Active fairness auditing

Authors: Tom Yan, Chicheng Zhang

ICML 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Finally, in Appendix F, we empirically explore the performance of Algorithm 3 and active learning, and compare them with i.i.d sampling. As expected, our experiments confirm that under a fixed budget, Algorithm 3 is most effective at inducing a version space with a small µ-diameter, and can thus provide the strongest manipulation-proofness guarantee. and F. Experiments
Researcher Affiliation Academia 1Carnegie Mellon University 2University of Arizona.
Pseudocode Yes Algorithm 1 Minimax optimal deterministic auditing and Algorithm 3 Oracle-efficient Active Fairness Auditing
Open Source Code No The paper does not include an unambiguous statement that the authors are releasing their code for the work described, nor does it provide a direct link to a source-code repository.
Open Datasets Yes The first is COMPAS (Larson et al., 2016), where the two groups are defined to be Caucasian and non-Caucasian. And the second is the Student Performance Dataset
Dataset Splits No The paper mentions training a model on the COMPAS and Student Performance datasets but does not explicitly provide details about training, validation, or test dataset splits (e.g., percentages, sample counts, or cross-validation methodology).
Hardware Specification No The paper describes the experimental setup and results but does not provide any specific hardware details such as GPU/CPU models, memory, or cloud computing instance types used for running the experiments.
Software Dependencies No The paper mentions training a logistic regression model but does not provide specific software dependencies with version numbers, such as programming languages, libraries, or frameworks (e.g., Python, PyTorch, scikit-learn versions).
Experiment Setup No The paper mentions training a logistic regression model but does not provide specific experimental setup details such as hyperparameter values (e.g., learning rate, batch size, number of epochs) or optimizer settings.