Cascade and threshold models are widely used to predict information diffusion in social networks, yet their characterization of networks as static and monoplex limit their ability to accurately predict how information propagates in dynamic, multiplex social environments. Using data from a peer-led HIV prevention intervention for homeless youth, we determine whether manipulating the baseline social network by (1) adding ties observed at later time points, and (2) accounting for alternative relational contexts improves each model’s predictive accuracy. Results show that the addition of new ties improves the performance of both models, while substituting the context of interaction yields only minor improvements.

级联和阈值模型被广泛用于预测社交网络中的信息扩散，但是它们对网络的描述是静态的和单一的，这限制了它们准确预测信息在动态的，多元的社交环境中传播的能力。使用来自同行的针对无家可归的年轻人的艾滋病毒预防干预措施的数据，我们确定是否通过以下方式操纵基线社交网络：（1）添加在以后的时间点观察到的纽带，（2）考虑其他关系背景是否可以提高每种模型的预测准确性。结果表明，添加新关系可以改善两个模型的性能，而替换交互上下文只会产生较小的改进。