On the Impact of Algorithmic Recourse on Social Segregation

Authors: Ruijiang Gao, Himabindu Lakkaraju

ICML 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Extensive experimentation with real-world datasets demonstrates the efficacy of the proposed approaches.
Researcher Affiliation Academia 1University of Texas at Austin 2Harvard University. Correspondence to: Ruijiang Gao <ruijiang@utexas.edu>, Himabindu Lakkaraju <hlakkaraju@hbs.edu>.
Pseudocode Yes Algorithm 1 Explicit Balanced Recourse
Open Source Code Yes Our code is available at this URL.
Open Datasets Yes We use the real-world datasets German (Asuncion & Newman, 2007), Give-me-some-credit (GMC) (Kaggle, 2011), Law (Wightman, 1998) and COMPAS (Angwin et al., 2016)
Dataset Splits No The paper mentions training models ('For the synthetic dataset, the classifier is trained on empirical data using a Logistic Regression. For real-world datasets, the classifier is MLP with relu activations.') but does not provide specific training/validation/test splits (e.g., percentages or absolute counts) or references to predefined splits.
Hardware Specification No No specific hardware details (such as GPU or CPU models, memory, or cloud instance types) used for running the experiments are mentioned in the paper.
Software Dependencies No The paper mentions using specific models like 'MLP', 'VAE', 'Gaussian Mixture Network', and 'Logistic Regression' but does not provide specific version numbers for any software dependencies, libraries, or frameworks used (e.g., Python, PyTorch, scikit-learn versions).
Experiment Setup Yes λ is set as 0.01 in the experiments." (for Wachter) and "The step size of the growing hyperspheres is set as 0.2." (for GS) and "As we shall see in Section 6.4, the hyperparameter β can trade off recourse cost and social segregation." and "We include our results for ablation studies in Table 5. Here we conduct ablation studies on the hyperparameters of EBR and IBR on our synthetic examples in Section 6.2 to examine their effect on social segregation and recourse cost."