Duncan J. Watts poses this question in his book Six Degrees: Suppose you’re in a foreign city and there are two similar-looking restaurants side by side, one bustling with patrons, the other practically deserted. Price being equal, which one do you choose to eat at? His answer is that unless you hate crowds, are in a great hurry or feel unreasonably sorry for the forlorn waiter at the uncrowded eatery, you’ll choose the popular restaurant. This is because you’ll assume that all those customers know something in the restaurant’s favor or they wouldn’t be there. This seems true to me, and even if I’m a committed contrarian, I’ll likely only act out my contrariness in my home town, among restaurants I know. But notice a paradox: Once I go into the popular restaurant, I’m part of the crowd—part of its popularity—and so the next guy who passes by, and makes the same choice I made, is doing so partly on the basis of my presence in the restaurant, on the false assumption that I know something.
Whereas actually the two of us, and who knows how many of the diners already inside, are drawn by the popularity, not by the food.
Watts is a sociologist, so he actually tests his hypotheses. He conducted an experiment showing that, in effect, even if we had chances to sample the food at the different restaurants, we’d still be likely to choose the popular one—except the experiment didn’t involve the popularity of restaurants but of songs, in the form of downloadable MP3 music files. He and his colleagues Matthew J. Salganik and Peter Sheridan Dodds set up a website that offered free downloads of music by little-known bands, inviting almost 1,500 kids from a teen-interest website to partake. Everyone got to listen to the music, was asked to rate each track on a scale of 1 to 5 and was given the opportunity to download it, if he or she wanted. But the participants were not all given the same information. Members of one group of participants—called the independent group—simply saw the name of the song and the band name, but had no idea how many others had downloaded a track. In contrast, members of eight other groups—called social-influence groups—got to see how often a track had been downloaded by previous participants. However, they were only seeing the download information for their own group. So in effect the experimenters had set up nine separate universes, with nine different download histories
Taking the taste of the independent group as a measure of a track’s inherent appeal, Watts and crew found that when totaled over all the groups, the appealing songs did somewhat better and the unappealing songs somewhat worse; but within each of the eight social-influence groups results were all over the place, and were more extreme. That is, in the social-influence universes the popular songs were far more popular than in the independent group, and the unpopular songs were far less popular, but you couldn’t really predict which would be popular and which wouldn’t. A song that the independent group rated as appealing might be popular in some of the universes and middling in the others, and a middling song might be really popular in some universes and unpopular in others, etc.
Before going, “Oh my God, this shows that people choose popularity over quality!”—which is only partially true, since the songs with appeal did do somewhat better overall—look at what’s actually interesting in the result. It’s not that people choose popularity, which you’d expect from the restaurant example anyway, and Watts surmises that not only do we assume the crowd knows something, but we’re also interested in hearing what our fellows are listening to.
What’s truly interesting is that the winners and losers differ from universe to universe. This means that what wins and loses can’t be predicted with confidence, even if you know in advance what people like. Although you’re still better off putting your money on something you believe appeals to music buyers, since it’s got a better shot, there’s no guarantee here. And this principle is ironclad, that there will always be an element of chance, no matter how many focus groups you conduct and how penetrating your understanding of consumer taste is.
The reason for this unpredictability is simple: To paraphrase Watts, small differences in a song’s appeal or even purely random fluctuations can get locked in early and lead to very large inequalities over time. In situations like the download experiment, a lot depends on which people show up first, their individual tastes and their changing moods and momentary preferences. Watts believes this is somewhat comparable to the famous “butterfly effect” in chaos theory. I’d never actually heard of the butterfly effect (so much for fame), but according to Wikipedia it’s a metaphor for the fact that a small, possibly random event can cascade into a large result: for instance, a butterfly flapping its wings in the right air current at the right time can create enough change to cause or prevent a tornado. This isn’t such a good metaphor, actually, since we’re going to get tornadoes anyway, so the butterfly doesn’t change the overall climate; whereas if by chance a James Brown or an Elvis hadn’t gotten popular, modern music would be incomprehensibly different.
In any event, as Watts points out, their experiment has a result that’s counterintuitive: The more people know about each other’s choices, the more likely they are to come to agreement. In retrospect, this strong agreement can make an outcome seem as if it had been inevitable: Look, all these people agree, so this must reflect the taste of the public, or the quality of the merchandise. But in fact the experiment shows the exact opposite. The more people know about each other’s choices, the less predictable the outcome.
To emphasize this: As social influence increases, 1.) agreement increases as well, but 2.) the connection between the magnitude of the popularity of a song (or an opinion or an idea or a belief) and the magnitude of its underlying appeal decreases (that is, its underlying appeal had there been no social influence). This doesn’t necessarily mean that the popular song, opinion, idea, etc. has less inherent appeal than something that’s less popular, just that its popularity is all out of proportion to that appeal, and chance may play a big role. By the way, this principle applies just as much to Beethoven and The Beatles as it does to Fergie and Timbaland—which isn’t meant to imply that they’re all equal, just that the underlying appeal isn’t what’s locking them into their current popularity.
“Inherent” or “underlying appeal” is actually an imaginary notion, given that in the real world no song, opinion, idea or belief exists outside of social influence. In Watts’ experiment the members of the “independent” group are by no means outside of all social influence. Sure, they don’t know the popularity of the songs they’re rating, but they know the popularity of the various musical styles they’re hearing, and their basic taste was formed in the context of knowing other people’s likes and dislikes.
To extrapolate wildly on Watts’ counterintuitive result: The more consensus there is as to an idea, the less likely an individual’s adherence to that idea will be based on his or her examination of the idea, or his or her deep involvement with it. This extrapolation isn’t altogether valid (for instance, consensus can possibly be due to individuals independently coming to the same conclusion, e.g., that night is dark), but in a lot of cases it makes sense. If I believe something that everybody else believes, I’m less likely to test the belief or even give it a lot of attention. To be honest, I’ve never looked at the data that prove that the Earth and Mars and the other planets revolve around the sun.
A final question, which I’ll leave you with (and of course, people’s ideas do change, and lots of ideas and bands fall out of popularity, a subject I’ll put off for the time being): Watts’ idea seems applicable to everything that’s popular, and to all received wisdom. That is, it can’t explain why Shakespeare or Danielle Steel are popular, or why anything in particular is popular; it can only explain why there always will be great discrepancies in popularity, and why popularity can never be fully predicted. So what use is the idea?
I’ve got a whole bunch of answers (for instance, it further emboldens me to grit my teeth when people say, “I believe everything happens for a reason”). Here’s just one: Back in 1973 I was one of the few people I knew who had anything good to say about the New York Dolls and The Stooges. Anyway, now those bands are considered classic, at least by rock-critic types and by those literary intellectuals who pay attention to rock criticism. So you could say that history has vindicated me. Except I’d say that this “vindication” is because people like me—rock-critic types and literary intellectuals—are the ones who talk the loudest about popular culture and the ones who write the history. And since these days I’m not much liking the music of the people who venerate the Dolls and The Stooges, just what does this vindication tell us, and what is it worth?
So one benefit I gain from reading Watts is that I don’t gain easy comfort from phrases such as “the verdict of history.”
Keep the conversation going at firstname.lastname@example.org. Read previous Rules of the Game columns at lasvegasweekly.com.