Rumor diffusion and information cascades
http://science.sciencemag.org/content/359/6380/1146
While I was doing my readings for another class (INFO 2450), I couldn’t help but connect concepts I learned in this class with one of the papers. This assigned paper, titled “The spread of true and false news online”, is a study that analyzed true and false tweets on twitter and how fast these tweets spread depending on their accuracy. It concluded that “falsehood diffused significantly farther, faster, deeper and more broadly than the truth”. The most notable explanation for these results is that false news is found to be more novel than true news which is more appealing and fascinating to viewers. In particular, their discussion on rumor diffusion reminded me exactly of the network cascades that we learned about. In fact, this study directly addresses large-scale information cascades with regards to spread of false information.
It would be intriguing to graph out rumor diffusion with nodes representing twitter users and the cascade of users spreading the false news. I am curious for this study to be carried out on a college campus. False news about friend drama, professors, campus events and other gossip happens every day. Additionally, organizations like Greek life, large club organizations and smaller campus communities perpetuate clustering of networks and would make it even more fascinating to examine the various cascades and compare those with true news that spreads around campus. In class, we learned about the dynamics of networks and how nodes can switch until equilibrium. Given a situation in where there are two sides, one true and one false side, to the same story, we could graph out a network to examine how the network model impacts what side is believed or not. Decisions are based off of who you are connected to in a network; therefore, you are more likely to match the behaviors of your neighbors. Through calculation of the fraction choosing a particular decision, total number of neighbors and the payoffs of either decision, we learned that a node will switch to decision a (out of 2 chooses a or b) if f (fraction choosing a) is greater than or equal to b over a plus b. From this, we can calculate a “threshold”. This threshold rule determines how likely a node is to adopt a given decision and can be influenced by different biases. In the context of a college campus rumor, these biases could be if a story has a political leaning that a majority of the school has. This means that you could gain more payoff in forms of say, social acceptance, from one decision. Ultimately, I found it most interesting that concepts from both of the information science classes I am taking this semester connected so well together. Network concepts are applicable in so many other contexts beyond this classroom!