Geographic Biases are ‘Born, not Made’: Exploring Contributors’ Spatiotemporal Behavior in OpenStreetMap

Citation: Jacob Thebault-Spieker, Brent Hecht, Loren Terveen (January 7–10, 2018) Geographic Biases are ‘Born, not Made’: Exploring Contributors’ Spatiotemporal Behavior in OpenStreetMap. ACM Conference on Supporting Group Work 2018 (RSS)
DOI (original publisher): https://doi.org/10.1145/3148330.3148350
Semantic Scholar (metadata): https://doi.org/10.1145/3148330.3148350
Sci-Hub (fulltext): https://doi.org/10.1145/3148330.3148350
Internet Archive Scholar (search for fulltext): Geographic Biases are ‘Born, not Made’: Exploring Contributors’ Spatiotemporal Behavior in OpenStreetMap
Tagged: Sociology (RSS) peer production (RSS), organizational behavior (RSS), openstreetmap (RSS), collaboration (RSS), bias (RSS)

Summary

This paper examines production biases in the Open Street Map project. Open Street Map is a mapping platform that allows individuals to contribute geographical information (such as features, roads, and buildings) to a shared repository. This geographical information is then incorporated into a variety of FLOSS and commercial products. OpenStreetMap's coverage is understood to be geographically biased, such that wealthier urban areas have substantially more detailed information available to them than poorer and more rural areas. The authors examine OpenStreetMap contributor behavior through a what they term a "spatiotemporal" lens, that is, how geographically-identified contributor behaviors change over time.

The authors found that the behaviors of contributors to this project echoes a dynamic which has been reported to occur Wikipedia: the "users are born, not made" dynamic described by Panciera, Halfaker, and Terveen in 2009 (DOI: 10.1145/1531674.1531682). This shorthand refers to the notion that long-term contributors to a peer production platform can be immediately predicted and distinguished from those contributors who will not persist in their engagement with the site. This suggests that individuals join a platform already possessing traits or tendencies which suggest either longer-term or shorter-term engagement.

These "born not made" contributors, who are characterized by their origin from counties which tend to be wealthier and urban, tend to contribute to the project in areas which are close to their areas of origin. The authors find that the geographic bias of Open Street Maps is thus also "born not made" -- persistent contributors tend to be wealthy/urban, and since they contribute inside a constrained area, their contributions are likewise liable to be geographically biased. The paper also considers the relatively small subset of contributors who seem to deviate from the trend and seeks explanations for these deviations. For example, some of the prolific individuals who do contribute to rural/high-poverty areas seem to be orienting their contribution to national parks and forests, or else seem to indicate a regional focus. However, some of the prolific individuals contribute in high-poverty areas not consistent with a "parks" explanation. The authors suggest that this finding should be further investigated and may inform producer recruitment efforts.

Theoretical and Practical Relevance

Scope, Data, and Method:

This paper's scope is limited to the continental US. The authors drew upon data from the US Census, the full history of OpenStreetMap through February 2014. Following the example of Panciera, Halfaker, and Terveen (2009), the paper classes contributors, which divides contributors into the top 1% of all contributors (by edit quantity), the next 9% of all contributors, and the bottom 90% of all contributors. Geographic distances are characterized by relative distances from a geometric center and by identifying a county wherein they make the most contributions. Human geographic factors are characterized using census information about how rural a county is and what percent of county residents live in poverty.

The paper develops evidence for its conclusions by producing descriptive statistics of relationships among these data, including a histogram of how long contributors persist, and distribution graphs displaying the relative geographical spread of different classes of contributors.

Contributions:

This paper contributes to the understanding of how geographical biases may arise in peer production projects. The work also offers a model to follow in terms of using descriptive statistics to characterize spatiotemporal peer production.