Authors: Renee Sieber*, McGill University, Jeremy Crampton*, Newcastle University
Topics: Geographic Information Science and Systems, Quantitative Methods
Session Type: Virtual Paper
Start / End Time: 11:10 AM / 12:25 PM
Room: Virtual 47
Presentation File: No File Uploaded
With few exceptions and despite a long history of using machine learning (ML), geographers' voices have been largely absent in the critique of artificial intelligence (AI)/ML, especially where it concerns deep learning. This absence exists despite the prevalence of geographic examples and concepts to illustrate societal impacts of AI, and critiques from computer science of popular deep learning methods used in geography. We assert that geographers can play a key role in developing a more geographically responsible AI, a critical GeoAI.
In the rush to ‘disrupt’ via GeoAI there is a danger of neglecting lessons learned from critical GIS and cartography. Data, algorithms, and representations are power claims about the world. A critical GeoAI might, for example, leverage lessons through a commitment to local and bottom-up designs that include traditional-looking concepts of co-creation and use of place-based training data. Critical GeoAI can also be examined from a scalar and a policy perspective. AI systems developed in and trained on data from one locale can become universalized to transfer to another locale, especially if the target destination is data “poor”. There may be little critique of the sensitivity of algorithms to spatial feature engineering. Who holds jurisdictional power for accountability in Responsible AI is uncertain.
We articulate an area of study called Critical GeoAI that analyzes AI through a spatial lens, either by leveraging the distinct perspective of geographers or by examining explicitly spatial AI tools. What are our disciplinary strengths and our “value proposition”?