Bridging Causality, Individual Fairness, and Adversarial Robustness in the Absence of Structural Causal Model

Ahmad Reza Ehyaei, Golnoosh Farnadi, Samira Samadi

Abstract: Despite the essential need for comprehensive considerations in responsible AI, factors such as robustness, fairness, and causality are often studied in isolation. Adversarial perturbation, used to identify vulnerabilities in models, and individual fairness, aiming for equitable treatment of similar individuals, despite initial differences, both depend on metrics to generate comparable input data instances. Previous attempts to define such joint metrics often lack general assumptions about data and were unable to reflect counterfactual proximity. To address this, our paper introduces a \emph{causal fair metric} formulated based on causal structures encompassing sensitive attributes and protected causal perturbation. To enhance the practicality of our metric, we propose metric learning as a method for metric estimation and deployment in real-world problems in the absence of structural causal models. We also demonstrate the applications of the causal fair metric in classifiers. Empirical evaluation of real-world and synthetic datasets illustrates the effectiveness of our proposed metric in achieving an accurate classifier with fairness, resilience to adversarial perturbations, and a nuanced understanding of causal relationships.

Submission Length: Regular submission (no more than 12 pages of main content)

Changes Since Last Submission:

Thank you for your feedback. In this revision, we have addressed the specific concerns raised in the final decision and internal reviewer discussions:

- **Clarification in the Abstract and Introduction:** We have explicitly outlined the gap in prior work, the impossibility proposition, and our proposed approximation to ensure a clearer motivation for our approach.
- **Section 6:** The discussion has been added to highlight the identified gaps more explicitly and improve the logical flow of our arguments.
- **Proposition 6.3:** We have added the explanation of its scope, addressing concerns about its contextual fit. Additional justifications have been provided to clarify its relationship with prior results.

Code: https://github.com/Ehyaei/Causal-Fair-Metric-Learning

Assigned Action Editor: ~changjian_shui1

Submission Number: 3444