Consider an experiment where 100 people, each in their own cubicle, are asked to rank three different types of chocolate bars based on their taste. Imagine that 54 people prefer chocolate bar A, 38 people prefer chocolate bar B, and only 8 people prefer chocolate bar C. Given these results most people would be happy to assert that chocolate bar A is the best tasting chocolate bar– I know I would.
Now consider a similar, but slightly different situation. One hundred people are asked to rank three different types of tea. This time however the participants vote one at a time and after each person votes a leader board is updated indicating the current rankings of the tea. If after the first 20 people, for example, tea A had 8 votes, tea B had 6 votes, and tea C also had 6 votes then, person number 21 will have access to this information. Imagine that in a similar way to our first experiment, the results show that 54 people prefer tea A, 38 people prefer tea B, and 8 people prefer tea C. Can we still conclude that tea A is the best tea? This time it is unclear.
The difference between the two experiments is a matter of independence. In the first experiment each of the participants made their decisions independently in their own cubicles. But in the second experiment, each participant (except the first one) also has the knowledge of the votes of previous people. This difference is important because the lack of independence introduces a confound. It is not clear whether tea A is popular because it was objectively the best tea, or whether it became popular as a result of people copying the votes of people who came before them.
In this article my goal is to examine our notion of popularity. We often think that the most popular things are popular because they are the best. It might be, however, that effects of social-influence (like in our tea experiments) may actually be playing an important role.
Let’s make this discussion more concrete. Consider a study published in Science in which a group of experimenters invited 14,000 particpants to register at a Web Site called Music Lab where they were asked to listen, rate and, if they chose, download songs by bands they had never heard of. Some of the participants were only shown the names of the songs and the bands, while other participants were also shown how many times each song had been downloaded by other participants. This second group of people was labeled the “social influence” group and was split further into eight parallel “worlds” so that participants could only see the downloads of other people in their world. At the start of the experiment all of the rankings of the songs were set at zero. Then as the experiment began each world developed it’s own set of rankings independent of the others.
This setup allowed the experiments to test the notion of popularity in the following way: if people know what they like regardless of what they think other people like, then the most popular songs in the independent condition should be the same songs that are popular in the social influence condition.
The findings were the exact opposite. The particular songs that became hits were different in each of the worlds. In other words popularity was not just the result of “intrinsic” quality as measured in the independent condition but was also affected by social influence in each of the worlds. As one of the experiments put it, “introducing social influence into human decision making made the hits more unpredictable.”
But what was the scale of this effect. Although the experimenters found that “good” songs were rated better on average than “bad” songs, it was still found that the impact of a listener’s own reactions is easily overwhelmed by his or her reactions to others. One song, “Lockdown”, by 52metro, for example, was ranked 26th out of 48 in the independent condition, but was the No.1 song in one of the social-influence worlds and 40th in another.
This finding has surpisingly deep consequences for our understanding of popularity. Consider the comments of one of the experiments, Duncan Watts:
“Because the long-run success of a song depends so sensitively on the decisions of a few early-arriving individuals, whose choices are subsequently amplified and eventually locked in by the cumulative-advantage process, and because the particular individuals who play this important role are chosen randomly and may make different decisions from one moment to the next, the resulting unpredictability is inherent to the nature of the market. It cannot be eliminated either by accumulating more information — about people or songs — or by developing fancier prediction algorithms, any more than you can repeatedly roll sixes no matter how carefully you try to throw the die.”
In other words, it seems that often popularity is more the result of random events that occur early on than of any innate quality.
Profound as this experiment may be, it is not an entirely new observation. Self-fulfilling prophecy’s are already a well known phenomenon in a number of fields. In medicine it is know that a placebo– a drug with no active medical ingredients– can still have a positive effect on patients if they believe it will have a positive effect. Similarly, in education if a teacher believes that she is teaching a group of high acheiving students then those students will subsequently perform better on tests.
In fact, it seems that social-influence effects plays a role in any decision that is the least part social.
As a final thought consider the notion of popularity as it applied in high school. (It is often rumored that the careers of social sciences are based on explaining their high school experiences.) Although some of you may have occupied the enviable position of one of the popular kids in high school a lot of you probably felt left out on the side. This, however, may not have been your fault. It is quite possible that small friendships formed through random events in the early years of high school may have set the hierarchies of popularity in an almost completely arbitrary manner.
On a more serious note, if nothing else, I think that this research should create a sense of humility. To those who have done well, yes it is true that you are most likely better than most at what you do. To some extent though, it is likely that random social-effect also played a role in your success. Success is never just an individuals making, randomness always plays an important role too.