Collective attention to social media evolves according to diffusion models

Abstract

We investigate patterns of adoption of 175 social media services and Web businesses using data from Google Trends. For each service, we collect aggregated search frequencies from 45 countries as well as global averages. This results in more than 8.000 time series which we analyze using economic diffusion models. The models are found to provide accurate and statistically significant fits to the data and show that collective attention to social media grows and subsides in a highly regular manner. Regularities persist across regions, cultures, and topics and thus hint at general mechanisms that govern the adoption of Web-based services

The diffusion of mobile social networking: Exploring adoption externalities in four G7 countries

“…In this case, the S-shape for non-interactive innovations should be less pronounced, more right skewed, having a smaller q/p ratio than that for interactive innovations, which are more left skewed. Bauckhage, Kersting, and Rastegarpanah (2014) also use the Bass and G/SG models jointly for comparing )…”

Using Google Trends Data to Gauge Interest in Virtual Worlds

“…Research has shown that Google Trends can outperform surveys in predicting consumer behavior (Vosen & Schmidt, 2011). In this section, we will critically examine a study by Bauckhage, et al (2014aBauckhage, et al ( , 2014b that used Google Trends data to gauge the public interest in 175 social media products (e.g. Facebook, YouTube, Twitter), including some social virtual worlds.…”

“…In studying the diffusion of social media usage, Bauckhage, et al (2014aBauckhage, et al ( , 2014b assumed that Google Trends popularity was a proxy for "collective attention." To model and predict changes in collective attention, they tested the effectiveness of three different diffusion models in fitting Google Trends data: the Bass model (Bass, 1969), the shifted Gompertz model (Bemmaor, 1992), and a third function also used in diffusion studies, the Weibull model (Rinne, 2008).…”

Kernel Archetypal Analysis for Clustering Web Search Frequency Time Series

“…This, in turn, rules out the use of many commonly used and well established pattern recognition techniques. However, previous work on Web intelligence has shown that general trends in time series of human attention dynamics are well accounted for by statistical life time distributions and that lifetime distribution provide plausible socio-mathematical models of the growth and decline of attention to novelties [11], [13], [22]- [24]. In particular, the work in [13] has established the Weibull distribution as a theoretically well founded and empirically valid model of general trends in search frequency data.…”