Source-linked AI summary
Circadian patterns of Wikipedia editorial activity: A demographic analysis
Taha Yasseri, Róbert Sumi, János Kertész
TL;DR
Privacy restrictions leave basic information about Wikipedia editors rarely studied systematically. The paper analyzes temporal activity across 34 Wikipedias and uses time-shifted activity profiles to estimate geographic contributions; it reports a common daily pattern alongside cross-language differences and estimates substantial European contributions to English and simple English Wikipedias.
Problem
Privacy restrictions prevent access to committed editors’ personal data, leaving basic characteristics of Wikipedia’s editor community rarely studied systematically and without bias.
Source: Introduction · p. 2Method
The authors compare timestamped activity patterns across Wikipedias and fit profiles as mixtures of shifted Standard curves to estimate contributions from different time zones.
Source: Geographical distribution of editors · p. 6Results
The study reports a common circadian pattern across Wikipedias and estimates considerable European contributions to English and simple English, with North America below expectations.
Source: Discussion · p. 10Takeaways & Limitations
Estimated editor distributions provide a basis for investigating article biases, heterogeneous topical coverage, and the origins of conflicts and editorial wars.
Source: Discussion · p. 7Timestamp-only estimates cannot determine editor distributions across longitudes, and regional resolution cannot exceed one-hour bands.
Source: Discussion · p. 9
Abstract
from arXiv · showhide
Wikipedia (WP) as a collaborative, dynamical system of humans is an appropriate subject of social studies. Each single action of the members of this society, i.e. editors, is well recorded and accessible. Using the cumulative data of 34 Wikipedias in different languages, we try to characterize and find the universalities and differences in temporal activity patterns of editors. Based on this data, we estimate the geographical distribution of editors for each WP in the globe. Furthermore we also clarify the differences among different groups of WPs, which originate in the variance of cultural and social features of the communities of editors.
Introduction
The paper addresses the lack of systematic, unbiased information about Wikipedia editors, whose identities are protected by privacy restrictions. It studies daily and weekly activity across 34 Wikipedias and uses temporal patterns to estimate editors’ geographic distribution.
Introduction
- Privacy restrictions have left basic information about Wikipedia editors, including nationality and individual editing patterns, rarely studied systematically and without bias.The authors identify privacy as the main barrier to obtaining personal data from committed editors.Source: Introduction · p. 2
- Because registered editors’ identities and IP addresses are hidden, the paper seeks to infer their geographic distribution from their temporal editing patterns.IP-based estimates for unregistered editors cover less than 10% of edits in many Wikipedias and may not represent average users.Source: Introduction · p. 2
- The study characterizes circadian activity across 34 Wikipedias, estimates editors’ geographic distribution, and compares longer-term weekly patterns across societies.The authors focus on cyclic behavior; other forms of temporal inhomogeneity are reported elsewhere.Source: Introduction · pp. 2–3
Methods
The analysis uses large Wikipedias and timestamps of individual edits to characterize activity over the day and week. Initial local-time offsets are assigned by language, while the authors note that editor locations are unavailable for improving those assumptions at this stage.
Methods
- The sample contains 34 large Wikipedias, generally selected for having more than 100,000 articles, with 3 million to 455 million edits and 46,000 to 14 million editors each.Volapük and Waray-Waray were excluded; simple English was included despite having around 70,000 articles.Source: Methods · p. 3
- The authors use timestamps for every edit to calculate activity by time of day and day of week, initially assigning each language a local-time offset.For widely spoken languages they use the offset of a known language origin; English and simple English initially use US Central Time.Source: Methods · p. 3
- Because registered editors’ IP addresses are unavailable, the initial language-based time offsets cannot be improved at this stage.The authors aim to estimate contributions from different geographic units using average editor behavior.Source: Methods · p. 4
Results
Across the studied Wikipedias, daily activity generally reaches a dawn minimum and rises through the day, while deviations in timing and amplitude vary across language communities. The authors also estimate geographic contributions from temporal profiles and identify distinct weekly activity patterns.
Circadian patterns
- Most Wikipedias show a daily activity minimum around 6 A.M., a rapid rise to noon, a slight increase until around 9 P.M., and a nighttime decline.The authors describe these shapes as qualitatively similar to activity patterns in other human systems.Source: Circadian patterns · p. 4
Deviations
- Spanish and Portuguese activity shifts later and has flatter amplitude, while English and simple English strongly deviate from the common pattern, indicating complex origins.The paper suggests the larger Latin American speaker population contributes to the Spanish and Portuguese shift and flatter profile; cultural features may also contribute.Source: Deviations · pp. 4–5
- The four activity-profile categories distinguish localized European languages, Asian languages with more night activity, and widely distributed contributor groups with higher night activity.The last category includes Arabic, Persian, Chinese, Spanish, Portuguese, English, and simple English.Source: Deviations · p. 5
- Sleep depth ranges from 2.3 for simple English to 5.6 for Italian; Spanish and Portuguese have intermediate values of 4.4 and 4.2, respectively.The paper treats sleep depth, the difference between maximum and minimum activity, as a possible measure of contributor locality.Source: Deviations · p. 5
Geographical distribution of editors
- A mixture of 3 to 6 shifted Standard curves often fits activity profiles, and demographic constraints support estimates of regional contributions for 9 Wikipedias.The Standard curve is averaged from the 10 Wikipedias with the deepest sleep; weights represent contribution volume from time zones.Source: Geographical distribution of editors · p. 6
Weekly patterns
- Weekly patterns classify 28 of 34 Wikipedias into four groups: some peak on working days, while others show weekend activity, including Friday activity in Arabic and Persian.The authors associate Friday activity in Arabic and Persian with Friday being a public holiday in their countries of origin.Source: Weekly patterns · p. 7
Discussion
The discussion highlights how estimated editor geography can inform questions about Wikipedia content and participation, while noting limits in the method’s spatial resolution. It also relates temporal patterns to working hours and reports that seasonal analysis yielded few clear findings.
Discussion
- Editor geography may help investigate article bias, uneven topical coverage, and the origins of conflicts, though the English Wikipedia’s estimated North American share remains a puzzle.The discussion also reports an even larger European share for Simple English than for English.Source: Discussion · p. 7
- IP-based editor locations and activity patterns broadly agree where unregistered edits are more common, but diverge for Wikipedias with few unregistered edits.Unregistered editors’ activity has a deeper nighttime minimum; the authors suggest these editors may better represent occasional readers than committed contributors. Open proxies can further weaken IP-based estimates.Source: Discussion · p. 8
- Persian Wikipedia’s small sleep depth and estimated 45% Iranian edit share may reflect internet restrictions in Iran and intellectual communities formed by emigrants.The authors present these as possible explanations, not established causes.Source: Discussion · p. 8
- Low contributions from North American editors to French Wikipedia and Egyptian editors to Arabic Wikipedia may relate to divergence between spoken dialects and standard languages.The authors note a separate Egyptian Arabic Wikipedia and suggest contribution estimates may inform decisions about local-dialect editions.Source: Discussion · pp. 8–9
- Geographic estimates cannot resolve longitude from timestamps alone, and regional distinctions finer than roughly one-hour time zones are not justified.The authors note that summer-time practices and insensitivity of fit to nearby time-zone weights constrain finer estimates.Source: Discussion · p. 9
- Asian-language Wikipedias show higher evening and night activity and more weekend activity, which the authors relate partly to working hours exceeding 40 hours per week in the cited countries.The discussion also reports later-evening editing in European countries with longer working times; seasonal analysis was difficult to interpret, with reduced activity approaching New Year holidays the only significant large-scale pattern.Source: Discussion · p. 9
- Diurnal activity is broadly universal across Wikipedias, but weekly patterns differ, with working-day and weekend-active groups.Geographical estimates place a considerably large part of English and Simple English edits in Europe, while the North American share of English edits is below expectations.Source: Discussion · p. 10
Tables
The figures show editor activity patterns across Wikipedias, including daily and weekly categories, a decomposition of English activity, and estimated regional contributions. Table 1 reports Wikipedia statistics and assigned time offsets and sleep depths.
Tables
- Figure 1 compares normalized daily activity across Wikipedia groups, including Spanish and Portuguese, English and simple English, and an average and standard curve.Its panels also show unregistered-user activity alongside the average and standard curves.Source: Figure 1 · p. 13
- Figure 2 shows four categories of deviation from the average activity pattern, with sleep depths of 4.5±0.2, 4.3±0.2, 4.9±0.1, and 2.8±0.2 for categories a–d, respectively.The gray dotted line gives each category’s average deviation.Source: Figure 2 · p. 14
- Figure 3 decomposes English Wikipedia activity into five shifted, weighted standard curves whose linear superposition models the empirical curve.The empirical data are blue, the standard curve is yellow, and the superposition is red.Source: Figure 3 · p. 15
- Figure 4 estimates each region’s share of contributions to each Wikipedia by combining optimization outputs with language demographic data.Some nearby regions are unified when reporting results.Source: Figure 4 · p. 16
- Figure 5 categorizes weekly editor activity into four subcategories, with (b)–(d) representing weekend activity and (a) weekday activity.The caption notes that category (d) includes languages spoken mostly in Muslim countries with different weekend schedules.Source: Figure 5 · p. 17
- Table 1 reports Wikipedia size and activity statistics, user counts, unregistered-edit share, assigned UTC offsets, and sleep depths; its demographic data are not analyzed in the paper.The table’s demographic information is intended to give an impression of the Wikipedias.Source: Table 1 · p. 18, Tables · p. 18