Figure 2
(A) Our dynamical theory describes an evolving population of boundedly-rational individuals21 pursuing individual gain in a setting where there are winners and losers. Each sporadically receives information I and decides how to act using their individual p value, and changes it when it becomes unreliable. The distribution of p values at time t is the heterogeneity distribution P(p, t). (B) Evolution of polarization in our controlled laboratory experiment (orange line, see Methods and Data) and our theory (blue line), with R ≥ 1. The theoretical curve is an average over 1000 runs where the L/N value is chosen in the [0.45, 0.55] interval, while the error bars represent the standard deviation. (C,D) Two next-generation social media algorithms which actively connect individuals with the aim of promoting alignment and diversity respectively. See Methods and Data for meaning of C(Sij). (E) Our results are insensitive to how these two algorithms are implemented, i.e. choosing k individuals (nodes) or λ links. (F) Example flow-chart of these two algorithms.
