Webinar: Does GeoAI Promise an Ethical Future for Spatial Analytics?
Date and Time: Tuesday, February 8, 2022 8:00 am - 10:30 am U.S. Pacific Time

Chair(s) of the Webinar and Organizing Committee Member(s): Michael F. Goodchild. Co-chairs include Stewart Fotheringham, Wenwen Li.
Host(s) of the Webinar: Co-organized with ASU Spatial Analysis Research Center (SPARC), School of Geographical Sciences and Urban Planning (SGSUP)

Will there be a recording? Yes

In recent years, Geospatial Artificial Intelligence (GeoAI) has become the focus of a new wave of data-driven spatial analytics. This emerging technology benefits from recent advances in deep machine learning and high-performance computing, and has the capability to process large volumes of geospatial data to support knowledge discovery at scale. In just a few years, we have seen a flourishing of research projects and publications in this area, introducing exciting new techniques ranging from analysis of remotely sensed images to information extraction and natural language processing, and to the spatial and semantic reasoning of large-scale knowledge graphs. These techniques have played an increasing role in understanding the ever-evolving social and environmental systems found in the real world, as well as the interaction between them, such as modeling the spread of infectious diseases, quantifying the melting rate of sea ice in the Arctic, and predicting presidential voting outcomes.

However, because of its data-driven and theory-free nature, GeoAI research possibly generates as much skepticism as enthusiasm. For example, many GeoAI methods involve supervised learning, meaning that training data containing ground-truth information are needed for the algorithm to learn new patterns and make predictions. While training data contain potential bias in terms of sampled locations and socioeconomic factors, the transferability and replicability of a trained model will also be questioned. In addition, GeoAI models can be hard to interpret given their highly complex and non-linear nature, thereby reducing confidence in their solutions. In addition, despite the technical advances, GeoAI still faces challenges in its ability to conduct self-supervised or weakly supervised learning, such that predictive performance can be less dependent on the quality, size, and diversity of the training data. Other challenges including how to improve model efficiency to enable real-time GeoAI for time-critical applications. These issues are worthy of further attention from the geospatial community.

Recently numerous aspects of scientific practice have come under scrutiny, and concerns have been raised about the degree to which they reflect ethical ideals. Many of the questions raised in the early 1990s about GIS—equity in training and access, bias and oversimplification in representation, the potential for surveillance—can also be raised about GeoAI.

Consequently, topics of the workshop include:

• GeoAI for surveillance
• Transparency in GeoAI
• Explainable GeoAI
• Reproducibility and replicability of GeoAI
• Propagation of data uncertainty through GeoAI
• Real-time GeoAI
• Next-generation GeoAI models
• What makes an AI application “GeoAI”?


This webinar is being organized by ASU/SPARC, a multi-year conversation over the ethical issues that arise in geographic research, as the 2022 online edition of the SPARC workshop series (https://sgsup.asu.edu/sparc/workshops), and as one of the GeoEthics series of webinars that was initiated in February 2021.

The webinar will be scheduled for 2.5 hours. The topic will be introduced by the co-chairs, who will explain the structure of the webinar and its objectives, and the importance of follow-on actions and activities. The moderator will then introduce each of the speakers, each of whom will be asked to speak for no more than 20 minutes on some mix of the topics. These initial presentations will be followed by a moderated session of questions from the audience. The webinar will end with a review of follow-on activities by the co-chairs.

During the webinar, we will describe a process for follow-on activities. Participants will be given the opportunity to provide a 2-page position paper describing their interest in the topics of the webinar and in potential subsequent activities, along with a 2-page resume, within two weeks after the webinar. We will use these to create a number of topic areas for further exploration, and schedule a follow-on in-person workshop on ASU campus in October or November 2022 in which we can drill down to greater detail about GeoAI research and how it can be further advanced scientifically, and how it can be leveraged to create significant societal benefits. Participants of the webinar and those who submit a position paper will become potential invitees/speakers for the in-person workshop at ASU. We look forward to using this series of events to keep up the momentum in the community to advance GeoAI research, and to maintain interest in the GeoEthics project. Funds will be provided to cover the travel cost of invited participants. We are looking to diversify the attendees in terms of academic background, gender, race and other aspects. Women and underrepresented minorities will be especially encouraged to apply.

Michael F. Goodchild
University of California, Santa Barbara, Arizona State University

Stewart Fotheringham
Arizona State University

Peter Kedron
Arizona State University

May Yuan
University of Texas at Dallas

Yu Liu
Peking University

Renee E. Sieber
McGill University

Krzysztof Janowicz
University of California, Santa Barbara

Wenwen Li
Arizona State University