Like little stars.
Mathematical matchmaking sounds like a modern method for finding love. In truth, it predates dating Web sites, the computer, the field of statistics, or even Europe. The practice goes back to 1500 BC and the composition of the four Veda scriptures, part of the ancient Indian religion of Brahmin that later became Hindu. While the rules governing Western astrology are open, discoverable, and thus easy to ridicule, Vedic astrology as traditionally practiced was esoteric. Expertise was passed among a small, elite group with intimate knowledge of the ancient Veda texts. These Brahmins had a monopoly on the business of drafting personality profiles, or janam kundalis, which all well‑to‑do parents had to have before they could arrange marriages for their children.
Vedic astrology puts forward a mathematical explanation for human personality that has logical and weighted underpinnings, even if it is fabricated. Paramahansa Yogananda, whose “Autobiography of a Yogi” was one of the first accounts of the religion to be translated into English, explained the connection between an individual personality and planetary movement as a causal relationship.
“Astrology is the study of man’s response to planetary stimuli. The stars have no conscious benevolence or animosity; they merely send forth positive and negative radiations. Of themselves, these do not help or harm humanity, but offer a lawful channel for the outward operation of cause-effect equilibriums which each man has set into motion in the past.”
Today, most of us in technologically advanced societies view arranged marriage as an inherently chauvinistic practice seeing as its goal an exchange of daughters (mostly) for property or alliance formation and quite separated from anything resembling love. The history of the practice in the West would support that view. But janam kundali–based matchmaking in India, which is still practiced today, is somewhat less transactional. Compatibility is the goal. Vedic matchmakers consider dozens of elaborate weighted variables as part of a predictive model. It’s believed that a promised pair should share at least eighteen matching points in their thirty-six-point gun milan, in addition to moon position. All of this formulation is based on careful observation of planetary movements. Today, many Indian couples who were married according to Vedic tradition swear by the practice and have long, stable marriages to bolster their case.
That’s not to say this method for pairing people holds up to any sort of serious scrutiny. The endurance of arranged Vedic marriages has nothing to do with moon position but with partner commonality. Social convention biases the results because throughout most of Indian history it was customary for individuals to marry within their caste. This ensured that couples had similar socioeconomic backgrounds and a wide web of mutual affiliations, two factors that have been proven to be predictors of marriage longevity. And, of course, divorce was not really an option.
Today, we use better instruments and more complex math to map the movements of planets. While the field of physics has yet to reconcile the divergent equations of Sir Isaac Newton with those of Albert Einstein, it has given us a serviceable understanding of how matter and energy interact. There exists neither a plausible theoretical basis—nor any evidence—to support the idea that the position of the moon, the sun, or Neptune on the day of a person’s birth will have any measurable effect on her personality, relationship, or the future.
In some ways, science has triumphed over superstition. But in many other crucial ways, the power of the esoteric has only grown. Humanity is even more open to the idea of matchmaking by math, even math we don’t understand, than we were thousands of years ago. Many of the 40‑plus million Americans who have tried online dating sites that use propriety (i.e., secret) matching algorithms aren’t that different from the poor souls who paid hefty sums to Brahmin wise men for predictive models that couldn’t be proven. To date, no online matching site has demonstrated the success of their algorithm in any way that would allow an independent skeptic to check their work or repeat their results.
The one online dating site that is relatively open about their matching methodology is the very popular OKCupid. The system uses two scores: how you answer questions and the importance you place on a potential mate’s response to the same questions. The more questions you answer, the more information the system has to improve its matches. Users are presented with questions ranging from the humorous (e.g., “Have you ever murdered anyone?”) to the personal (e.g., “How often are you open with your feelings?”) to the basic (e.g., “Is the earth bigger than the sun?”) to the political (e.g., “Is gay marriage a sin?”). The most provocative questions are extremely personal, dealing with sexual preferences, interests, boundaries, and history.
After a user responds to a question she is presented with an importance scale to categorize how she wants a partner to respond. Answers here range from “irrelevant” to “mandatory,” indicating that the user does not want to be matched with people who avoid that question. The point value is logarithmic so the values increase exponentially as opposed to linearly.
How important is this question to you?
A little important 1
Somewhat importan 10
Very important 50
To calculate how well you match with someone else, the system takes your scores, the scores of the person whose profile you’re viewing, multiplies these together, and takes the square root. If you’re a 95 percent match with someone, that means you answered many of that person’s most important questions “correctly” (i.e., in the way she indicated she wanted that question answered) and she answered your most important questions the same way. The logarithmic system ensures you’re not matched with someone who just happened to share lots of trivial things in common with you but was miles away on the important stuff.
Because users submit their own questions, there’s a seemingly endless supply of them, and because the more questions you answer, the higher the likelihood of getting a good match, OKCupid is one of the very few Web start-ups outside health care that offers a real and tangible benefit for giving away more personal information.
When OKCupid purges a user for violating the terms of service as happens from time to time, that user has no means of getting her questionnaire data out of the system or verifying that’s it’s been destroyed. If you leave OKCupid voluntarily, you don’t get your data back, either.
In its early days, OKCupid was critical of such dating sites as Match.com, which charges a subscription fee of $17. Christian Rudder, one of OKCupid’s founders, explained the start‑up’s philosophy in a 2010 blog post: “The practice of paying for dates on sites like Match.com and eHarmony is fundamentally broken.”
The primary malfunction was an imbalance of parties. As Rudder observed, “Men drive interactions in online dating. Our data suggest that men send nearly 4 times as many first messages as women and conduct about twice the match searches. Thus, to examine how the problem of ghost profiles affects the men on pay dating sites is to examine their effect on the whole system.”
OKCupid was eventually purchased by IAC/InterActiveCorp, the same company that owns Match.com, which also bought all of OKCupid’s data. Sam Yagan, another cofounder of OKCupid, soon rose to become head of the entire portfolio of IAC dating sites including Match.com. He had a change of heart about paid dating.
“I think that there was a time where I believed that dating would be a winner-take-all market in the same way that Craigslist, eBay, PayPal, were all winners-take-all marketplaces. I think dating is different,” he told me. Because Match.com has a much higher number of users across more age groups, (93 million monthly site visits) there’s a better chance of users finding a date, which is worth paying for, he says. More users also help both Match.com and OKCupid better understand the science of love.
In his first post on the blog, titled “Rape Fantasies and Hygiene by State,” which showed a state‑by‑state breakdown of people who answered questions about their willingness to act out rape scenarios in bed at a partner’s request, OKCupid cofounder Chris Coyne boasted about the utility of OKCupid as a living social-science lab: “Old media could only get 3,050 people to answer a poll about Obama. And it was enough to call the election with confidence. OKCupid, on the other hand, can ask the world’s most personal questions and get hundreds of thousands of answers.”
OKCupid, as a Web site, is indeed a provocative tool for measuring the attitudes, beliefs, and sexual peccadillos of millions of people. But no matter how many questions users answer about themselves on the site, their matching percentages with other users aren’t any real indication of how likely one or another of them is to enter a long-term relationship.
In a 2012 article from the journal Psychological Science, Northwestern University psychologist Eli J. Finkel and his colleagues show that at best, even very good online dating sites can help you rule out whom not to go on a date with, but can’t tell you if the person you’re on a date with is at all likely to become your lifelong partner. Likewise, it can’t give you any advice on making that relationship stronger.
Here’s why relying on profile compatibility alone doesn’t work. When agents are free to select potential matches from a menu of profiles, certain profiles receive a lot more attention than others; and the more popular a person’s profile, the more messages, chat requests, and invitations she receives, the less likely she is to answer any of them. The result is that the best candidates feel overwhelmed and don’t want to participate in the network, and a lot of lesser candidates send out requests that aren’t answered. Eventually, they lose interest, too.
You might call this the prettiest-girl‑in‑the-room syndrome. It’s really just the social manifestation of survival of the fittest. From an evolutionary perspective, it makes perfect sense that we would constantly seek to associate with people who are a bit out of our league, even when sexual reproduction isn’t an issue, but we don’t want to be so outside our league that we give up any shot of scoring. Online dating sites mask the true odds.
Christian Rudder has acknowledged that the online environment makes the prettiest-girl‑in‑the-room syndrome worse. “You’ve got to make sure certain people don’t get all the attention. In a bar, it’s self-correcting. You see ten guys standing around one woman, maybe you don’t walk over and try to introduce yourself. Online, people have no idea how ‘surrounded’ a person is.”
We experience online dating profiles in the same context we view Amazon items. We shop for what on paper looks best. But with online dating, there’s an added illusion of one‑on‑one bonding. When we happen upon a profile that speaks to us we feel like we’re getting to know someone for the first time, intimately. We can’t see that dozens, perhaps hundreds of people are having the same reaction to the same profile. Dating Web sites would provide more value if they could predict which profiles were going to get the most attention. This is the score that really matters, and we’ve barely begun to understand how to tally it.
How Facebook Will Soon Be Able to Predict Whom You Will Like
In 1946 psychologist Fritz Heider first proposed a methodology, albeit a simple one, to quantify how what you liked affected your relationships, and how your relationships affected what you liked.
His methodology, since dubbed balance theory, holds simply that when the people we like don’t like the same things we like, we grow to either like those people less, tolerate their bizarre affections a bit more, or convince ourselves that the discrepancy is an illusion or irrelevant. Any one of the above options brings the relationship back into balance. And because balanced relationships require less energy to maintain, they are more sustainable. This is how he explained it: “p likes his children, people similar to him; p is uneasy if he has to live with people he does not like; p tends to imitate admired persons; p likes to think that loved persons are similar to him.”
Balance theory, in other words, suggests that it’s easy to be friends with those who are also friends with your friends, and your friends’ enemies make suitable enemies for you as well. You could plot this out on a triangle with a series of pluses and minuses, and depending on how the pluses and minuses matched up, the relationship would be either balanced or imbalanced. In relationships that are out of balance, where you have two friends who dislike each other (and neither can pretend the other does not exist), you’re very likely to drop the relationship that has the highest cost. That means that if you can accurately ascertain which of your friends like one another and how much each relationship “costs,” you can predict which of your friends you will keep and drop.
In 2004 a computer scientist and one of the inventors of RSS, Ramanathan V. Guha, proposed an alternative—but still complimentary— theory that status was a bigger factor than balance in who liked what.
Guha, who today works for Google, was the chief architect of Epinions.com, a product review site claiming to offer “real reviews by real people.” In order to distinguish more trustworthy product reviews from less trustworthy ones, Guha created a “Web of trust” system where Epinions users could rate their fellow reviewers as authoritative or not authoritative. As tends to happen on online communities, some users quickly established more clout than others. Guha observed that those individuals who had the highest status had the most pluses attached to them (they were adored) and sent the most number of minuses out.
Back to our problem. Status theory explains prettiest-girl‑in‑the-room syndrome but balance theory much better represents what a functional relationship is like. So which theory works to predict romantic matches? The answer is both.
In 2010 Stanford’s Jure Leskovec, who has done work with the Facebook Data Science Team, applied social balance theory and status theory to Epinions, Wikipedia, and Slashdot users to see which theory predicted how people would form alliances. Epinions and Slashdot allow users to designate “enemies” as well as friends, and Wikipedia allows users to edit the work of others (which can be a signal of an antagonistic relationship).
Using sixteen feature vectors, he found that he could predict friend and foe relations with up to 90 percent accuracy. Now apply this to Facebook, which doesn’t allow “unfriendships” or “dislikes” but does allow comments on posts. Those comments that are not accompanied by likes can be correlated with dislikes. It wouldn’t be hard to run comments through a semantic machine-learning algorithm to determine key dislike phrases that could more clearly indicate a negative edge (big minus sign). If you’ve got one friend who is constantly sharing material you aren’t fond of—election season tends to bring this stuff out like nothing else can—and you find yourself arguing with her posts, there’s a good chance you don’t regard your friend as having a terribly high status; you perceive little downside to picking a fight with her. If you notice that one of your friend’s friends tends to agree with you, there’s a good chance you will wind up being that second friend’s friend before too long (friends on Facebook, anyway). Facebook, which also serves as a dating site for millions of users, gives a clearer window into status.
If you want a more precise understanding of someone’s rank on Facebook, look beyond their friend count to the number of updates they post and the number of likes they get for them. While you may not have much of an interest in predicting which of your friends’ friends you will connect with on Facebook, Facebook does have an interest here.
Status and balance scores are what’s missing from online dating sites yet the reason for their absence is obvious. No one would use a dating site that made him feel like a loser with a terrible status score.
But Finkel’s research shows that that the long-term survivability of a romantic relationship is predictable on the basis of three variables. Similarity between partners, a category that includes music, religion, educational attainment, income, location, and a host of other things that can (today) be discovered online is just the first one. The other two are how partners collaborate and interact on a day‑by‑day basis and how partners react to stressful events.
You are more than a sexual fetish. For that matter, you’re more than an income bracket, more than your last educational degree obtained, height-weight proportionality, facial-feature symmetry, location, political affiliation, or musical taste. Most pay-dating sites try to quantify your personality and some even try to give a number to how you react to different events. But they do this via survey and that’s the problem, because you’re more than who you are when you sit down and fill out a form describing who you are. When you reduce yourself to a dating site profile, the result may be closer to the ultimate you than the position of the moon when you were born, but perhaps not by that much.
The second variable in predicting relationship longevity that Finkel identifies is collaboration style, a category including communication signals such as how well someone listens, how often or forcefully he interrupts people, and whether he laughs at his own jokes or never at all. Collaboration style includes subtle and nonverbal forms of communication: fidgeting, hand waving, posture, flirtatious glances, and disconcerting stares. These are factors that come into play when people talk about clicking with someone on a first or second date. But collaboration style also comes into play in working relationships and can include such factors as how likely someone is to ask for help when they need it; if she waits until the last possible moment to deliver uncomfortable news; if she seems to whine about every little thing. Whether the answers to these questions are deal breakers for a relationship depends on the unique nature of the couple and the way their communication influences each another. Opposites do sometimes attract because some communication styles have to be complementary, rather than reflective, in order to work. These are the sorts of qualities that simply don’t make it into an online dating form, at least not yet. Measuring how two people communicate and how they collaborate has been historically extremely difficult.
That’s beginning to change.
Excerpted from “The Naked Future: What Happens In a World that Anticipates Your Every Move” by Patrick Tucker. Published by Current, a Penguin Random House Company. Copyright © 2014 by Patrick Tucker. Reprinted with permission of the publisher. All rights reserved.
Like little stars.
World's best pie apple. Essential for Tarte Tatin. Has five prominent ribs.
So pretty. So early. So ephemeral. Tastes like strawberry candy (slightly).
My personal fave. Ultra-crisp. Graham cracker flavor. Should be famous. Isn't.
High flavored with notes of blood orange and allspice. Very rare.
Jefferson's favorite. The best all-purpose American apple.
New Hampshire's native son has a grizzled appearance and a strangely addictive curry flavor. Very, very rare.
Makes the best hard cider in America. Soon to be famous.
Freak seedling found in an Oregon field in the '60s has pink flesh and a fragrant strawberry snap. Makes a killer rose cider.
Ben Franklin's favorite. Queen Victoria's favorite. Only apple native to NYC.
Really does taste like pineapple.