Population Research Discovery Seminars

Deep Fake in Geography? When Geospatial Data Encounter Artificial Intelligence

REGISTER for Zoom Seminar HERE

The recent convergence of Artificial Intelligence (AI) and GIScience has been hailed for its transformational capacity in intelligentizing geospatial technologies, and such capacity could also unintendedly or purposefully generate problematic geospatial data. The more recent emergence of “Deep Fake” - an especially controversial use of AI, although not widely applied in GIScience yet, has stimulated widespread attention to its severe social and political impacts. To explain how geospatial data can be affected by AI and to discuss the concomitant implications, this paper focuses on three geospatial spoofing cases: first, the game player trajectories generated by bot; second, the fake locational information spread out on twitter by its filtering algorithm; and third, simulated satellite imagery made by a deep learning algorithm. Confronted with the above-mentioned complex issues, we further propose geospatial inconsistency detection principles to understand their inherent mechanisms and hopefully inform their impacts on society better. In this paper, we encourage GIScientists, geospatial data users, and the public to adjust the pure optimism about AI’s merits or pessimism of its drawbacks, thereby recognize and understand AI’s complex implications on individuals and human societies.

Bo Zhao is an assistant professor at UW's geography department. He is specialized in GIScience, Web Mapping, and VGI (Volunteered Geographic Information). His recent research mainly focuses on a) he geovisualization and geo-narrative of the coupled human and natural system (e.g., climate change, water use, coastal hazards, etc; b) The social implications of mapping, geovisualization and/or other forms of geospatial technologies; and c) The study of different spoofing phenomena in geography towards a post-truth reflection on the value of geospatial technologies