Novelty and Collective Attention

The subject of collective attention is central to an information age where millions of people are inundated with daily messages. It is thus of interest to understand how attention to novel items propagates and eventually fades among large populations. We have analyzed the dynamics of collective attention among one million users of an interactive website – \texttt{digg.com} – devoted to thousands of novel news stories. The observations can be described by a dynamical model characterized by a single novelty factor. Our measurements indicate that novelty within groups decays with a stretched-exponential law, suggesting the existence of a natural time scale over which attention fades.

💡 Research Summary

The paper investigates how collective attention to novel information spreads and eventually fades within a large online community. Using a massive dataset from the social news platform Digg.com, the authors tracked the behavior of roughly one million users as they interacted with thousands of newly posted news stories. For each story they recorded the time‑stamped “digg” votes, which serve as a proxy for the level of public interest, and aggregated these votes in hourly intervals to construct an attention curve A(t) that captures both the rapid rise and the gradual decay of attention over time.

Initial descriptive analysis revealed a characteristic pattern: a steep increase in votes during the first two to three hours, a peak typically reached within that window, followed by a long tail of low‑level activity that can persist for many hours. Traditional diffusion models—such as simple Poisson processes or classic epidemic‑type models—fit the early surge reasonably well but systematically underestimate the magnitude of the long‑term tail. This discrepancy motivated the authors to introduce a single “novelty factor” N(t) that modulates the effective propagation rate of a story as it ages.

The dynamical model is expressed by the differential equation

 dA/dt = λ · N(t) · A(t)

where λ is a baseline propagation constant (empirically estimated around 0.45) and N(t) is a time‑dependent decay function bounded between 0 and 1. The authors hypothesized that N(t) follows a stretched‑exponential (Kohlrausch) form:

 N(t) = exp