Insights into Social Media Data using Entropy Theory

Statistical mechanics, built upon the concept of entropy from classical thermodynamics, is used to reveal properties of physical systems and describe them in great detail based on properties such as thermodynamic states (entropy and enthalpy) and constants (heat capacitance). This method of gaining insight into macroscopic behaviour of a system by treating it as a set of microscopic interactions is particularly effective when a large number of entities are concerned–such as particles in a gas or ions in a plasma cloud.

An exploratory study is being conducted into whether these methods can be used to describe the macroscopic behaviour of a social media platform by prescribing analogous measures of density, temperature, and particle velocity to traits such as the number of tweets on a topic, their engagement metrics, and the growth of interest over time.

Findings could indicate suitability of the method to answer key questions often asked of social media data:

• Does the total amount of potential “energy” (user engagement) of a platform fluctuate over time? If so, how?
• Can we estimate how much of it is currently being channeled?
• Can we estimate the relative potential energy of different topic areas (politics, celebrity gossip, sports, etc.)?
• And most importantly: if we can analyse and understand these factors, could this provide tangible benefit to users, such as points in time where certain topic areas have optimal engagement potential?

Researchers posit this data mapped as entropy-over-time will be a more effective analogy for the lifetime of a trend that the oft-applied technology adoption curve or other metrics. Using only Twarc and a bunch of good old-fashioned scripting, we present and evaluate initial findings from an investigation into historical datasets from Twitter, and discuss what this means for treating social media platforms as closed, intra-connected systems for analysis.