When ideas go viral: Early warning signals in theoretical and real-world social contagion systems

Lexi Lerner, a biology major from Brown University, worked with Dr. John Drake to look for critical slowing down in consumer fad systems.

Abstract: Consumer fads are often characterized by unpredictable explosive outbreaks. Early warning signals have been retroactively successful at anticipating critical phenomena of complex systems such as infectious disease epidemics, but they have yet to be extended to consumer fads. Here we study both theoretical and real-world consumer fad systems to see whether their approach to an epidemic is characterized by critical slowing down (CSD). We propose a new model of social contagion transmission that includes the accumulation of buzz, or aggregate ubiquity, around an idea. We derived deterministic and stochastic solutions for this model and showed that it can push a subcritical system to criticality. We evaluated four candidate early warning signals by their sensitivity and specificity using various rolling window bandwidths to understand CSD detection performance. We also analyzed data for a faddish product line over a two-year timespan to determine if CSD was detectable in real-world systems. Our results show that variance was the best-scoring and most consistent predictor (AUC > 0.8) of CSD in the theoretical stochastic SIBR model. It was also determined that CSD is present in the product line data and, at small bandwidths, is characterized by “breaking points” that can be traced back to events in the original time series. We hope these findings will help guide statistical analysis of consumer behavior and further early warning signal analysis of economic interest.

Download (PDF, 1.96MB)