Yagan, however, is still convinced that mobile phone data is going to make online dating more fun and more effective. But finding you a date at this moment will be less important than improving your overall chances of finding a good date in the future.
“Dating is one of the categories of consumer Web products where the end goal is an in-person interaction. It comes into play in a way it wouldn’t for your Twitter mobile app. We can leverage that in a more transformative way than other categories. Your dating app will go on your dates with you,” says Yagan. The data that Yagan is seeking would look a lot like the customer feedback that eBay uses to rate transactions and sellers. In the case of Match .com, the product in question would be a date with someone from the site.
“Right now,” Yagan continues, “because we don’t know when dates take place, the algorithms optimize for on-site communication. It’s hard for us to tell if a conversation you’re having on the site is going to lead to a date or is just superficial chat. And because we don’t know if you went on the date, we can’t ask you for feedback on if the date went well.”
Match.com’s research suggests that people are relying a lot less on bars to meet people, so the service has gone into the business of hosting and sponsoring its own events, from happy hours to wine tastings to cooking—events that, according to Yagan, have attracted hundreds of thousands of people around the world.20 “That’s a service level that OKCupid doesn’t offer but that Match does.” Every day, people go to bars to meet other people. For these folks, a sushi-rolling class doesn’t seem to be a better opportunity to meet your soul mate than a boozy happy hour, but because it offers a better opportunity for actual collaboration around a project, it probably is.
Bottom line, these services still suffer from the prettiest-girl-in-the-room syndrome but they’ve put researchers closer to curing it.
Introductions are important, but facilitating meet-cutes between people is only a first step in solving the deeper mystery of collaboration dynamics. Partnership, romantic and otherwise, is built on patterns of communication: What happened after who said what? How did A respond to B? How quickly? What was the tone? What does that mean? Our brain evolved expressly to help us navigate this highly chaotic give-and-take. In fact, research suggests that we evolved the ability to detect certain types of lies almost at the same time that we learned to vocally express complex ideas.21 A truly quantitative approach to love requires a system that remembers every useful signal that’s traded back and forth.
Sandy Pentland’s Honest Signals
Alex “Sandy” Pentland, the head of the Human Dynamics Lab at MIT, has spent thousands of hours recording human interactions. Doing this in a lab setting would have been cost exorbitant and yielded results that were biased by the artificiality of the lab environment. So in 2002 Pentland made the world into his lab. He put together an infrared (IR) transceiver, an 8-kHz microphone, two accelerometers (the sensors in an iPhone that measures movement), a 256-MB flash drive, and four AAA batteries on a sensor board, encased in plastic. The first sociometer, as he calls these devices, was a bulky shoulder mount that looked like something a futuristic soldier would wear in a 1980s science-fiction movie. But it worked to record key aspects of human interaction, including pitch, volume, acceleration of speech, and use of hand gestures.
Pentland wasn’t going to all this trouble to record the content of the conversations. There was no shortage of material that had been recorded for its content. Instead, he was out to record the signals we deliver and process unconsciously. He calls these signals “honest.” You could also call them predictive.
In his 2008 book, Honest Signals: How They Shape Our World, he lays out four signaling types that make up the basis of human interaction: influence, “measured by the extent to which one person causes the other person’s pattern of speaking to match their own pattern”; mimicry, which he calls reflexive copying of nods, smiles, other gestures; and activity, which includes a lot of hand usage, speaking quickly then slowly, softly then loudly, high and then low.22
These signals, working together like elements of a costume, serve to establish character in conversation in a way that’s hard to consciously notice but is nonetheless deterministic. A lot of activity indicates sudden excitement, which further suggests that the actor is “open to being influenced.” You’re not all that confident in what you’re saying but you’re certainly interested in what the other person may say. The last signaling type, consistency, refers to evenness in tone. It’s marked by focused attention and low movement. If you have a lot of it, this communicates a message of confidence and resolution.
How you and your conversational partner send or receive these signals indicates how interested you are in what the other person is saying, how open you are to her ideas or concerns, how certain you are about what’s coming out of your mouth, and your level of interest in taking the conversation to a new place. In short, these signals tell your conversational partner what role you intend to play in the interaction.
Are you displaying focused attention, mimicry, and agreement? You’re signaling that you will take the “teaming” role. Are you speaking first, modulating your voice little, and displaying muted activity? You’re taking a “leading” position. Are you bouncing around in conversation, talking with your hands, fidgeting, and throwing in personal asides? This is the “exploring” role and accepting it means you’re open to what the other person has to say. The final role, “active listening,” works about as you would imagine that it would. None of these character parts is better or worse than any other. They all serve a precise function and we slip into and out of them at different times; it depends on whom we’re talking to and the subject at hand.
With just thirty seconds of sociometer data, Pentland can predict what roles two people will take in a given interaction. Your brain takes about the same amount of time to unconsciously make the same prediction when you chat with someone. But unlike your brain, once Pentland has predicted the role, his model can anticipate what’s going to happen, specifically who is going to get her way. And the model can do so with as high as 95 percent accuracy. Bosses who took the leadership role during salary negotiations with underlings were far more likely to walk away from the table with the better end of the deal. Pentland writes that this was a “better predictor of negotiated salary than any other factor.”23 Poker players who engage in active listening by showing their openness to being influenced and suppressing excitement are probably holding a much better hand than they’re letting on. This is the amateur’s poker face. It’s largely the same around the world and arises from the attempt to suppress the fight-or-flight instinct that people experience when engaging in a high-stakes behavior, an adaptation that was paramount to our survival in the wild. When amateur players have a bad hand they act normally but when they have a particularly good hand they go wooden. “People are not very good at judging how much nervous energy they’re showing, so they say to themselves, ‘To avoid looking excited, I’ll look like a dead fish,’” Pentland explained to me.
When a poker player tries to suppress the functioning of the entire automatic nervous system, she’s using a very small portion of her brain to override the functioning of a very big part of the brain. A lot of energy gets used up. To compensate, the brain shuts down other actions, like fidgeting. The result is a rather unnatural stiffness that’s largely absent when there are no strong stimuli to inspire the fight-or-flight response, as when you’re playing your cards straight. Pentland observed that silence and stiffness were correlated with bluffing 68 and 67 percent of the time, respectively.24 He became so good at recognizing this signal that today he can effectively outperform virtually any semipro poker player. But he has no interest in that. He hates Las Vegas. He himself displays very few of the signals of a hard-nosed negotiator. Sandy Pentland is a terrible bluffer.
Pentland’s office is on the fourth floor of the MIT Media Lab. When you meet him in person
, after reading his book, you notice that his voice rises and falls easily. He can grow quite animated when he’s discussing some aspect of his work about which he’s particularly passionate. For someone who knows more about the value of adopting particular affectations than anyone else on the planet, he seems incapable of faking who he is. That’s the entire point, according to Pentland. Our unconscious signals are our honest signals and a better indication of what we are really thinking, what we’re really like, than anything we say. If Pentland, the man who discovered these signals, can’t easily hide them, it seems there is little hope for the rest of us.
“What’s cool is that now we can measure this stuff. And it’s the same pattern we see in bees, in small human groups, and in large networks. It’s a different way of thinking about things. Everybody else wants to look at the content of ideas. I just want to look at the flow of information. If the flow is good, good ideas will surface.”
Once you learn to recognize honest signals, it’s impossible not to see their impact everywhere. A great example is the 2012 U.S. presidential race. In the first debate Mitt Romney stuck with an even tone. His voice did not ascend or descend in either volume or pitch very often and he spoke with little hesitation even when he was going on the attack. Barack Obama, conversely, began many statements with long pauses underscored by the rasping sigh that has become fodder for mimics, thus displaying greater variance in pitch and voice than his opponent. Later independent analysis shows that the president made far fewer inaccurate statements than Mitt Romney did in that debate, but that’s not how the public decided the winner. In surveys conducted immediately after the contest, people described Barack Obama as “disengaged.” In the end this made it appear to the television audience as though Mitt Romney was in expert command of his material, especially when he was attacking the president. The president meanwhile looked insecure and unstudied even when defending his own record, subject matter about which he should have been deeply familiar.
Pentland admits a certain cynicism about the political process. He doesn’t believe what candidates say in these sorts of exchanges but that doesn’t mean the debates aren’t good viewing if you know what to look for. “If you watch them this way, watch the meta-signals, you can tell what they get excited about. There’s a little extra in there or a lack of something,” he told me.
Pentland’s work also shows that honest signaling is more important to romantic relationships than profile matching.
To demonstrate this he and one of his research partners, Anmol Madan, attended a local Boston speed-dating event and asked the participants, all between ages twenty-one and forty-five, to wear the sociometers (a somewhat later version that was smaller and less intrusive). They observed sixty five-minute sessions. The participants themselves were instructed not to exchange any numbers directly. Rather, at the end of the evening, every participant was told to pick which sessions they thought went well and then numbers could be released. Pentland and Madan found that by paying attention to social signals between the couples they could predict, with 73 percent accuracy, which women were going to release their numbers. (The baseline, flat-out guessing, will get you to 20 percent accuracy.) Mimicry was the most obvious giveaway. A woman who in some way modeled her conversational partner’s speech or, in particular, his movements was the most likely to release her phone number. A slightly less important tell was variation in emphasis, speaking fast and slow, high and low, and revealing an openness to influence. Together these two modes of signaling were a “display of the exploring role.”
Here’s what was most significant about Pentland’s study: the response of the male speed-daters could also be predicted on the basis of female signaled interest. While the male signaling had relatively little effect on the women, female signaling had a big effect on the men.
Pentland had expected that every male would release his number to every female he had a session with because everyone knows men make themselves more sexually available to everyone in order to maximize the chances for sex. Instead, what Pentland found was that the men were more likely to give their number to a female who had also elected to give her number to them.
Here is where the prettiest-girl-in-the-room syndrome met its match. The guys came in with their Lovegetys set to 10 but left with an understanding of how each interaction with each conversational partner unfolded, and that changed each guy’s own level of interest. Every male liked the women who liked him back and lost interest in the women who seemed to be shooting him down.25
Anthropologist and bestselling author Helen Fisher, Match .com’s chief scientist, is also interested in capturing unconscious signals that explain personality in a way that’s a bit beyond what the average online dater even understands about herself or puts on a form. Fisher is leading a research effort to pin “dimensions of temperament” to individuals based on biology. In essence she’s proposing a unified theory for personality. And because dating sites sell matches based on personality, it would be a big competitive edge for Match.com if her theory proves valid. She’s looking to show that your dating personality is a function of your individual brain anatomy. She says that individual brain chemistry “could be useful in defining some of the primary biological structures of personality.”
Her central argument is this: our personalities are governed by how different neurotransmitters interact. Dopamine-led people are, in Fisher’s words, “curious and energetic.” This category is for eccentrics, people who love to travel, can’t stop asking questions, display a certain degree of impetuousness, and are eager to explore the world. It’s the category in which she places herself. If you’re more cautious and inclined to follow the rules, you’re serotonin led. An analytical and tough-minded arguer? Your temperament dimension is testosterone based. Finally there’s the estrogen-based personality characterized by empathy and attentiveness. The estrogen profile is the quintessential nice guy or nice girl.
Fisher is not the first researcher to propose a matrix to pigeonhole people into different personality types. Ernest Tupes and Raymond Christal’s five-factor model, which scores the personality on the basis of openness, conscientiousness, extroversion, agreeableness, and neuroticism is perhaps the most widely known. The fundamental flaw with these personality matrices is they aren’t based on biology, says Fisher. They start with an assumption rather than a hypothesis that could be falsified with real evidence from fMRI data. “Make the questionnaire based on what you know from biology, and then go back to the biology to study whether, in fact, what you say you’re studying, you actually are studying,” she says. As she explains it, “The bottom line is, I am not studying your culture. I’m studying your biology. I’m only studying traits that I know have a biological basis.”
To demonstrate the validity of her theory, Fisher and her colleagues conducted fMRI scans on a wide assortment of people and have also designed a questionnaire (she refers to this as the Fisher Temperament Inventory), which seeks to measure “the degree to which one expresses aspects of these four trait constellations.”
She’s also played a key role in a major survey effort at Match .com called Singles in America, which is America’s largest survey thus far on dating habits. It’s a first-of-its-kind big data study on dating, featuring responses from more than 107 million people who took the survey on Chemistry.com (also part of IAC’s portfolio of dating sites) and Match.com. Among the findings from the survey released in the most recent study:
A friends-with-benefits arrangement is becoming more mainstream every year. Some 44 percent of surveyed adults admitted to having been in a friends-with-benefits relationship in 2012, up from 20 percent in 2011.
Your friend with benefits should be monogamous. More women than men say they would prefer to be in an exclusive relationship before going to bed with a partner (37 percent of single women in 2012, 31 percent in 2011, and 25 percent in 2010).
Perhaps the best finding for Match.com’s bottom line: t
he surveys show that more people are looking for love on the Internet first and other places second. Far more people meet online today than they do in bars. Nearly 30 percent of the singles surveyed had dated someone that they met online and 20 percent had met their most recent partner through a Web site. Only 7 percent said they had dated someone they met at a bar.26
“What big data can do for you is find patterns in behavior and personality that you would never find in a small sample—never find it because there’s too much noise,” says Fisher. “The vast majority of personality questionnaires, and all kinds of questionnaires, are based on the college population, for heaven’s sake.”
Like Yagan, Fisher is also interested in using data from mobile apps to take profile matching to an entirely new level. But she’s looking for something rather more ambitious than a simple hookup app. What she wants is “the kind of app that uses what we know from evolutionary psychology, body language, linguistic studies, that really sums up a deeper understanding of who you’re talking to . . . I mean, we spend our lives trying to size up the people around us constantly. And I think we’re missing an awful lot of data, unless we start using apps that really do talk about word usage, body language.”
This is what makes Pentland’s work with the sociometer so potentially valuable to the future of dating and marriage. Because the sociometer provides a means to continuously see how your personality is affecting the person closest to you, it can provide a telemetric monitor for relationship health. After the experiment concluded, Anmol Madan actually created a program called a Jerk-O-Meter, which ran on a Zaurus VOIP phone.27
The program worked just as you might imagine. When the owner’s attention or interest in the phone conversation began to flag, the program would send out such helpful notifications as “stop being a jerk.” When Madan revealed the app, it received coverage from CNN, the New York Times, and other major media outlets, but it never emerged as a viable commercial product. Like Leonardo da Vinci’s helicopter or Friendster, the program paid the price for being ahead of its time. The Zaurus was not a popular device commercially and the iPhone was still in its infancy at the time of development. Moreover, the idea of self-monitoring personal conversations for “degree of jerk” probably struck consumers as bizarre. People have been having conversations since, well, before there were people. Why should we now use a device to help us do the most natural thing in the world?
The Naked Future Page 20