Authors: Susanne Schröder-Bergen*, University Erlangen-Nürnberg, Germany, Georg Glasze, University Erlangen-Nürnberg, Germany, Finn Dammann, University Erlangen-Nürnberg, Germany, Boris Michel, University Halle, Germany
Topics: Cultural Geography, Cartography, Development
Keywords: Digital Geographies, OpenStreetMap, Artificial Intelligence, Machine Learning, Volunteered Geographic Information, Critical GeoAI, Participation, Local Knowledge, Cultural Geography, Political Geography, Postcolonial Studies, Humanitarian Activism, Digitization
Session Type: Virtual Paper
Start / End Time: 11:10 AM / 12:25 PM
Room: Virtual 47
Presentation File: No File Uploaded
OpenStreetMap (OSM) is considered the most successful Web 2.0 geo-data project. Thousands of volunteers are building up a global geo-database, which is used for numerous applications. OSM has therefore been welcomed as a building block for the opening of geoinformation. However, critical research has shown that OSM is also characterised by numerous inequalities: Last but not least, the data density between the Global North and the Global South still differs significantly.
Recently, large western technology companies as well as humanitarian organisations have also become active in OSM. They are now beginning to introduce AI applications in OSM by automatically deriving cartographic classifications from remote sensing data. These classifications should then be validated by the crowd and thus the AI application is trained.
We highlight fundamental areas of conflict of this AI-application in OSM: On the one hand, the use of AI can improve the density and timeliness of OSM data, especially in those regions of the world that have been underrepresented so far. On the other hand, the question arises as to what kind of geographical knowledge is emphasised - not least: to what extent are machine learning models from the global North, and thus also certain worldviews, transferred? To what extent does a "cartographic land grab" associated with a marginalisation of local knowledge repeat itself? Basically, the question arises as to the relationship between local mapping communities and these AI applications: Is there a strengthening of local mappers or are local mappers rather used as inexpensive click-workers?