Transparency and Trust in Collaborative Mapping: Concerns and Dilemmas in AI-Assisted Road Integration within OpenStreetMap
The influx of machine-generated data from geospatial artificial intelligence (AI), has grown significantly in less than a decade. AI-assisted mapping, where human validation refines machine-generated output, is increasingly used to update crowdsourced databases, such as OpenStreetMap (OSM). OSM cont...
Saved in:
| Main Authors: | , , , , , |
|---|---|
| Format: | Article (Journal) |
| Language: | English |
| Published: |
08 Dec 2025
|
| In: |
Annals of the American Association of Geographers
Year: 2025, Pages: 1-22 |
| ISSN: | 2469-4460 |
| DOI: | 10.1080/24694452.2025.2589286 |
| Online Access: | Resolving-System, kostenfrei, Volltext: https://doi.org/10.1080/24694452.2025.2589286 Verlag, kostenfrei, Volltext: https://www.tandfonline.com/doi/full/10.1080/24694452.2025.2589286 
|
| Author Notes: | Francis Andorful, Benjamin Herfort, Edson Augusto Melanda, Nathan Damas Antonio, Alexander Zipf, and Silvana Philippi Camboim |
| Summary: | The influx of machine-generated data from geospatial artificial intelligence (AI), has grown significantly in less than a decade. AI-assisted mapping, where human validation refines machine-generated output, is increasingly used to update crowdsourced databases, such as OpenStreetMap (OSM). OSM contributors, however, have expressed mixed sentiments about the presence of AI content in the database, raising questions about trust and authenticity. We first analyzed community discussions identifying emotional concerns and the evolving role of human mappers. Next, we assessed whether AI-assisted roads (AI-aR) can be reliably detected within OSM, using machine learning (ML) models as diagnostic tools to reveal transparency limitations and advocate for improved tagging practices. Community debates reveal tensions over inadequate tagging, local context loss, and corporate influence. ML models perform best on human-free benchmark data, worse on AI-aR edits, but improve when temporal patterns are included. Although geometric or temporal patterns might help identify AI-aR, these approaches remain uncertain and unstable over time. Our findings underscore the danger of data quality erosion through issues like validation-loop bias and the accountability sink effect. The distinction between AI-aR and human-generated roads will continue to blur with of the growth of human-AI collaboration. Current models might perform best on fresh AI contributions with minimal human modification. |
|---|---|
| Item Description: | Gesehen am 05.01.2026 |
| Physical Description: | Online Resource |
| ISSN: | 2469-4460 |
| DOI: | 10.1080/24694452.2025.2589286 |