About 62,000 Hunch users have answered the following Teach Hunch About You (THAY) question:

Here’s how they answered.  Basically, you Hunch users love Wired, and then you go for Time and People, trailing off after that.  Only a small percentage (<1%) choose the AARP magazine. (That’s because Hunch doesn’t yet have a large audience of people in that age group… we’re working on that.)  Pretty logical.

You all are a bunch of Wired geeks (although you do have *some* other interests as well)

Here’s another THAY, which has been answered by about 64,000 Hunch users:

Make a decision...please

Now, let’s cross correlate the two questions to see how music genre choice varies (or not) by the magazine people choose.  You’ll see that for the most part (but with some exceptions), the overall ranking is pretty consistent.  That is, most people say “It depends”, then they choose Rock, ‘Other music’ (any idea what that might be?), Dance, Pop, etc.

It’s somewhat surprising that so many people answered “Just depends” when the question asked for a favorite choice.  Just goes to show that music is really about state of mind and many people have a hard time nailing down a single answer.

Overall ranking of music genre preference is very consistent

Let’s take another look at how well people were able to make a decision when asked about their favorite music. Turns out that people who prefer Sports Illustrated are the least likely to say “it depends.”  They know what they want (dammit) and aren’t afraid to say it.  The most wishy-washy group who were the most like to say “It depends?” That would be those who prefer Vogue.

Note that we’ve pulled the oldest stats trick in the book and magnified the vertical scale in the following chart to better illustrate our point.

Make a decision, please!

Now let’s take a look at the data a different way, creating indices for each magazine audience’s affinity for each type of music, compared to the overall Hunch user preferences.  Let’s also normalize for the “it depends” people, taking them out of the mix for each magazine group and grossing up the rest of the numbers so that they still add to 100%

So in the following chart, if 10% of Hunch users overall prefer pop, but 20% of the people choosing People Magazine prefer pop, that’s an index of 20%/10%*100 = 200.  Straightforward enough?

Mmmmm...indices

A few observations on the chart above:

• Green cells represent particularly high indices.  So if you see someone reading the AARP Magazine or Reader’s Digest, chances are they have classical music playing on their headphones.
• Yellow cells represent particularly low indices. A Sports Illustrated fan is likely to smack you upside the head if you put on the classical station in the car (they’re more likely to thank you for Hip Hop, though).  For People magazine readers, avoid Jazz.   And for Wired magazine readers, avoid Hip Hop. When it comes to pop, play it for the People and Vogue readers, not the AARP crowd.
• The secondary table at the bottom looks at the average, min., and max. of the indices.  The indices aren’t weighted to take into account the size of each group/cell, but this is still an interesting exercise.  Note that when you look at the spread between the highest and lowest index, there’s a big range. That spread is essentially a measure of how homogeneous or heterogeneous each magazine’s audience is when it comes to their expressed taste in music.  You can see that Car & Driver and Time readers have a low spread; that is, their indices tend to be not too far from the norm.  People and Sports Illustrated showed much wider spreads, showing their audience has tastes that are more divergent than the average.

Note that the above analysis was based on THAY questions. Hunch has dozens of actual topics which go much deeper in terms of recommending magazines, books, and music you might like.  Take a browse through Hunch’s various media categories and give it a go.

In recent weeks, Hunch’s content team has been re-reviewing every contribution made to the site.  Looking back over each of them was a sentimental duty with plenty of old favorites popping up.  (I’m looking at you, Butter Battle).  During this review effort, we specifically sought out which uses of Hunch have been the most successful at providing both quality recommendations and attracting top-notch, creative contributions.  We figure that if we focus Hunch.com on the topics and uses that best accomplish these two goals, the site will be on a winning path.

That is why today we are making a change to the type of topics which will be on Hunch.com: We’ll be removing Yes/No topics (also known as Either/Or topics) from the site, as well as what we call ‘continuous scale’ topics which are generally formulated as “How much [x] should I [y]?”

The reason? Yes/No and Either/Or topics don’t fit well with Hunch’s focus on taste-based, personalized recommendations.  Binary topics also tend to be more diagnostic and factual (which is not the point of Hunch), and frankly much of their content is usually better presented in a broader context. For example, a topic like “Should I watch the movie Scarface?” should really just be one of many results in a broader Movies topic.

Similarly, the “How much [x] should I [y]?” formulation tends to be primarily a factual (and often mathematical) function of many complex inputs (consider “How much should I invest?”, “When should I retire?”, How much should I raise for my startup?”, or “How much computer memory do I need?”)  These are good questions, but again, they don’t lend themselves to Hunch’s core algorithm and its focus on personalized recommendations that are driven by taste.

If you have previously created a Yes/No or “How much…” topic, you will keep any banjos you earned as a result of creating the topic and having people use it, but the topic won’t generate any new banjos for you since it won’t be accessible on the site any more.

We know it’s a bit of a drag to create something and then not have it last forever.  We (as staff) have also created many of these topics which will be retired, so we’ll be waving a nostalgic goodbye to many of our own creations as well.  We tip our glass in thanks to your contributions and hope you understand the rationale for this change.

Finally, we’ll also be streamlining the ‘Create topic’ wizard so that the ‘Yes/No’ option in the first step is removed.  Going forward, all topics will thus be of the same type: a recommendation (person, place, or thing) from amongst a set of more than 2 possible results.

We hope that the community of contributors on the site will join us in seeing that this change is a positive step toward simplifying Hunch and focusing on what we do best.

If you have questions or comments about this change, you can post in this forum.

Fortunately, late season travel deals reward the impulse traveler. In the U.S., Expedia.com will mail you a $50 debit card reward when you book three or more nights at a hotel between now and Labor Day weekend. And travel websites such as cheapflights.com, kayak.com, or fly.com can help you scour the Internet to find affordable Labor Day weekend getaways.

Travel is one of the robust categories on Hunch, with more than 400 topics — on everything from honeymoon destinations to airline frequent flyer programs — fully developed and ready to offer recommendations.

We’ve already mined Hunch’s travel data for two previous posts. In the first we redefined the notion of “escape.” In the second, we developed traveler profiles based on their vacation preferences for beaches, historic cities, urban cities or way out there places. (If you’re at all curious about who prefers USA Today over the New York Times on holiday, or which travelers are most likely to be technoheads, this one’s a must-read.)

Always curious about Hunch’s users, we took a look this time at some of the most popular Travel topics to see what sorts of preferences skew in a particular way among specific demographics.

Male vs. Female

Does it sound just too stereotypical to say that men skew towards preferring an activity-focused vacation, such as skiing, or staying home and do nothing at all, while women on Hunch skew towards being pampered at a spa for a weekend or visiting family?

Possibly, but that’s indeed what the numbers tend to say. And it makes us even more curious about the less obvious findings, such as the fact that women enjoy heading to Europe, to civilized countries including Belgium, Ireland and the cradle of civilization, Greece, while men show interest in some of the more remote areas of the world: Maine, Chile, the Czech Republic and even space (Virgin Galactic).

Vacation-related recommendations on Hunch that skew significantly male or female:

Ages 18-34 vs Ages 50-64

Looking at the vacation aspirations of these two age-focused demographics, it’s fair to say that both groups want to get out of town — the difference being that the younger ones, the 18-34 year olds, are heading to international destinations that are overflowing with culture, cuisine and nightlife. On the target list: Rio de Janeiro, S.E. Asia, Germany, New York.

On the other hand, the demographic with a few more years (and probably a lot more traveling) under its belt is more interested in remote escapes with sublime natural beauty: Banff National Park, Alaska, a B&B in the country. Indeed, there’s something to be said for the thoughtful reflection that the vastness of nature can inspire.

Lives in North America vs. Lives in Europe

Not all the data here was robust enough to make a true head-to-head comparison, but we weren’t surprised to find out that the U.S. states Europeans want to visit first are those bicoastal bookends, California and New York. In other categories, Europeans revealed their spirit for adventure, showing interest in such disparate destinations as Osaka, Finland, Luang Prabang, and Laos. How’s that for a multi-pronged trip?

North Americans, on the other hand, keep it closer to home, opting for a road trip or a visit to the U.S. capital. When they do get abroad, their sights are set on some of the closest possible destinations, including Belgium and the United Kingdom — or maybe a Caribbean cruise.

Ready to get away? Log into Hunch to get personalized travel recommendations that are customized just for you.

• Registered Hunch users have answered more than 50 million “Teach Hunch About You” questions.
• There are currently more than 60 million distinct “items” in the taste graph.  An item can be a person, a place, an opinion, a movie, a pair of shoes…or just about anything that you might either agree with, disagree with, like or dislike.
• There are more than 10 BILLION connections (proper math terminology would denote them “edges”) in Hunch’s taste graph. An example of a connection/edge is the fact that a particular person likes the Honda Civic, or whether someone happens to be a vegetarian. It’s these edges that give Hunch the ability to make smart predictions even when it only knows a tiny bit about someone.

Since we changed Hunch to login only, our overall site traffic has dropped but the number of users registering daily has tripled to about 3,000 per day, growing aggregate accounts by about 15% every month.  The accuracy of recommendations has gotten a lot better since Hunch is much smarter when users have an account.  Here are some recent tweets to that effect:

And since Hunch continues to get smarter the more it collects feedback, recommendations will become even better as more users come on board.  We will also be announcing a number of partnership deals soon where Hunch’s taste graph will help personalize other websites.

Hunch is all about correlations: we personalize the internet by discovering the relationships between people’s preferences. If you like spaghetti, do you prefer romantic comedies? By asking enough questions and comparing your answers to thousands of other responses, we might be able to predict the kind of person you’ll find most attractive.

That being said, most Hunch users claim to be above such superficial stuff. In fact, a whopping 77% would rather be intelligent than attractive.

When it comes to attraction, people speak colloquially of having a “type”. Some people prefer brunettes and others are struck by redheads. As superficial as it may be, attraction seems to matter: research shows that marriages in which the woman is slightly more attractive than her husband last the longest, and that seeing an attractive person touch an object makes that object more appealing.

We also know that the environment plays a role in what we find attractive. We might seek out a mate with the traits we admired in our parents (or with those our parents were missing). Women born to older parents tend to be less impressed by signs of youth in a potential partner’s face than their counterparts born to younger parents, for example.

The eye of the beholder

Much about attraction seems to be individualized, or “in the eye of the beholder,” and some features considered beautiful change as fashions wax and wane. Yet there do seem to exist some universally agreed upon elements of beauty: symmetry is general seen to increase ratings of health and sometimes of facial attractiveness.

At the same time, a great deal of pseudo-science surrounds any attempt to quantify attraction. From theories that attempt to link our pre-historic inclinations to current motivations, to those colorful pills by the convenience store cashier that claim aphrodisiac qualities, the world is full of people claiming inside information about the “secrets of attraction”.

Attraction deconstructed

Hunch is no pill peddler, but we do have a virtual gaggle of responses about what people find attractive. Now, on to the burning question: who do you find attractive? First, let’s look at some faces:

Based on the responses of more than 200,000 Hunch users, almost half (47%) claimed Eva was the most attractive in their eyes, with George a distant second:

"Who do you find most attractive?" All responses

I may love you, man (but I’m not going to admit it)

If we look at the responses broken down by gender, however, we can begin to see what’s driving these percentages. In total, about 60% of Hunch users report their gender as male. And it’s the men who are the biggest fans of Eva. Overall, men show a 93% preference for one of the two women.

Hunch doesn’t have a direct Teach Hunch About You question which asks about sexual orientation, but for the sake of argument let’s assume that 10% of both men and women on Hunch prefer same-sex partners.

Only about 8% of men on Hunch chose either of the two men as most attractive, while nearly 26% of women chose one of the females as being most attractive. Our conclusion? Women, straight or not straight, are simply much more comfortable than men are in acknowledging that someone of their same gender is attractive.

As far as individual preferences among female Hunch users, George catapults to the top, enjoying a 50% preference advantage over and above Mario, compared to a less than 30% gain when both genders are considered (it turns out that men, in fact, like Mario more than George). Both men and women prefer Eva to Nicole, and by about the same margins.

"Who do you find most attractive?" Breakdown by gender (male top, female bottom)

Do gentlemen prefer blondes?

We next consider what general features people find attractive. With regard to hair color, gentlemen do indeed prefer blondes, at least on a relative basis when compared to the preferences of their female counterparts (by a 2 to 1 ratio). But men like redheads even more: here, the ratio is 3 to 1. On the whole, however, both genders prefer brunettes. And more women than men claim that their tastes are “broader” than mere hair color.

More striking is something that’s been confirmed by scientific study of groups and attraction: we tend to be most attracted to people who share some of our features. Blondes like other blondes relative to their peers with differently colored hair, brunettes like brunettes, and redheads redheads. Grey- and white- haired folk also give blondes a comparative advantage, whether because their own hair is similar in hue, because these folks (who, presumably, are older) prefer a bygone fashion, or for some other reason.

In the table below, people with differently colored hair (along top row) prefer different colors of hair in others (listed in leftmost column). Individuals give a boost to those with their own hair type in rating attractiveness. Notice how the highlighted cells show a greater preference percentage than other cells on the same row.

Birds of a feather flock together

A Fine Romance

Back to our celebrities. What does finding one of these people attractive say about your own relationship status? Well, if you’re anything like other Hunch users, your selection of Mario as the most attractive is likelier to signal a troubled relationship or state of singlehood. Those in troubled relationships are also likely to have the hots for George. On the other hand, if you’re relationship is “going great” or even “just ok” you might like Nicole.  Of course, we don’t discount the possibility of a third variable, such as the tendency of women or men to claim troubled relationships, as a factor driving these correlations.

Do opposites attract?

We’ve already seen above that people rate as attractive those who share similar features to their own. Yet Hunch users believe that opposites attract: about 70% of those surveyed answered this question affirmatively. It’s a fine thing to believe in, and not necessarily inconsistent with the first set of responses (the question could be read as “opposite personalities”, rather than looks, attract).

Those who believe that opposites attract are more likely to find Eva good-looking, while those who think that birds of a feather flock together would like to nest with George. Similarly, risk-takers (those who say they’ve been skydiving) like Eva and the risk-averse skew towards George. Again, we have not controlled for gender here: being male explains at least in part both a preference for Eva and a propensity for risk.

You might be wondering about the role of intelligence when it comes to attraction. Almost 80% of Hunch users say they’d rather be intelligent than attractive: that figure is slightly higher for men than for women (and psychologists have shown that status and intelligence are more important to women in mate-selection, while looks are more important to men).

But the smartest cookies are the ones least likely to believe that opposites attract, perhaps because they don’t want to date someone who’s dumb as rocks. Of those in the top 5% of their high school class, 29% believe opposites don’t attract (versus 64% who believe they do). Among those who report being in the bottom 50% of their high school class, 73% believe opposites attract (and only 23% believe they don’t).

Wondering what your object of attraction says about you?  Try some of Hunch’s relationship-oriented topics to see for yourself.

It’s not that we don’t want other languages on Hunch; it’s just that our content teams are not able to review every language in the world, so we couldn’t ensure that non-English topics would have the same quality standards we normally encourage and enforce in English topics.

At the same time, we wanted to be sure that we would only show non-English language content to users who were likely to want it and understand it.

We think we’ve found a good compromise to address the issues above:

• Non-English topics will generally be promoted from the Workshop to the main Hunch Corpus.  However, those topics will have a “U” icon next to them to indicate that they are unreviewed/unmoderated.  So use at your own risk!  They might have stylistic or content issues which you wouldn’t normally see in a promoted English language topic.
• We will attempt to match users to languages they understand, and will show you foreign language content when we find a match. We’ll do this in a couple of ways: by checking the language code which your browser sends when it connects to Hunch, or by noting if you’ve ever created or contributed to a topic in a language other than English.
• If Hunch thinks it finds a match for a non-English language you understand, it will generate a Hunchmail to let you know.  You can change the language setting, add additional languages, or opt out completely from non-English content by clicking ‘Settings’ on the top of any page, and then visiting the ‘Preferences’ tab to adjust your language settings.
• Because non-English topics are generally not reviewed or edited by Hunch staff, the Hunch community could doubly use your support to jump in and edit, correct, or adjust non-English content which you think could be improved.
• Finally, note that this is just an initial baby step.  For the time being we will not be actually translating any topics (or THAY questions) to or from any language, nor are we going to localize the site itself into other languages in the near-term.  But we’re glad that original non-English content now has a rightful place to live within Hunch!

Many more refinements coming soon!

Author Devin Leonard spent a lot of substantive time with us while researching the article.  And in an impressive display of “this is how real journalism is done”, Wired sent us around 150 fact points to check, giving us the opportunity to confirm or clarify each. We were very happy with the result, and we hope you like the article.

Grab, get, or make some Hunch SWAG (anything with the Hunch logo on it), take a picture of it on yourself, someone else, or just about anything, and post the photo on a public website (such as Flickr, or your own blog).

Then, just send us an email by September 3, 2010 to contest@hunch.com to let us know the link where to find your photo(s).  Photos will be judged on originality, creativity, and cleverness.  So please: be original, creative, and clever! We’ll announce a winner by September 10, 2010.

You can find more info on the main contest page, or if you relish in wading through long legalese, check out the official rules. Happy SWAGging and photographing.

Here’s a very quick peek at some findings:

• If you want to make a crowd-pleasing sandwich, note that on the whole people prefer hard cheese, romaine lettuce and multigrain bread.
• However, if you come across someone who professes a love for moldy cheese, you might serve him a sandwich with arugula. If a friend likes red leaf lettuce, it’s a fair bet he’ll also be partial to white bread.
• When it comes to pairing fries with beer, consider steak fries with pale ale, bistro frites with dark ale, or McDonald’s fries with no beer at all.

For more graphics and fun food clustering facts, check out the full report in HTML or PDF version. And don’t forget to check out some of the hundreds of food-related topics on Hunch to see how your own preferences translate into Hunch’s food recommendations for you.