The Danger of Location-Based Discrimination in Artificial Intelligence Models
USC researchers have revealed that artificial intelligence decision-making systems can treat people unfairly based on where they live due to geographic data.
USC researchers have announced that artificial intelligence-based decision mechanisms can unintentionally produce discrimination through geographic location—even without considering protected characteristics such as race and ethnicity—leading to financial disadvantages.
The Hidden Role of Location in Decision Processes
Artificial intelligence systems can disadvantage individuals by focusing on where they live rather than who they are. Even individuals with good careers and high credit scores can face higher insurance premiums or unfavorable loan rates simply because of their postal codes.
The Consequences Uncovered by Spatial Biases
Nripsuta Ani Saxena, leader of the research team, states that people unjustly pay higher insurance premiums due to spatial biases and experience a loss of access to certain services that would make their lives easier.
Reflection of Historical Inequalities in Data
Because residential patterns reflect decades of housing discrimination and unequal investments, artificial intelligence systems can learn these historical patterns within geographic data. This lays the groundwork for the algorithm to base its decisions on this past data.
Search for Solutions for Fairer Systems
It is stated that location information cannot be completely eliminated because it provides necessary information for insurance companies and mortgage lenders. The primary challenge is to prevent algorithms from turning geography into a proxy for undesirable attributes.