Diffusion
Source: Akbarpour, Mohammad, and Matthew O. Jackson. “Diffusion in Networks and the Virtue of Burstiness.” PNAS, National Academy of Sciences, 24 July 2018, https://www.pnas.org/content/115/30/E6996.
In their work, Akbarpour & Jackson examine the varying effects different types of individual behavior have on the spread of information throughout a network (2018). According to the authors, individual behavior is influenced by “time” and “history”: the former being extremely influential with respect to the likelihood of different nodes adapting to a certain kind of behavior. Akbarpour & Jackson discuss “burstiness”, which is a term used to describe “consecutive periods of activity followed by consecutive periods of silence” (2018). Prior to the results of this study, burstiness was a phenomenon that thwarted diffusion. However, in this study, the researchers discovered that “burstiness” enhances diffusion when the network is composed of a diverse group of individuals with varying levels of “activity” (2018). There were three different types of activity (or behavior) within the network model: sticky nodes represent individuals who maintain a certain type of activity — whether that be active or inactive — for a long duration of time (2018). On the opposite end of the spectrum are reversing nodes: thus, these nodes represent individuals who shift between active and inactive regularly (2018). Lastly, “Poisson” nodes represent individuals whose levels of activity are “randomly on or off in every period with the same probability independently of history” (2018). The researchers conclude that optimal diffusion occurs when a sticky node is transmitting information and a reversing node is “receiving” (Akbarpour & Jackson 2018).
This study relates to conversations in class regarding information cascades and the circumstances in which certain behavior is adopted. In contrast to this study, we looked at the adoption of information with respect to clusters, which represent tightly knit communities comprised of individuals with similar values. Rather this study breaks the diffusion process down even further by observing the behavior of each node and accounting for individuality. This article puts into perspective how complex an information cascade is given other influences such as time, probability, and a node’s past behavior in addition to thinking about the spread of information at the individual level rather than at the community level.