Fairness Shields: Safeguarding against Biased Decision Makers

Authors: Filip Cano, Thomas A. Henzinger, Bettina Könighofer, Konstantin Kueffner, Kaushik Mallik

AAAI 2025 | Venue PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Our empirical evaluation demonstrates the effectiveness of these shields in ensuring fairness while maintaining cost efficiency across various scenarios.
Researcher Affiliation Academia 1Graz University of Technology 2Institute of Science and Technology Austria (ISTA) 3IMDEA Software Institute EMAIL, EMAIL, EMAIL, EMAIL
Pseudocode No The paper describes algorithms in prose and mathematical equations but does not include a distinct pseudocode block or algorithm figure.
Open Source Code No The paper refers to an extended version on arXiv (Cano et al. 2024b) but does not provide an explicit statement or link for the source code of their methodology.
Open Datasets Yes We performed our experiments on the datasets Adult (Becker et al. 1996), COMPAS (Kirchner et al. 2016), German Credit (Hofmann 1994), and Bank Marketing (Moro et al. 2012).
Dataset Splits No The paper mentions using well-known datasets but does not explicitly provide details about training/test/validation dataset splits, percentages, or specific splitting methodologies used for reproduction.
Hardware Specification No The paper does not provide specific hardware details (e.g., exact GPU/CPU models, processor types with speeds, memory amounts, or detailed computer specifications) used for running its experiments.
Software Dependencies No The paper does not provide specific ancillary software details, such as library or solver names with version numbers, needed to replicate the experiment.
Experiment Setup Yes We synthesized shields to ensure DP and Eq Opp with thresholds κ {0.05, 0.1, 0.15, 0.2}. For all models and datasets, Fin Hzn shields were synthesized with T = 100 for DP and T = 75 for Eq Opp. We synthesized Static-Fair, Static-BW, and Dynamic shields with T = 50 for DP and Eq Opp, and simulated them for 10 periods.