Unstable diffusion in social networks
本文提出一个分析框架，比传统平均场方法更准确地描述复杂网络上的扩散动态，发现多种活动的扩散表现为鞍点路径，本质不稳定，且当活动可替代性高时，即使事前吸引力相同，也会随机出现一种活动主导另一种。
How and to what extent will new activities spread through social ties? To answer this question, we present an analytical framework that allows us to describe the diffusion dynamics on complex networks more accurately than the conventional mean-field approach. Based on two classes of network games, we find that the spread of multiple activities is expressed as a saddle path, and thus, inherently unstable. In particular, when the two activities are sufficiently substitutable, either of them will dominate the other by chance even if they are equally attractive ex ante. We argue that, in environments where such symmetry-breaking occurs, any average-based approximation method may not correctly capture the Nash equilibrium — the steady state of an actual diffusion process.