# The statistical mechanics of Twitter

## TL;DR

This paper applies statistical physics models to Twitter data, capturing social state distributions and revealing criticality and scaling behavior, thus offering a new perspective on complex social systems.

## Contribution

It introduces maximum entropy models for Twitter social states, accurately describing higher order correlations and revealing critical phenomena in social networks.

## Key findings

- Models match pairwise correlations in Twitter data
- Parameters are near critical surfaces indicating phase transition behavior
- Scaling behavior observed under coarse-graining of data

## Abstract

We build models for the distribution of social states in Twitter communities. States can be defined by the participation vs silence of individuals in conversations that surround key words, and we approximate the joint distribution of these binary variables using the maximum entropy principle, finding the least structured models that match the mean probability of individuals tweeting and their pairwise correlations. These models provide very accurate, quantitative descriptions of higher order structure in these social networks. The parameters of these models seem poised close to critical surfaces in the space of possible models, and we observe scaling behavior of the data under coarse-graining. These results suggest that simple models, grounded in statistical physics, may provide a useful point of view on the larger data sets now emerging from complex social systems.