Traditional social networks fueled Twitter’s spread

February 2012

PROBLEM

Conventional wisdom tells us the Internet has flattened the world, creating social networks that are independent of geography and socioeconomic status, and making it possible for ideas to spread quickly across these new social networks without the help of news media. For instance, Facebook and the microblogging platform Twitter allow friendships, news and ideas to travel around the world instantaneously via email, texts and tweets. These and other innovations like smartphone applications are also affecting the way physical networks and systems are used, and influencing the types of new infrastructure required. Yet, we lack a scientific understanding of the “contagion process” of the spread of the new technology platforms themselves.

APPROACH

Traditional marketing studies of the diffusion of innovations have characterized the adoption of expensive, durable goods (such as cars and household appliances), but these models don’t apply to Internet technology platforms that are low-cost or free and that achieve success only when adopted by many people. Similarly, traditional contagion models of the spread of disease can’t be applied effectively to the contagion process of social network platforms, because those models often rely on overly simplified parameters, such as the homogeneous mixing of populations.

Professor Marta González and graduate student Jameson Toole used statistical physics and network theory to model the contagion process of the growth in the number of Twitter user accounts in 408 U.S. cities during the company’s first three years of existence: from May 2006, when there were a couple of hundred users, through August 2009, when Twitter users reached about 3 million. Using marketing terminology that describes individuals, González and Toole characterized cities as early adopters, early majority adopters, late majority adopters or laggards, depending on when Twitter accounts in a city reached critical mass — defined as the day when user accounts reached 13.5 percent of the total number of Twitter users in that city. As a means of quantifying media influence, they used historical search-engine data to count news stories about Twitter.

FINDINGS

González and Toole found that Twitter’s growth in the U.S. relied on traditional social networks that are based in geography and homophily (a preference for friendships with people who share similar tastes and live nearby one another), rather than the more random contagion process predicted by conventional wisdom. They also found that media influence was essential to this process, but had to be included as random spikes, not the more customary constant, because the random spikes more accurately reflect the way the growth in the number of Twitter user accounts influenced media coverage and vice versa.

Twitter’s popularity initially spread from its birthplace in San Francisco via young, tech-savvy innovators to greater Boston. It then changed routes and began traveling short distances, which implies that face-to-face interaction was influencing its adoption. Media attention — which spiked in August 2009 when Ashton Kutcher reached 1 million followers and Oprah Winfrey ceremonially sent her first tweet — was responsible for quadrupling the number of Twitter user accounts.

IMPACT

This work creates a new way of modeling the effects of social network structures on the adoption of innovative technologies by combining in a single analysis three different dimensions: geography, social networks and media coverage. This new modeling methodology makes it possible to more accurately predict a social contagion or the adoption pattern of innovative technologies at the city-to-city scale, which may influence other models of human dynamics in transportation and other network systems.

MORE

A paper on this work by González, Toole and Professor Meeyoung Cha of the Korea Advanced Institute of Science and Technology appeared in the Jan. 19 issue of PLoS ONE. González integrates methods of complex systems with statistical physics approaches, computational sciences, geographic information systems and network theory to characterize and model human dynamics.

Each circle corresponds to a city’s total number of Twitter user accounts during the three years of data collection. Watch a video of Twitter adoption over time.  Image / Jameson Toole

Download this issue as a pdf.