|Scooped by luiy|
In order to identify generic features of online diffusion structure, we study seven diverse examples comprising millions of individual adopters. As opposed to biological contagion, our domain of interest comprises the diffusion of adoptions, where “adop- tion” implies a deliberate action on the part of the adopting individual. In particular, we do not consider mere exposure to an idea or product to constitute adoption. Conta- gious processes such as email viruses, which benefit from accidental or unintentional transmission are therefore excluded from consideration.
Although restricted in this manner, the range of applications that we consider is broad. The seven studies described below draw on different sources of data, were recorded using different technical mechanisms over different timescales, and varied widely in terms of the costliness of an adoption. This variety is important to our con- clusions, as while each individual study no doubt suffers from systematic biases arising from the particular choice of data and methods, collectively they are unlikely to all ex- hibit the same systematic biases. To the extent that we observe consistent patterns across all examples, we expect that our findings should be broadly applicable to other examples of online—and possibly offline—diffusion as well.
The remainder of this paper proceeds as follows. After reviewing the diffusion liter- ature in Section 2, in Section 3 we describe in detail the seven domains we investigate. We present our main results in Section 4, showing that not only are most cascades small and shallow, but also that most adoptions lie in such cascades. In particular, it is rare for adoptions to result from chains of referrals. Finally, in Section 5 we discuss the implications of these results for diffusion models, as well as the apparent discord between our results and the prevalence of popular products, such as Facebook and Gmail, whose success is often attributed to viral propagation.