Improving Fairness and Privacy in Selection Problems

Authors: Mohammad Mahdi Khalili, Xueru Zhang, Mahed Abroshan, Somayeh Sojoudi8092-8100

AAAI 2021 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Moreover, the experiments on real-world datasets show that the exponential mechanism can improve both privacy and fairness, with a slight decrease in accuracy compared to the model without post-processing.
Researcher Affiliation Academia Mohammad Mahdi Khalili,1 Xueru Zhang, 2 Mahed Abroshan, 3 Somayeh Sojoudi 4 1 CIS Department, University of Delaware, Newark, DE, USA 2 EECS Department, University of Michigan, Ann Arbor, MI, USA 3 Alan Turing Institute, London, UK 4 EECS Department, University of California, Berkeley, CA, USA khalili@udel.edu, xueru@umich.edu, mabroshan@turing.ac.uk, sojoudi@berkeley.edu
Pseudocode No No explicit pseudocode blocks or algorithm listings were found in the paper.
Open Source Code No The paper does not state that the source code for the methodology is openly available or provide a link to a repository.
Open Datasets Yes Case study 2: FICO score. We conduct two experiments using FICO credit score dataset.4 FICO scores are widely used in the United States to predict how likely an applicant is to pay back a loan. The FICO credit score dataset has been processed by Hardt et al. (Hardt, Price, and Srebro 2016) to generate CDF and non-default rate (i.e., Pr(Y = 1|R = ρ)) for different social groups (Asian, White, Hispanic, and Black). 4Find the dataset here: https://bit.ly/3di5NOC
Dataset Splits No The paper mentions using synthetic data and the FICO dataset, but it does not specify explicit train/validation/test splits for these datasets for model training or evaluation of the post-processing method.
Hardware Specification No No specific hardware details (e.g., GPU models, CPU types, memory amounts) used for running experiments are provided in the paper.
Software Dependencies No The paper does not provide specific version numbers for any ancillary software or libraries used in the experiments.
Experiment Setup No The paper describes parameters for synthetic data generation and the FICO dataset's use but does not provide specific hyperparameters or system-level training settings for a supervised learning model, as its focus is on a post-processing step for an already trained model.