Skip to main content

Information Cascade in Music

This summer, I attended Jon Kleinberg’s lecture entitled, “What Facebook, Amazon and Google teach us about society and about ourselves,” which described the appearance of large networks and how information flowed through them.

During the lecture, he brought up a very interesting study of popular songs in music-sharing websites which created “alternate reality.”  This was achieved by grouping the site’s visitors into several groups, and putting each of those groups onto a different version of the website.  These versions were identical except they accurately reflected their perspective group’s music choices, and posted listings of music that was most popular within the given group.  The music that all versions of the website shared was limited to lesser-known Indie music; that had not been completely subjected to mainstream pop culture.

After running the website for some time, the study compared the lists of “top” (highest-rated) songs among the different versions of the website.  They noticed that each list was pretty different with the exception that the lowest rated songs in one group were never the highest rated songs in another group.  This shows that while different groups shared some commonality in recognizing truly bad music, which songs became the most popular was highly random, often reflective of music that the first few users of the website had liked.  This situation is a perfect example of an information cascade.

An information cascade can be defined as a situation where even against logic; a person will discard their private knowledge regarding a matter and follow a crowd.   This occurs when there is a choice to be made, and people make the decision sequentially.  Each person has his or own knowledge of the situation, but does not know anyone else’s knowledge, simply his or her actions.

In the music-sharing website, the first few users had knowledge of their own tastes, and chose to rate songs according.  As more people began using the website, people saw what others had listened to and liked.  As in most information-cascade situations, the new user became affected by the actions of the crowd.  Instead of picking one of the many random songs that the music-sharing website offered, users listened to songs others had liked.  Even if the new user did not like a song that much, they would still rate it higher because of other’s actions.  As a result, if a song is initially liked, it will likely exhibit ever-increasing popularity.   The list of “top” songs does not actually reflect a group’s true music tastes as much as it reflects the first few users’ music tastes due to the persuasion of the information cascade.



Leave a Reply

Blogging Calendar

November 2011
« Oct   Aug »