Reconciling Predictive and Statistical Parity: A Causal Approach

Authors: Drago Plecko, Elias Bareinboim

AAAI 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Finally, we demonstrate the importance of our findings on a real-world example. We now apply our approach in the context of criminal justice using the COMPAS dataset (Angwin et al. 2016), and demonstrate empirically the trade-off between SP and PP.
Researcher Affiliation Academia Drago Pleˇcko and Elias Bareinboim Department of Computer Science, Columbia University, New York, NY 10027 dp3144@columbia.edu, eb@cs.columbia.edu
Pseudocode Yes Algorithm 1: Business Necessity Cookbook
Open Source Code Yes see https://github.com/dplecko/sp-to-pp/blob/main/sp-pp-compas.R for source code
Open Datasets Yes We now apply Alg. 1 to the COMPAS dataset (Angwin et al. 2016), as described in the following example.
Dataset Splits No The paper does not provide specific training/validation/test dataset splits. It mentions using the COMPAS dataset and bootstrap repetitions for confidence intervals, but not how the data was partitioned into these specific sets for model training and evaluation reproducibility.
Hardware Specification No The paper does not provide any specific details about the hardware used to run the experiments, such as GPU/CPU models or specific computational resources.
Software Dependencies No The paper mentions using the "fairadapt package" and "random forests" but does not specify version numbers for any software dependencies.
Experiment Setup No The paper does not provide specific details about the experimental setup, such as hyperparameter values, learning rates, batch sizes, or other system-level training settings for the models used.