by Bruno Ribeiro (last updated Feb 6 2014)
Carnegie Mellon University Press Release: In Age of Information Overload, Ability To Sustain Attention Determines Success | CMU School of Computer Science Press Release
In light of the recent Princeton x Facebook "feud", I would like to weigh in with my recent peer-reviewed study to appear at the upcoming 23rd International World Wide Web Conference, where I propose a model to predict social media websites’ popularity and demise by modeling Internet users’ attention as a scarce resource that must be “consumed” by a website. Users’ interactions with the website also create content that consumes the attention of other users. For instance, if the attention we dedicate to Facebook incites us to create content that helps Facebook capture a larger share of attention of our friends, then our friends will also produce more content (on average), which in turn captures a larger share of our attention. This positive attention feedback loop drives Facebook user activity up. But as the attention we are willing to dedicate to Facebook is limited, our (average) activity eventually stabilizes and remains stable. Conversely, if the share of our attention consumed by the website does not drive enough attention and content creation from our friends to keep our current attention level, then the negative attention feedback drives the website to its death. In this context my model predicts that Facebook popularity will remain stable, at least until other websites capture a larger share of our attention. The model also accurately predicted the steep popularity decay of the Tea Party’s two main sites, whose popularity have dropped seven fold since their peak in 2010.
The model fit is parameter-free, fully automated (no tweaking allowed). My study also argues that website popularity cannot be measured by Google search data (one of many methodological flaws in the much critiqued Princeton study). The model classifies websites into self-sustainable versus unsustainable and whether their startup growth is mostly driven by marketing & media campaigns or word-of-mouth. While this is a preliminary study and more research is needed to further understand the “marketplace of attention” and its impact on Internet startups, the preliminary results are promising and point to an even more sensationalist and competitive future for social media.
My work was inspired by a 1969 lecture of Nobel Laureate Herbert A. Simon in which Simon argued that an information processing system (say, Facebook) will reduce our net demand of attention only if it absorbs more information, previously received by others, than it produces – that is, if it listens and thinks more than it speaks. I believe that Facebook and many similar membership-based sites are "talkers", information processing systems that speak more than they help us digest information, thus making us live in a "marketplace of attention" where companies such as Facebook and Twitter, and social movements such as the Tea Party and Occupy Wall Street only survive online by relying on their users (followers) to intensely generate content to compete for attention.
PS: I disagree with Facebook's argument that "correlation is not causation" is the source of the methodological flaws in Princeton's Cannarella and Spechler study. Yes, the Cannarella and Spechler study is riddled with methodological flaws, but mixing correlation with causation is not one of them. A historical example to illustrate my point: Kepler, despite not knowing what caused the celestial orbits, was able to predict the trajectories of all outer planets with great accuracy.Tweet
Pointer to the article:
Bruno Ribeiro, Modeling and Predicting the Growth and Death of Membership-based Websites, WWW 2014 [arXiv:1307.1354].