From Parity to Preference-based Notions of Fairness in Classification
Authors: Muhammad Bilal Zafar, Isabel Valera, Manuel Rodriguez, Krishna Gummadi, Adrian Weller
NeurIPS 2017 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Finally, we experiment with a variety of synthetic and real-world datasets and show that preference-based fairness allows for greater decision accuracy than parity-based fairness. |
| Researcher Affiliation | Academia | Muhammad Bilal Zafar MPI-SWS mzafar@mpi-sws.org Isabel Valera MPI-IS isabel.valera@tue.mpg.de Manuel Gomez Rodriguez MPI-SWS manuelgr@mpi-sws.org Krishna P. Gummadi MPI-SWS gummadi@mpi-sws.org Adrian Weller University of Cambridge & Alan Turing Institute aw665@cam.ac.uk |
| Pseudocode | No | The paper does not contain structured pseudocode or algorithm blocks (i.e., no clearly labeled algorithm sections or code-like formatted procedures). |
| Open Source Code | Yes | An open-source code implementation of our scheme is available at: http://fate-computing.mpi-sws.org/ |
| Open Datasets | Yes | We experiment with three real-world datasets: the COMPAS recidivism prediction dataset compiled by Pro Publica [23], the Adult income dataset from UCI machine learning repository [2], and the New York Police Department (NYPD) Stop-questionand-frisk (SQF) dataset made publicly available by NYPD [1]. |
| Dataset Splits | No | We randomly split the corresponding dataset into 70%-30% train-test folds 5 times, and report the average accuracy and group beneļ¬ts in the test folds. No explicit mention of a separate validation split. |
| Hardware Specification | No | The paper does not provide specific hardware details (e.g., exact GPU/CPU models, processor types, or memory amounts) used for running its experiments. |
| Software Dependencies | No | The paper discusses the use of logistic regression and SVM classifiers but does not provide specific software dependencies (e.g., library or solver names with version numbers) needed to replicate the experiment. |
| Experiment Setup | No | The paper mentions using logistic regression classifiers with L2-norm regularization and describes the train-test split, but specific hyperparameter values (e.g., learning rate, regularization strength) are not provided in the main text, with regularization details deferred to an appendix. |