In this paper we develop a model to show that the network of agents adopting a particular technology may have characteristics that differ significantly from the social network of agents over which the technology spreads. For example, the network induced by a cascade may have a heavy-tailed degree distribution even if the original network does not. This provides an alternate explanation for certain properties repeatedly observed in data sets, for example network densification, shrinking diameter, and network community profile. These properties could be caused by a sort of sampling bias rather than by attributes of the underlying social structure. By generating networks using a cascade over traditional models, we can reliably reproduce these properties using traditional network generation models that do not themselves contain these properties.
This provides evidence that online social networks may look fundamentally different from social networks. This indicates that using data from digital social networks may mislead us if we try to use it to directly infer the structure of social networks. We then discuss implications, including learning properties of the underlying network assuming such a model.