# Novelty and collective attention - Author(s): Fang Wu and Bernardo A. Huberman - Date: 2007 - Publication: PNAS - [Link](https://www.pnas.org/content/104/45/17599) - Note Created: 2022-02-12 @ 10:03 --- ## Summary The authors analyze the rise and fall of articles on [digg.com](https://digg.com/). Specifically, they look at how long articles stay on the front page, which (at least at the time) appears to be 100% driven by "diggs" (upvotes) of users. The nature of the site's dynamics allowed them to examine the distribution of time at which articles remained on the front page (i.e., how long articles remained popular). Since each article also earned "diggs", the authors could analyze the distribution of the *total diggs* earned by each article, offering some measure of "total popularity". At the end of the day, they devise a simple dynamical model with a single "novelty" factor. This model allows them to estimate how an article's popularity decays over time. They show that popularity decay's with a stretched-exponential law, suggesting that there is some natural time scale over which attention fades. The exact time half-life that they estimate articles will naturally fall from the front page is 69 minutes, which they indicate agrees with the typical 1-2 hours they observe in their data — allowing for variation in new story topic, and other factors. ### Figures ![[attentionDynamics_noveltyAttention1.png]] ![[attentionDynamics_noveltyAttention2.png]] ![[attentionDynamics_noveltyAttention3.png]] --- #### Related [[attention_dynamics]] [[modeling]] [[collective_attention]]