Notice: The reproducibility variables underlying each score are classified using an automated LLM-based pipeline, validated against a manually labeled dataset. LLM-based classification introduces uncertainty and potential bias; scores should be interpreted as estimates. Full accuracy metrics and methodology are described in [1].
Understanding Fairness Surrogate Functions in Algorithmic Fairness
Authors: Wei Yao, Zhanke Zhou, Zhicong Li, Bo Han, Yong Liu
TMLR 2024 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | In this work, in order to deeply understand them, taking a widely used fairness definition demographic parity as an example, we show that there is a surrogate-fairness gap between the fairness definition and the fairness surrogate function. Also, the theoretical analysis and experimental results about the gap motivate us that the fairness and stability will be affected by the points far from the decision boundary, which is the large margin points issue investigated in this paper. [...] Finally, we provide empirical evidence showing that our methods consistently improve fairness and stability while maintaining accuracy comparable to the baselines in three real-world datasets. |
| Researcher Affiliation | Academia | Wei Yao EMAIL Gaoling School of Artificial Intelligence Renmin University of China, Beijing Beijing Key Laboratory of Big Data Management and Analysis Methods, Beijing Zhanke Zhou EMAIL TMLR Group, Department of Computer Science Hong Kong Baptist University Zhicong Li EMAIL Gaoling School of Artificial Intelligence Renmin University of China, Beijing Beijing Key Laboratory of Big Data Management and Analysis Methods, Beijing Bo Han EMAIL TMLR Group, Department of Computer Science Hong Kong Baptist University Yong Liu EMAIL Gaoling School of Artificial Intelligence Renmin University of China, Beijing Beijing Key Laboratory of Big Data Management and Analysis Methods, Beijing |
| Pseudocode | Yes | Our balanced surrogates approach mitigates unfairness by treating different sensitive groups differently using a parameter being updated during training. The key idea of the updating procedure is making the magnitude of gap as small as possible. [...] The algorithmic representation of the balanced surrogates can be found in Appendix B.1. [Appendix B.1: The Balanced Surrogates Algorithm, Algorithm 1: Balanced Surrogates] |
| Open Source Code | No | The paper does not contain any explicit statements about code release, links to a code repository, or mention of code in supplementary materials. |
| Open Datasets | Yes | We use three real-world datasets: Adult (Kohavi, 1996), Bank Marketing (S. Moro & Rita, 2014) and COMPAS (Julia Angwin & Kirchner, 2016), which are commonly used in fair machine learning (Mehrabi et al., 2021). |
| Dataset Splits | Yes | The dataset is randomly divided into training set (70%), validation set (5%) and test set (25%). |
| Hardware Specification | No | The paper does not provide specific hardware details such as exact GPU/CPU models, processor types, or memory amounts used for running 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 | The parameter setting is discribed in Appendix B.3. [Appendix B.3: The Parameters] General sigmoid surrogate. The regularization coefficient λ in (8) lies in the grid [0.1, 0.2, , 5]. The parameter w in general sigmoid is chosen from {1, 2, 4, 8, 16} according to the fairness performance on the validation set. Balanced surrogate. The smoothing factor α = 0.9. The termination threshold η = 0.01. |