Modeling Discrimination with Causal Abstraction
Milan Mossé, Kara Schechtman, Frederick Eberhardt, Thomas Icard·January 14, 2025
Summary
The paper introduces a framework for reasoning about racial discrimination, treating race as a high-level abstraction. It addresses the challenge of modeling discrimination causally by allowing precise, explicit statements about social construction, ensuring modularity. The abstraction framework clarifies disagreements in the literature, pinpointing where precise causal accounts of discrimination differ. Causal discrimination theory posits that direct racial discrimination occurs when a person is treated worse due to their race. This concept is supported by audit studies, like Bertrand and Mullainathan's (2004) experiment, which found racial bias in the labor market. The theory applies to other protected attributes and is used in algorithmic fairness criteria and legal contexts, such as the U.S. Title VII, which prohibits actions "because of" sex. The causal notion of discrimination is invoked when the outcome would change if the attribute were different, aligning with but-for causation. This approach promises to address limitations in probabilistic notions of discrimination and is computationally feasible, with methods to mitigate causal discrimination under uncertainty.
Introduction
Background
Overview of racial discrimination in society
Importance of understanding discrimination causally
Objective
To introduce a framework for reasoning about racial discrimination using a high-level abstraction approach
Framework Overview
High-Level Abstraction
Explanation of treating race as a high-level abstraction
Benefits of this approach in modeling discrimination causally
Modularity
Importance of modularity in the framework
How it allows for precise, explicit statements about social construction
Causal Discrimination Theory
Definition
Explanation of the concept of causal discrimination
Distinction from other forms of discrimination
Audit Studies
Bertrand and Mullainathan's (2004) experiment as an example
Findings and implications for the labor market
Application
Extension to other protected attributes
Use in algorithmic fairness and legal contexts
Causal Notion of Discrimination
Theoretical Foundation
Alignment with but-for causation
Explanation of the concept in detail
Practical Implications
How the causal notion addresses limitations in probabilistic discrimination
Computational Feasibility
Methods for Mitigation
Overview of techniques to address causal discrimination under uncertainty
Computational approaches to implement the framework
Conclusion
Summary of the Framework
Recap of the key points and contributions
Future Directions
Potential areas for further research and development
Basic info
papers
computers and society
artificial intelligence
Advanced features