Source-linked AI summary
Novelty and Collective Attention
Fang Wu, Bernardo A. Huberman
TL;DR
The paper asks how novelty-driven attention propagates and fades in very large natural groups, where prior evidence was mainly from small studies and theory. It analyzes one million Digg users and thousands of stories with a stochastic propagation model, finding that collective novelty decay follows a stretched-exponential law with a characteristic timescale of about one hour.
Problem
Large-group collective attention lacked empirical evidence from a natural, non-laboratory setting despite its importance for information propagation, advertising, and viral marketing.
Method
The authors analyze popularity growth and decay for thousands of Digg stories among one million users and validate a stochastic model with a time-dependent novelty factor.
Results
rt fits a stretched exponential, rt ∼e−0.4t0.4, and has a half-life τ = 69 minutes, or about one hour.
Takeaways & Limitations
A single novelty factor describes the growth and decay of collective attention and determines the natural timescale over which attention fades.
Takeaways & Limitations
The analysis assumes story digg numbers can only grow because burying a story rarely happens, and the complex front-page qualification algorithm is not discussed.
Abstract
from arXiv · showhide
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.
Abstract
The paper studies collective attention by analyzing one million users and thousands of Digg stories, combining large-scale observations with a stochastic model of story propagation and novelty decay. Attention growth is followed by stretched-exponential decay, with a characteristic timescale of about one hour.
- Motivation and data: Large-scale Digg data address the lack of empirical evidence on collective attention in very large, natural groups.The study analyzes one million users interacting with a user-driven news website and tracks thousands of stories.
- Motivation and data: Digg users submit stories, assign explicit popularity through diggs, and determine which stories reach or leave the front page.Stories that gain enough diggs quickly become popular and appear on the front page, while older stories are replaced or remain longer as Top 10 stories.
- Dynamical model: The model represents story growth as discounted random multiplicative propagation, with a positive novelty factor that decreases over time.The dynamics are Nt = (1 + rtXt)Nt−1, where rt starts at 1 and decreases toward zero, curtailing growth as novelty fades.
- Dynamical model: The model predicts approximately log-normal story popularity because log popularity is a discounted sum of random variables.The tracked distributions support this prediction, including a two-hour Kolmogorov-Smirnov normality-test p-value of 0.5605 for log(Nt).
- Empirical validation: 6.947 is the observed slope relating the sample mean and variance of log(Nt)−log(N0) across 1,110 stories, as predicted by the model.The points for times from 1 to 1,440 minutes lie roughly on a line through the origin.
- Attention decay: rt follows a stretched-exponential form, rt ∼e−0.4t0.4, with a half-life τ = 69 minutes, or about one hour.The decay is faster than a power law and slower than an exponential; its stretched form is consistent with multiple characteristic relaxation times.