The Anatomy of Large Facebook Cascades
https://www.facebook.com/notes/facebook-data-science/the-anatomy-of-large-facebook-cascades/10151549884868859/
In class we discussed the concept of information cascades and how the cascade actually begins. In an information cascade each person can observe what other people do but not what they can see, and they use the information deduced from other people’s actions to make a choice on what action they should take. In essence, a person first gathers information based on what they observe other people do, make inferences based on these observations, then use these inferences to change what they believe and act accordingly. In an information cascade, if enough people believe the same thing, then it will cause everyone around them to believe the same thing as well and act accordingly. Thus an information cascade has begun.
However, on Facebook there is a different but still similar kind of cascade. A prominent feature on Facebook is the share button and the phenomenon stemming from users ability to share posts is known as viral posts. It is similar to information cascades were a user is affected by the network around them in choosing its action (to share or not to share). The article above studies two posts that went extremely viral: Obama’s victory photo when he won the 2012 election, and a “million likes meme” posted by Facebook user Petter Kverneng. Both posts received hundreds of thousands of likes and shares but the factors that caused this cascade and how it spread are quite different.
The importance of the content being shared to the user is a large factor in whether or not the user will share the Facebook post or not. With Obama’s victory photo, it was shared in very large quantities in the first hour of its posting before sharply falling off afterwards. The results of a presidential election is an important piece of information that would have been widely shared by many users of Facebook, and it would be especially even more important to Facebook users of a Democratic political leaning. This is reflected in the data analyzed from the cascade of sharing Obama’s victory photo post. There were two bumps in the sharing of Obama’s victory photo, when Michelle Obama and Alicia Keys shared the post. These are two important and popular figures in America so them sharing the photo would have also increased the importance and therefore likelihood that a Facebook user would share the post. Additionally, about 50% of the sharers of Obama’s victory photo were people following Obama’s Facebook page and about 80% of the sharers were people with liberal political views. If you were someone who followed Obama already or were liberal, Obama winning the presidential election would be a more important post to you so you would be more likely to share the post on Facebook.
With the “million likes meme” the cascade occurred a little bit differently. Instead of an initial massive amounts of sharing like with Obama’s post, the sharing amounts as time went on was relatively even with an increase as the photo was nearing 1 million likes. After the photo reached 1 million likes the amount of shares dropped off. In this case I believe the importance of the post to the user also affected the share rates similarly. The meme was initially not that important and did not matter to many people in its first stages of viral-ness so the amount of sharing was relatively low. Many users may have felt that there sharing would not have an impact in the success of the million likes goal so may have ignored it. However, as it neared 1 million likes, users may feel that sharing it will more of an impact in getting it to a million likes and were more invested in the success of the post. This caused the recorded increase in shares at the 17 hour mark.
There are evidently some similarities between the behavior of resharing and information cascades, although the reason it develops may differ quite drastically.