Evaluating Bias in GPS-Measured Activities Using Computer Vision and Distributed Travel Routing
Topics: Quantitative Methods
, Remote Sensing
, Urban Geography
Keywords: big data, bias, gps, computer vision
Session Type: Virtual Paper Abstract
Day: Tuesday
Session Start / End Time: 3/1/2022 11:20 AM (Eastern Time (US & Canada)) - 3/1/2022 12:40 PM (Eastern Time (US & Canada))
Room: Virtual 26
Authors:
James Saxon, University of Chicago
,
,
,
,
,
,
,
,
,
Abstract
The last several years have seen explosive growth in the use of commercial GPS data from location-based services, with widespread applications in both industry and academia. Policymakers also relied on these data through the coronavirus pandemic, for insights about mobility and congested spaces.
However, work to directly evaluate fundamental data quality, sample rates, bias, and appropriate applications, has been far more limited. This paper suggests and implements technical and statistical strategies for directly measuring variation in device to person “sample rates" from location-based services. The technical strategies employed are, first, to build counts of park entrants from video streams, and second, to construct vehicle flows from GPS locations, which are in turn contrasted with administrative data. The subsequent analysis shows that sample rates in fact differ significantly by activity and demographic group. This paper underscores the care that must be taken in relying on these data and constructing observables from them, for the aforementioned applications.
It suggests the utility of incorporating complementary data streams. As the privacy practices of GPS data continue to evolve, new streams are likely to prove ever more necessary.
Evaluating Bias in GPS-Measured Activities Using Computer Vision and Distributed Travel Routing
Category
Virtual Paper Abstract
Description
This abstract is part of a session. Click here to view the session.
| Slides