Abstract: We introduce a model for predicting page-view dynamics of promoted content. The regularity of the content promotion process on Wikipedia provides excellent experimental conditions which favour detailed modelling. We show that the popularity of an article featured on Wikipedia’s main page decays exponentially in time if the circadian cycles of the users are taken into account. Our model can be explained as the result of individual Poisson processes and is validated through empirical measurements. It provides a simpler explanation for the evolution of content popularity than previous studies.
Introduction: The Challenge of Predicting Attention
Predicting the "shelf-life" of online content is notoriously difficult. On platforms like YouTube or Reddit, popularity is often a "rich-get-richer" phenomenon: highly-voted content receives more visibility, which leads to even more engagement, creating a feedback loop that obscures the underlying dynamics of human interest. To isolate the true nature of how collective attention fades, researchers turned to Wikipedia as a "natural laboratory."
Unlike other social media platforms, Wikipedia’s Main Page promotion is remarkably disciplined. A "Featured Article" is highlighted for exactly 24 hours and then moved to a "Recently Featured" list for the subsequent three days. This fixed exposure duration provides a clean, controlled environment to study how attention naturally decays without the noise of algorithmic ranking or variable promotion times.
Methodology: Decycling and Modeling
The core innovation of this study is a two-step approach involving Circadian Correction (Decycling) and Exponential Modeling.
Circadian Correction
Wikipedia traffic follows a strict circadian rhythm, primarily driven by the sleep schedules of its global user base. To observe the true decay of interest, the researchers had to "decycle" the data. They created a "redistributed time scale" ($t^*$) that essentially stretches and compresses time based on the average traffic volume of the Main Page. By removing approximately 16% of "background noise" traffic, they were able to isolate the specific impact of the promotion from the general fluctuations of site-wide traffic.
The Exponential Decay Model
Once the data was decycled, the researchers discovered that popularity follows a simple exponential decay. This was modeled using two primary parameters:
- $\beta$ (Interest Decay): A constant factor representing the rate at which interest fades hour-by-hour.
- $\gamma$ (The "Jump"): A negative jump in views occurring at the 25th hour, when the article transitions from the primary spotlight to the "Recently Featured" list.
Mathematically, this decay is consistent with a population of users acting as individual Poisson processes, where the probability of a "first visit" decreases over time as the pool of potential interested readers is exhausted.
Key Insights and Results
The Predictive Power of the First Hour
One of the most striking findings is that the model can predict the trajectory of views for four days using only the data from the very first hour ($v_1$). From just those first 60 minutes, the model correctly predicted page popularity for 50% of articles at the 25-hour mark. This suggests that the initial "spark" of interest is a highly reliable indicator of long-term engagement in a controlled environment.
Stability of Human Curiosity
The decay factor $\beta$ was found to be remarkably stable across different sets of articles, hovering around 0.987. This stability suggests a near-universal rate at which curiosity fades for Wikipedia readers, regardless of the specific topic being featured.
Popularity and the "Jump"
The researchers found that the "jump" parameter ($\gamma$) is not constant. More popular articles (those with a higher $v_1$) actually experience a sharper relative drop when moved off the Main Page compared to less popular ones. This indicates that the primary spotlight provides a disproportionately large boost to high-interest topics that cannot be sustained once the article is relegated to the "Recently Featured" list.
Limitations and Future Work
While the model is highly effective for Wikipedia, it has specific limitations. It assumes a "single-visit" behavior, which may not hold for complex or breaking-news topics where users might return multiple times. Additionally, the model is designed for internal drivers; it struggles to account for external "spikes" caused by social media mentions or real-world news events occurring mid-promotion.
Future research could explore whether this "decycling" method reveals similar exponential decays in other steady-state environments, such as the "Top Stories" section of online newspapers or mobile app store recommendations, further validating the Poisson-based approach to human attention dynamics.
Summary updated