Invisible Influence Read online

  But what does it mean that even experts get it wrong?

  * * *

  This question vexed Princeton sociologist Matthew Salganik as he was working on his dissertation. Hit books, songs, and movies are so much more successful than their peers that we tend to see them as qualitatively different.

  But if the best are distinctly better than the rest, why do experts have so much trouble identifying them? Why did so many publishers pass up the opportunity to sign J. K. Rowling?

  To find out, Salganik and his colleagues set up a simple experiment. They designed a website where people could listen to music and download it for free. No famous songs or well-known bands, just unknown songs from unknown artists. Local acts that were just starting out, or groups that had just put together their first demo. Bands with names like Go Mordecai, Shipwreck Union, and 52 Metro.

  The songs were organized in a list, one after the other. People could click on any song, listen to it, and, if they liked it, download it. Song order was shuffled for each listener to ensure that each song received equal attention. More than fourteen thousand people participated.

  In addition to the names of the bands and the songs, some people were given information about what previous listeners liked. For each song, they could see how many other people had downloaded it. If 150 people had downloaded “Lockdown” by 52 Metro, for example, the number 150 appeared next to the song.

  And just like a best-seller list, for these “social influence” participants the songs were ordered by popularity. The most downloaded song appeared at the top of the list, the next most downloaded second, and so on. The download numbers and song order automatically updated each time a listener downloaded a new song. Then Salganik examined which songs people downloaded.

  Simply providing information about others’ choices had a big impact. Suddenly people tended to follow their peers. Just like watching a point of light in a dark room, people listened to and downloaded songs that prior listeners had liked.

  Popularity became concentrated. The gap between the most and less popular songs increased. Popular songs became more popular and less popular songs got even less attention. The songs stayed the same, but social influence led the best to do better and the worst to do worse.

  But Salganik wasn’t finished. It was neat to see how people’s tendency to imitate others influenced popularity, but that still didn’t resolve the prior puzzle. Sure, certain songs or books might be more popular than others, but why couldn’t experts armed with market research predict those successes in advance?

  To find out, Salganik added one more detail.

  It’s impossible to rerun the real world. No one can stop time, go back, and see what would happen if things started anew. So instead of rerunning the same world, Salganik created eight different ones. Eight separate worlds, or distinct groups, that looked identical—at least initially.

  This decision was key.

  The beauty of a good experiment is control. In this case, each of the eight worlds started the same. Everyone had access to the same information. All songs started with zero downloads, and because people were randomly assigned to each world, even the participants in the different words were indistinguishable. So while some people might like punk music, and others might like rap, on average there were an equivalent number of people with each preference in each world. On every dimension possible, then, the worlds started the same.

  But while they started the same, each world evolved independently. It was almost as if eight different versions of earth were separately spinning next to one another.

  If success were driven by quality alone, each world should end up looking the same. Better songs should be more popular, worse songs should be less popular, and the songs that are popular in one world should be popular in all of them. If 52 Metro’s “Lockdown” was the most downloaded song in one world, it should be close to the top of the list in others. On average, preferences across the groups should be the same.

  But they weren’t.

  Song popularity varied widely from one world to the next. 52 Metro’s “Lockdown” was the most popular song in one world. In another, one of the least popular. Fortieth out of forty-eight. Almost dead last.

  Same song, indistinguishable groups of participants, completely different levels of success. Same initial conditions, different final outcomes.

  Why was success so variable?

  The reason was social influence. There weren’t more punk lovers in the world where 52 Metro was popular than in the world where it wasn’t. But because people tend to follow those who came before them, small, random initial differences snowballed.

  To understand why this phenomenon occurs, imagine parking at a county fair. There’s no real parking lot per se, or even someone directing traffic, just a big field where people leave their cars. People are generally indifferent about where they park, they just want to go eat cotton candy and ride the Ferris wheel. There are no white lines denoting where individual cars should go, so the first family that drives in can park wherever they want.

  The first car that drives up happens to be driven by the West family. They slightly prefer facing west when they park, so they drive in, turn right, and park their car facing west:

  Then the second family shows up. This family, the Souths, prefer parking facing south rather than west. But their preference is not that strong, and given that the first car is parked facing west, they pull up next to them and face west as well:

  Soon, more and more cars show up. While the people in each might have a slight preference here and there, they follow the cars ahead of them until the parking lot ends up looking like this:

  Makes perfect sense.

  But what if, rather than the West family showing up first, the South family had shown up first instead? What if the Souths had been the first to park in the lot?

  Instead of parking facing west, given the Souths’ slight preference for facing south, they go ahead and park that way:

  The Wests show up next. They would have slightly preferred to face west, but given a car is already facing south, they go ahead and do the same. More and more cars show up, all following the cars in front of them, until the lot ends up looking like this:

  Same eight cars, same overall parking preferences of people in the lot, but completely different outcome. Everyone is facing south rather than west. Just because of the preferences of whoever happened to park first.

  This same process drove the outcome of the music study. Imagine two of the social influence worlds at the start of the experiment. They are essentially identical. None of the songs have any downloads. Even the participants are the same, on average.

  Just like the Wests and the Souths, though, different individuals may have slightly different preferences. One person may have a slight preference for punk over rap, while another has a slight preference for rap over punk.

  And the order in which these two people express their preferences varies. In one world, the person who likes punk happens to go first. They listen to a few songs, find a punk song they like, and download it. Score one for the punk song, zero for the rap song. Then, when the second listener comes along, they use the first listeners’ choice as a guide. The punk song has more downloads, so it gets more attention. The second listener has a slight preference for rap, but they like punk and the song seems pretty good, so they download it. Punk 2, Rap 0.

  In the other world, the person who prefers rap happens to go first. The process is much the same, but with a different outcome. They listen to a few songs, find a rap song they like, and download it. Not because they hate punk songs, but, on the margin, they prefer rap. Punk 0, Rap 1. Then the punk liker comes along, but this time they are second in line. So rather than going with their slight preference, they are influenced by others, and download a rap song as well. Punk 0, Rap 2.

  Soon, those two once-identical worlds start to look a little different. One world has a punk rock song on top of the list and the other has a rap song.
/>   Again, one person liking a song isn’t enough to entirely change someone else’s preferences. But it’s enough to tip the scales. Songs at the top of the list got more attention, were more likely to be listened to, and as a result, more likely to be downloaded. Which made it more likely that the punk rock song would be downloaded again in the first world, and that the rap song would be downloaded in the second one. And the process repeated with the next listener.

  Slowly, but surely, just like the cars parked in the field, social influence pushed the once-identical worlds in different directions. Magnified over thousands of people making choices, it led to vastly different outcomes.

  The implications are both simple and shocking. Rather than being driven by quality, hits might sometimes just be driven by luck and the herd. If we reran the world again, Britney Spears (and J. K. Rowling, for that matter) might never have been popular. Britney’s video happened to land at the right time, some people liked it, and, because of that, others jumped on the bandwagon. But she might not be any better than thousands of aspiring musicians we’ve never heard of.

  * * *

  Does this mean that anything could be a hit? That terrible books and movies are just as likely to be popular as good ones?

  Not exactly. Even in Salganik’s experiment, quality was still correlated with success. “Better” songs, those downloaded more in the independent world, tended to get more downloads and “worse” songs tended to get fewer. The best songs never did terribly, and the worst songs never did extremely well.

  But there was still a lot of variation. And what that means is that quality alone is not always enough.

  There are thousands of books, movies, and songs vying for collective attention. And no one has the time to read every book jacket or listen to every sample clip. Most people don’t have the bandwidth to check out even a small percentage of the options.

  So we use others as a helpful shortcut. A filter. If a book is on the best-seller list, we’re more likely to skim the description. If a song is already popular, we’re more likely to give it a listen. Following others saves us time and effort and (hopefully) leads us to something we’re more likely to enjoy.

  Does that mean we’ll like all those books or songs ourselves? Not necessarily. But we’re more likely to check them out and give them a try. And given the thousands of competing options out there, this increased attention is enough to give those items a boost.

  Knowing others liked something also encourages people to give it the benefit of the doubt. Appearing on the best-seller list provides an air of credibility.12 If that many people bought it, it must be good.

  * * *

  J. K. Rowling unintentionally tested these ideas when she released a book under a pseudonym. After her success with Harry Potter, Rowling decided to write a detective novel called The Cuckoo’s Calling. While Potter brought Rowling notoriety, reviewers were critical of later books in the series, and Rowling worried her fame would bias response to the new novel. She wanted to let the writing speak for itself. So she released The Cuckoo’s Calling as Robert Galbraith. A combination of Robert F. Kennedy and her childhood fantasy name Ella Galbraith.

  Robert Galbraith’s novel had mixed success. Almost every person who read The Cuckoo’s Calling liked it. They called it “inspired” and “an engaging read.”

  Unfortunately, however, there just weren’t that many of them. Readers were few and far between. The Cuckoo’s Calling was released with little fanfare and sold only 1,500 hardcover copies in the first three months of being on sale.

  Then one day the book surged from 4,709th to become the best-selling book on Amazon. Soon, hundreds of thousands of copies had been sold.

  Had people realized Robert Galbraith’s genius? No. Had careful inspection of the writing of The Cuckoo’s Calling revealed it to be a literary masterpiece? No again.

  Someone had simply unmasked Rowling as the book’s actual author.

  Without J. K. Rowling, The Cuckoo’s Calling was merely one of the thousands of well-written detective novels competing for attention. With J. K. Rowling, it had a 450-million-copy seal of approval that made potential readers take a look. After all, how could millions of people be wrong?

  PUTTING SOCIAL INFLUENCE TO WORK

  These findings about the science of imitation have a number of important implications.

  When trying to persuade others or convince them to do something, we tend to default to rewards or punishments. The employee of the month gets $100 and their name up on the wall. Kids are told to eat their vegetables or they won’t get ice cream for desert.

  But while rewards and punishments are effective in the short term, they often undermine what they set out to achieve.

  Imagine you were stuck on an alien planet and they serve two things for dinner: Zagwarts and Gallblats. You’ve never heard of either, and both look a little weird, but you’re famished, so you have to eat something.

  Before you get a chance to pick one, your host says that before you eat your Zagwarts, you have to eat your Gallblats.

  Which one do you think tastes better? Zagwarts or Gallblats?

  Kids make similar inferences about ice cream and vegetables. They like ice cream, and while they might not love vegetables, the ice cream reward undermines any otherwise positive feelings they might have had. After all, if vegetables were good in the first place, why would they need a reward to eat them?

  An ice cream reward sends a subtle signal that vegetables aren’t worth eating on their own. That kids need to be paid (in ice cream) to eat them. And when parents stop paying, kids will stop eating. Whenever they get the opportunity to make their own food choices, vegetables will be tossed to the side. The same goes for employees. They start to infer that the only reason to be on time and give good service is because they’ll get paid more, not because they care about their job.

  Using social influence is more effective. Just like monkeys with the red and blue corn, people mimic others’ choices and actions. If their parents can’t seem to get enough broccoli, kids will follow suit.

  Unfortunately, many parents signal to their kids that vegetables are not tasty. Parents don’t put many vegetables on their own plate, and eat the chicken or steak or whatever else is being served first. And if their parents aren’t eating vegetables, why would kids want to?

  But if broccoli is the first thing on their parents’ plate, and the first thing their parents eat, kids will do the same. Even better if there’s a mock argument over which parent gets to eat the last piece. The more kids see their parents eating something—and liking it—the more likely they’ll be to do the same.

  Mimicry is also a helpful tool.

  Imagine you’re out to lunch one sunny spring day with a couple of colleagues from work. You’re sitting outside at a local pub, and after scanning the menu for a few minutes you know exactly what you’re going to get.

  The waiter comes by, asks you what you’d like, and the order rolls off your tongue: “The Brussels Burger, medium, with bacon and cheddar, and a side salad.”

  “Okay,” he says, “the Brussels Burger, medium, with bacon and cheddar, and a side salad, correct?”

  “Yes,” you reply, excitedly. You can already hear your stomach rumbling.

  Notice what happened? Probably not.

  Yet the same thing happens to each of us dozens, if not hundreds, of times each day. The waiter didn’t just take your order, he mimicked you. He could have just said “ok” or “coming right up!” But he didn’t. He repeated your order back to you, word-for-word, copying saying the exact thing you said.

  Seem trivial? Maybe.

  But research shows that this mimicry just increased the waiter’s tip by 70 percent.

  Whether trying to win a contract, get someone to do something, or just have people like us, subtly mimicking their language and mannerisms is an easy place to start. Even something as simple as mimicking their greeting style (e.g., “Hey,” “Hi,” or “Hello”) in e-mails increases affili
ation.

  * * *

  By understanding why people imitate, we can also learn to be less susceptible to influence ourselves.

  Group decisions often suffer from something called groupthink, where conformity and the desire for intragroup harmony lead groups to make worse decisions. Watch a focus group share opinions or a committee decide who to hire, and whoever goes first has a big impact on the outcome. Just as how songs became popular due to the preferences of the first few listeners, the direction of the discussion or vote depends on the opinion of whoever happens to lead off. Group members who were on the fence tend to conform, and unless someone has strong objections, they tend to keep their opposition to themselves. Without much of a peep, the group quietly goes one way when they could have just as easily gone the opposite. Groupthink has been blamed for everything from the space shuttle Challenger disaster to the Cuban missile crisis.

  People talk about the wisdom of crowds, but crowds are only wise when the group has access to everyone’s individual information. Aggregating these pieces can lead to better decisions than any person could have made alone. But if everyone just follows everyone else, or keeps their information to themselves, the value of the group is lost.

  Consequently, eliciting everyone’s idiosyncratic information is vital. So how do we do that? How do we encourage opposing voices to speak up?

  Turns out that even one dissenting voice can be enough. If just one prior person in Asch’s line experiment gave the correct answer, it was enough to free participants up to give the correct answer themselves. It didn’t need to be half the room, just one coconspirator. We don’t need to be in the majority to feel comfortable expressing our opinions, we just need to feel like we aren’t alone.

  Interestingly, the other minority voice doesn’t even have to have the same opinion. Even a dissenter giving the other incorrect answer (line A rather than line B) was enough to free people up to give the correct answer (line C) themselves. Just having another dissenting voice, even if it didn’t agree, made people feel more comfortable in expressing their own personal opinion.