Twitter and Power Laws
As we discussed in class, power laws are ubiquitous in the dissemination of social media. While our textbook tells us that determining the reasoning behind the robustness and consistency of power laws from the perspective of individual decision-making continues to be an elusive research goal, we can nonetheless build models like the “rich get richer” process we covered in class based on the propensity of people to copy each other and the consequences of this tendency.
We can see an application of the concept of power laws on Twitter. In particular, in the paper “On the Frequency Distribution of Retweets“, the authors propose that the frequency distribution of retweets follows a power law distribution. According to the authors, Twitter – and particularly the amplification of information driven by retweeting – plays a crucial role in the dissemination of information on the internet. The authors propose the concept of “preferential attachment and transmissibility of tweets” to attempt to explain why some tweets become more popular than others. Preferential attachment “means a user is more likely to retweet a tweet that has already been retweeted many times”. “Transmissibility of tweets” refers to a parameter intrinsic to the tweet that defines the probability of a tweet being retweeted.
50,000 tweets were collected over a span of a month with 42180 participants. Results indicated that users retweeted tweets that were already popular to begin with. The authors propose that both preferential attachment and transmissibility contribute to the probability of retweeting. That is, both the frequency with which the tweet has already been retweeted intrinsic characteristics about the tweet itself dictate the probability that the tweet will be spread.
Based on this information, the authors propose that the power law modeling retweet patterns has an exponent of 0.6-0.7 and are able to develop a model that successfully reproduces empirical data based on their assumptions about preferential attachment and transmissibility. It is another example the robustness of the power law in its ability to describe empirical data on social networks.