Is it free will or is it cascade?
https://www.princeton.edu/~mjs3/salganik_dodds_watts06_full.pdf
In SOC 2580, “Six Pretty Good Books”, the only book that has been listed on the syllabus for several consecutive semesters is “Everything is Obvious, Once You Know the Answer: How Common Sense Fails Us” by Duncan Watts (link). In his book, Watts summarizes one of his most famous study “Music Lab”.
When we think about famous arts, music, and literatures, the account for their success we offer is usually centered around their particular intrinsic quality. For example, many people conclude that “Harry Potter” books has all the element needed for a successful children’s book: for example, it has a character arc of the hero, who fights the evil guy, and a prophecy predestines his fate.
However, the quality argument ignores the fact that the famous book series was rejected by 9 publishers before its sudden rise to stardom, and another similar book that J. K. Rowlings wrote under a pseudonym received little splash.
On the other hand, it is very hard to predict what the next fad on the social media: the huge success of Gangnam Style was a surprise to everyone. Although someone argues that some of its features have lead to its epidemic popularity, a lot of people think its quality does not justify its fame.
Thus it leads us to the natural question: are the famous works success because of they’re good, or they’re just lucky?
Duncan Watts with two other scientists, Matthew Salganik and Peter Dodds, conducted the landmark study “Music Lab”. In the experiment, experimenters created a virtual “music market” comprised of over 14 thousands participants, who can rate and download from a list of previously unknown songs from unknown bands. The participants are randomly assigned to nine “parallel universes”: one of them is the “independent world”, in which participants make decisions only based on the names of bands and songs, and the other eight worlds are “social influence worlds”, where participants are provided with information of previous participants’ choices.
Compared to the model of blue and red marbles that we discussed in class, this setting offers a weaker signal or pressure to the participants. In Music Lab, when they’re offered other people’s opinions, the participants can rely solely on their own discretion and music taste to make the decision. However, in the urn model of red and blue marbles, everyone after the first student must base their decision on previous people’s choices.
In the independent world, therefore, the result (which songs get more downloads and which songs are unpopular) corresponds to the validity of the quality argument: based on the population’s preferences for music, the popularity a song receives is depended only on its quality. The result of the social influence world, on the other hand, sheds light on the influence of other people’s opinions of decision making, when compared with the independent world’s result. For example, if a song is downloaded many times in the social influence world but is not well received in the independent world, it would indicate that its success comes mainly from social influence, i.e., it’s popular because it’s popular. If a song is popular in the independent world (indicating that it’s “good”) but is not in the social influence world, it can be inferred that the unpredictability of popularity.
The results indicate that in social influence worlds, the disparity between popularity of songs is greater than that of the independent worlds, i.e., reviews are more polarized. The unpredictability between worlds is higher in the social influence setting, indicating that social influence increases unpredictability. What’s more, inconsistency across worlds is observed, in other words, a hit song in world A may be a bust in world B.
How do we explain between-world inconsistency? With a close examination of the process of the experiment, we can see a very similar cascade of opinions that is illustrated in the lectures.
With 48 songs to choose from, the participants in the Music Lab has far more leeways than the students in the urn model, in which they’re guessing from 2 options. With a wide range of songs one can go to, in a social influence setting, it is more likely for users to first try what is “trending” and only try what is “trending”. Therefore, the information of others’ opinions is still salient in the decision making progress. When we trace back in time to when the first user in one universe make a choice, and there’s no previous information available, the first mover would choose based on his own preferences. The second user, however, is more likely to choose listen to what the first users like than other songs, because when they know nothing about these songs (other than their names), information of the first mover’s preferences is all they could refer to when choosing which songs to listen to, assuming all users only want to listen to “good” songs.
Therefore, we see a cascade in the social influence worlds, and the first movers’ choices are the most influential to the whole world’s choices. It explains the between-world inconsistency: the first movers choose randomly, but their random choices can shape the whole world. We also observed it in the blue marble/red marble model in class, in which if the first two students both guess MB, then the following student would guess MB regardless of what they actually observe.
The Music Lab proves the existence of cascades in real life, and points out a direction for future researches on how public opinions are shaped. For example, it would be a very interesting question to ask ourselves: are my thoughts originated from free will or just from cascade?