Counterfactual Fairness

Authors: Matt J. Kusner, Joshua Loftus, Chris Russell, Ricardo Silva

NeurIPS 2017 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We demonstrate our framework on a real-world problem of fair prediction of success in law school. and We experimentally contrasted our approach with previous fairness approaches and show that our explicit causal models capture these social biases and make clear the implicit trade-off between prediction accuracy and fairness in an unfair world.
Researcher Affiliation Academia Matt Kusner The Alan Turing Institute and University of Warwick mkusner@turing.ac.uk Joshua Loftus New York University loftus@nyu.edu Chris Russell The Alan Turing Institute and University of Surrey crussell@turing.ac.uk Ricardo Silva The Alan Turing Institute and University College London ricardo@stats.ucl.ac.uk
Pseudocode Yes 4.1 Algorithm 1: procedure FAIRLEARNING(D, M) Learned parameters ˆθ 2: For each data point i D, sample m MCMC samples U (i) 1 , . . . , U (i) m PM(U | x(i), a(i)). 3: Let D be the augmented dataset where each point (a(i), x(i), y(i)) in D is replaced with the corresponding m points {(a(i), x(i), y(i), u(i) j )}. 4: ˆθ argminθ P i D l(y(i ), gθ(U (i ), x(i ) A)). 5: end procedure
Open Source Code No The paper does not contain any explicit statements about making the source code publicly available or links to a code repository.
Open Datasets Yes The Law School Admission Council conducted a survey across 163 law schools in the United States [35]. It contains information on 21,790 law students such as their entrance exam scores (LSAT), their grade-point average (GPA) collected prior to law school, and their first year average grade (FYA). and reference '[35] Wightman, Linda F. Lsac national longitudinal bar passage study. lsac research report series. 1998.'
Dataset Splits No The paper states 'We split the dataset 80/20 into a train/test set, preserving label balance, to evaluate the models.' which indicates train and test splits, but no explicit validation split is mentioned.
Hardware Specification No The paper does not provide any specific hardware details such as GPU models, CPU types, or cloud computing specifications used for running the experiments.
Software Dependencies Yes We use the probabilistic programming language Stan [34] to learn K. and reference '[34] Stan Development Team. Rstan: the r interface to stan, 2016. R package version 2.14.1.'
Experiment Setup No The paper mentions generating predictors using logistic regression and describes the model equations, but it does not provide specific experimental setup details such as hyperparameter values (e.g., learning rate, batch size, number of epochs, optimizer settings) used for training.