Preventing Fairness Gerrymandering: Auditing and Learning for Subgroup Fairness
Authors: Michael Kearns, Seth Neel, Aaron Roth, Zhiwei Steven Wu
ICML 2018 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | 5. Experimental Evaluation We now describe an experimental evaluation of our proposed algorithmic framework on a dataset in which fairness is a concern, due to the preponderance of racial and other sensitive features. We also demonstrate that for this dataset, our methods are empirically necessary to avoid fairness gerrymandering. |
| Researcher Affiliation | Collaboration | 1University of Pennsylvania, Philadelphia, PA, USA 2Mcrosoft Research, New York City, NY, USA. |
| Pseudocode | No | Pseudocode for the implemented algorithm is given in the full version. |
| Open Source Code | No | The paper does not explicitly state that the code for their methodology is released, nor does it provide a link. |
| Open Datasets | Yes | The dataset we use is known as the Communities and Crime (C&C) dataset, available at the UC Irvine Data Repository.1 1http://archive.ics.uci.edu/ml/datasets/ Communities+and+Crime |
| Dataset Splits | No | Thus for simplicity, we report all results here on the full C&C dataset of 1994 points, treating it as the true distribution of interest. |
| Hardware Specification | No | The paper mentions implementing algorithms and running experiments but does not provide any specific hardware details like GPU/CPU models or memory used for these experiments. |
| Software Dependencies | No | The paper describes algorithm implementation but does not specify any software dependencies with version numbers (e.g., specific machine learning libraries or frameworks). |
| Experiment Setup | Yes | 5.3. Algorithm Implementation The main details in the implementation of Fair Fict Play are the identification of the model classes for Learner and Auditor, the implementation of the cost sensitive classification oracle and auditing oracle, and the identification of the protected features for Auditor. For our experiments, at each round Learner chooses a linear threshold function over all 122 features. ... Auditor s model class consists of all linear threshold functions over just the 18 aforementioned protected race-based attributes. |