one kind of music I told it specifically I didn’t like.

I’m intrigued with Hunch’s mission to find the best content on the web for me. I always like taking quizzes about my likes and dislikes to get recommendations for what else I’d like. Rating movies to get recommendations was one of the things that drew me into Netflix. My tastes are so frequently outside my demographic that I’m curious to discover correlations other than gender and age.

Getting to Know You

Hunch learns about you in several ways. First you take a little multiple question quiz. I found the THAY (Teach Hunch About You) questions fun. There are over 1500 THAY questions. You have to answer 20 of them to get started. But because I like personality tests, I ended up answering 253 in my first go. By the end of the day, I’d answered 531. Hunch gives you the ability to edit or delete your answers, an option I find comforting.

Hunch also learns about your tastes by having you rate results. This makes more sense in some topics than others. It’s easy to rate movies, books, and other consumables. But how does one rate Myers-Briggs personality types? am I rating types of people I get along with? or am I rating the information provided about that type? And why isn’t Hunch smarter? If I rated a book in one “topic” (Classic Books), why doesn’t it remember my rating when it shows me the same book in another “topic” (Science Fiction Books).

Hunch also encourages you to rate other people comments on an item. Is this beginning to seem like a lot of work?

Hunch Demographics and Data Correlations

The THAY questions are amusing and some made me reflect. Even more fun, after you answer Hunch shows the percentage of people who answered the same way. The more I answered, the more I wondered about the Hunch demographic. Obviously they are a self-selecting group. (The experience reminded me of taking the Myers-Briggs test at ETI. The results were skewed because a huge percentage of the people who participated were intuitive thinkers–mostly because NT types love taking these kinds of tests. It would have been more useful if everyone in the company had taken it.) As I progressed through the THAY questions, I learned that among people currently on Hunch:

  • 81% think President Obama was born in the US and birthers should get a life
  • Only 50% believe in evolution
  • 46% are introverts
  • 38% believe time is a human construct, the same percentage of people (although not necessarily the same people) who visit their local libraries
  • 26% said they didn’t know what LGBT meant
  • 17% have eaten food that was still alive
  • 13% are divorced, 40% not divorced, 44% never married
  • 6% think Rush Limbaugh is a voice of reason
  • 4% don’t have a Facebook account

The data correlations are lots of fun to play with. According to Hunch, Republicans are correlated with Fiji Water and Democrats are often seen drinking Evian. Once again I’m an outlier; I drink San Pellegrino. I haven’t gotten to that question yet, so I don’t know if San Pellegrino is even a choice. Maybe a more insightful question would be Water: 1. tap; 2. bottled, still; 3. bottled, sparkling.

As I got deeper into the questions, I became more frustrated because I felt that none of the answers were applicable. For example, Religion: 1. Major religion; 2. Minor religion; 3. No religion/personal philosophy. I resented the assumption that if one had no religion that one had no personal philosophy. Or maybe I just misunderstood how the “/” is used in that answer.

More and more, I began refusing to answer questions unless I strongly agreed with one of the answers. I wanted to be able to flag the questions which I felt did not comply with Hunch’s guidelines. “Hunch forces users to choose one and only one answer to a question, so make sure that a question’s answers are truly mutually exclusive and applicable to everyone.” I began thinking more and more about the data correlations and how the data would be misleading if the choices offered did not accurately represent my opinions.

After digging around this morning, I found you can flag questions and even edit some of them–but these options don’t appear in all presentations of the cards. I noticed it only when I reopened a question to edit my initial answer.

User-Curated Topics

In playing around with the quizzes and ratings, I became completely distracted from the main promise of Hunch–that it would provide me with smart recommendations. Initially, I was confused and frustrated by the recommendations. I hadn’t read a single author, blog, or book that Hunch thought I’d like. Conversely none of my favorites were listed. Why did the results seem so off? I’d wanted to believe that Hunch would open up a whole new world for me to explore.

What I finally realized is that the results are limited because they rely on user-curated content, on the tastes and knowledge of other Hunch users. The recommendations are skewed toward the Hunch community’s interests and biases. And in areas where my interests are not that of the community, such as gardening, the results are appallingly bad. Anyone looking for gardening information would be better served reading a gardening book, a garden blog, or Googling information on a specific plant. The Internet contains a wealth of gardening information. But you wouldn’t guess that from the Hunch results.

Hunch has various strategies for the community to create and expand content. But I have zero interest in providing Hunch free content. All the cute badges in the world are not enough to entice me. I’d rather devote endless hours of work providing free content under my own brand. The one area I might spend some time is creating or editing THAY questions.

Bottom Line

Hunch is a visually attractive site. It’s one of the most beautiful sites I’ve visited lately: clean, white, uncluttered. And I enjoyed answering the THAY questions and seeing what percentage of people chose the same answer. Ultimately Hunch fails in its mission to find the best content on the web for me. It only knows what its users tell it. And that’s just a tiny chunk of the whole wide web. The early adopters will have to do a lot of work to make it robust enough to attract a mainstream audience.

I remain wary about content “designed” for me. Although I’m looking for relief from information overload, I’m very suspicious about being ghettoized–about being isolated by demographics and restricted from the larger world. A huge advantage of the Internet is that it enables us to interact with people, places, and ideas outside the normal scope of our lives. The idea of a “smart” web showing us only thing we’re interested in seems backwards–narrowing our world view rather than expanding it.