Dataset for "Ultraslow dispersion reveals a marginal stability boundary in collective attention"

This dataset was used to produce the figures in the article “Ultraslow growth of uncertainty reveals a stability boundary in collective attention” by Keisuke Okamura (2026). This dataset contains large-scale daily page-view data from the English-language edition of Wikipedia, collected using the REST APIs provided by the Wikimedia Foundation. For each calendar day between July 1, 2015 and December 31, 2020 (denoted by d), the set of highly viewed pages (“Topviews”) on that day was identified, up to the API-imposed maximum of K = 1,000 pages. For each identified page, the subsequent daily page views were collected over a one-year period (T = 365 days), starting from the day following d. Daily page-view counts were retrieved via the Analytics Query Service (AQS) API. Only user traffic was included, following the official Wikimedia definition under which bot-generated activity is excluded. The data collection procedure was repeated independently for each day d within the specified period. As a result, approximately (KT =) 365,000 page titles were collected per year, except for 2015, which covers a half-year period. For each reference day d, the resulting data form a matrix capturing the post-peak page-view trajectories of articles that attracted substantial collective attention on that day. The datasets are provided in STATA .dta format and are organised separately by calendar year.