Core Concepts
Urgent need to address spatial fairness in decision-making systems due to location biases.
Abstract
Abstract
Urges consideration of spatial fairness due to correlation with protected characteristics.
Introduction
Data-driven decision-making systems prevalent but can replicate historical biases.
Problems unique to spatial data
Dimensionality, computing spatial network distance, continuity of space, MAUP, spatial autocorrelation.
Limitations of current spatial fairness work
Legal soundness, limitations in techniques like closing the loop and taking agency away from people.
Guidelines and Future Directions
Close the loop, avoid disparate impact, consider location as a protected attribute, be resistant to MAUP.
Conclusion
Argues for importance of addressing spatial fairness and outlines guidelines for future research.
Stats
"Despite location being increasingly used in decision-making systems employed in many sensitive domains such as mortgages and insurance."
"The adoption of data-driven decision-making systems has skyrocketed across the board in the last two decades."
"For example, neighborhoods in the United States have been historically correlated with race."
Quotes
"Everything is related to everything else, but near things are more related than distant things." - Waldo Tobler