Daily Arxiv

This is a page that curates AI-related papers published worldwide.
All content here is summarized using Google Gemini and operated on a non-profit basis.
Copyright for each paper belongs to the authors and their institutions; please make sure to credit the source when sharing.

Modeling Discrimination with Causal Abstraction

Created by
  • Haebom

Author

Milan Moss e, Kara Schechtman, Frederick Eberhardt, Thomas Icard

Outline

This paper challenges the conventional view that race itself is the cause of discrimination. It points out that race is closely intertwined with other social factors, making it difficult to identify its independent causal role. It questions how causal relationships can be explained if race is a socially constructed concept. This paper proposes a new framework that views race as an abstraction of lower-level characteristics. This framework allows race to be modeled as a direct cause of discrimination, ensuring modularity by explicitly stating assumptions about social construction through alignment between race and its constituent elements. These assumptions face normative and empirical challenges, leading to diverse perspectives on the timing of discrimination. By distinguishing between constitutive and causal relationships, this paper clarifies inconsistencies in existing research on discrimination modeling and provides a precise causal explanation.

Takeaways, Limitations

Takeaways: By presenting a new analytical framework to clearly explain the causal relationship of racial discrimination, this study overcomes the limitations of existing research and opens up the possibility of incorporating diverse perspectives. By considering the complex relationship between race and social factors, it enables a more sophisticated and balanced analysis of discrimination. By distinguishing between constitutive and causal relationships, it clarifies the discussion and establishes a foundation for systematically comparing and analyzing diverse perspectives on discrimination.
Limitations: Further validation is needed to determine the practical applicability and utility of the proposed abstract framework. There is a lack of specific methodological guidance on how to establish and measure alignment between race and its constituent elements. Accurately isolating the causal role of race while fully considering all social factors may still be challenging. Further discussion is needed on the criteria for normative judgment and securing objectivity within the framework.
👍